4 * Copyright (C) 1991, 1992 Linus Torvalds
8 * 'fork.c' contains the help-routines for the 'fork' system call
9 * (see also entry.S and others).
10 * Fork is rather simple, once you get the hang of it, but the memory
11 * management can be a bitch. See 'mm/memory.c': 'copy_page_range()'
14 #include <linux/slab.h>
15 #include <linux/init.h>
16 #include <linux/unistd.h>
17 #include <linux/module.h>
18 #include <linux/vmalloc.h>
19 #include <linux/completion.h>
20 #include <linux/personality.h>
21 #include <linux/mempolicy.h>
22 #include <linux/sem.h>
23 #include <linux/file.h>
24 #include <linux/fdtable.h>
25 #include <linux/iocontext.h>
26 #include <linux/key.h>
27 #include <linux/binfmts.h>
28 #include <linux/mman.h>
29 #include <linux/mmu_notifier.h>
31 #include <linux/nsproxy.h>
32 #include <linux/capability.h>
33 #include <linux/cpu.h>
34 #include <linux/cgroup.h>
35 #include <linux/security.h>
36 #include <linux/hugetlb.h>
37 #include <linux/seccomp.h>
38 #include <linux/swap.h>
39 #include <linux/syscalls.h>
40 #include <linux/jiffies.h>
41 #include <linux/futex.h>
42 #include <linux/compat.h>
43 #include <linux/kthread.h>
44 #include <linux/task_io_accounting_ops.h>
45 #include <linux/rcupdate.h>
46 #include <linux/ptrace.h>
47 #include <linux/mount.h>
48 #include <linux/audit.h>
49 #include <linux/memcontrol.h>
50 #include <linux/ftrace.h>
51 #include <linux/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>
82 #define CREATE_TRACE_POINTS
83 #include <trace/events/task.h>
86 * Protected counters by write_lock_irq(&tasklist_lock)
88 unsigned long total_forks; /* Handle normal Linux uptimes. */
89 int nr_threads; /* The idle threads do not count.. */
91 int max_threads; /* tunable limit on nr_threads */
93 DEFINE_PER_CPU(unsigned long, process_counts) = 0;
95 __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
97 #ifdef CONFIG_PROVE_RCU
98 int lockdep_tasklist_lock_is_held(void)
100 return lockdep_is_held(&tasklist_lock);
102 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
103 #endif /* #ifdef CONFIG_PROVE_RCU */
105 int nr_processes(void)
110 for_each_possible_cpu(cpu)
111 total += per_cpu(process_counts, cpu);
116 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
117 static struct kmem_cache *task_struct_cachep;
119 static inline struct task_struct *alloc_task_struct_node(int node)
121 return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node);
124 void __weak arch_release_task_struct(struct task_struct *tsk) { }
126 static inline void free_task_struct(struct task_struct *tsk)
128 arch_release_task_struct(tsk);
129 kmem_cache_free(task_struct_cachep, tsk);
133 #ifndef CONFIG_ARCH_THREAD_INFO_ALLOCATOR
134 void __weak arch_release_thread_info(struct thread_info *ti) { }
137 * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a
138 * kmemcache based allocator.
140 # if THREAD_SIZE >= PAGE_SIZE
141 static struct thread_info *alloc_thread_info_node(struct task_struct *tsk,
144 struct page *page = alloc_pages_node(node, THREADINFO_GFP,
147 return page ? page_address(page) : NULL;
150 static inline void free_thread_info(struct thread_info *ti)
152 arch_release_thread_info(ti);
153 free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
156 static struct kmem_cache *thread_info_cache;
158 static struct thread_info *alloc_thread_info_node(struct task_struct *tsk,
161 return kmem_cache_alloc_node(thread_info_cache, THREADINFO_GFP, node);
164 static void free_thread_info(struct thread_info *ti)
166 arch_release_thread_info(ti);
167 kmem_cache_free(thread_info_cache, ti);
170 void thread_info_cache_init(void)
172 thread_info_cache = kmem_cache_create("thread_info", THREAD_SIZE,
173 THREAD_SIZE, 0, NULL);
174 BUG_ON(thread_info_cache == NULL);
179 /* SLAB cache for signal_struct structures (tsk->signal) */
180 static struct kmem_cache *signal_cachep;
182 /* SLAB cache for sighand_struct structures (tsk->sighand) */
183 struct kmem_cache *sighand_cachep;
185 /* SLAB cache for files_struct structures (tsk->files) */
186 struct kmem_cache *files_cachep;
188 /* SLAB cache for fs_struct structures (tsk->fs) */
189 struct kmem_cache *fs_cachep;
191 /* SLAB cache for vm_area_struct structures */
192 struct kmem_cache *vm_area_cachep;
194 /* SLAB cache for mm_struct structures (tsk->mm) */
195 static struct kmem_cache *mm_cachep;
197 static void account_kernel_stack(struct thread_info *ti, int account)
199 struct zone *zone = page_zone(virt_to_page(ti));
201 mod_zone_page_state(zone, NR_KERNEL_STACK, account);
204 void free_task(struct task_struct *tsk)
206 account_kernel_stack(tsk->stack, -1);
207 free_thread_info(tsk->stack);
208 rt_mutex_debug_task_free(tsk);
209 ftrace_graph_exit_task(tsk);
210 put_seccomp_filter(tsk);
211 free_task_struct(tsk);
213 EXPORT_SYMBOL(free_task);
215 static inline void free_signal_struct(struct signal_struct *sig)
217 taskstats_tgid_free(sig);
218 sched_autogroup_exit(sig);
219 kmem_cache_free(signal_cachep, sig);
222 static inline void put_signal_struct(struct signal_struct *sig)
224 if (atomic_dec_and_test(&sig->sigcnt))
225 free_signal_struct(sig);
228 void __put_task_struct(struct task_struct *tsk)
230 WARN_ON(!tsk->exit_state);
231 WARN_ON(atomic_read(&tsk->usage));
232 WARN_ON(tsk == current);
234 security_task_free(tsk);
236 delayacct_tsk_free(tsk);
237 put_signal_struct(tsk->signal);
239 if (!profile_handoff_task(tsk))
242 EXPORT_SYMBOL_GPL(__put_task_struct);
244 void __init __weak arch_task_cache_init(void) { }
246 void __init fork_init(unsigned long mempages)
248 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
249 #ifndef ARCH_MIN_TASKALIGN
250 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
252 /* create a slab on which task_structs can be allocated */
254 kmem_cache_create("task_struct", sizeof(struct task_struct),
255 ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL);
258 /* do the arch specific task caches init */
259 arch_task_cache_init();
262 * The default maximum number of threads is set to a safe
263 * value: the thread structures can take up at most half
266 max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
269 * we need to allow at least 20 threads to boot a system
271 if (max_threads < 20)
274 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
275 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
276 init_task.signal->rlim[RLIMIT_SIGPENDING] =
277 init_task.signal->rlim[RLIMIT_NPROC];
280 int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst,
281 struct task_struct *src)
287 static struct task_struct *dup_task_struct(struct task_struct *orig)
289 struct task_struct *tsk;
290 struct thread_info *ti;
291 unsigned long *stackend;
292 int node = tsk_fork_get_node(orig);
295 prepare_to_copy(orig);
297 tsk = alloc_task_struct_node(node);
301 ti = alloc_thread_info_node(tsk, node);
303 free_task_struct(tsk);
307 err = arch_dup_task_struct(tsk, orig);
313 setup_thread_stack(tsk, orig);
314 clear_user_return_notifier(tsk);
315 clear_tsk_need_resched(tsk);
316 stackend = end_of_stack(tsk);
317 *stackend = STACK_END_MAGIC; /* for overflow detection */
319 #ifdef CONFIG_CC_STACKPROTECTOR
320 tsk->stack_canary = get_random_int();
324 * One for us, one for whoever does the "release_task()" (usually
327 atomic_set(&tsk->usage, 2);
328 #ifdef CONFIG_BLK_DEV_IO_TRACE
331 tsk->splice_pipe = NULL;
333 account_kernel_stack(ti, 1);
338 free_thread_info(ti);
339 free_task_struct(tsk);
344 static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
346 struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
347 struct rb_node **rb_link, *rb_parent;
349 unsigned long charge;
350 struct mempolicy *pol;
352 down_write(&oldmm->mmap_sem);
353 flush_cache_dup_mm(oldmm);
355 * Not linked in yet - no deadlock potential:
357 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
361 mm->mmap_cache = NULL;
362 mm->free_area_cache = oldmm->mmap_base;
363 mm->cached_hole_size = ~0UL;
365 cpumask_clear(mm_cpumask(mm));
367 rb_link = &mm->mm_rb.rb_node;
370 retval = ksm_fork(mm, oldmm);
373 retval = khugepaged_fork(mm, oldmm);
378 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
381 if (mpnt->vm_flags & VM_DONTCOPY) {
382 long pages = vma_pages(mpnt);
383 mm->total_vm -= pages;
384 vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
389 if (mpnt->vm_flags & VM_ACCOUNT) {
390 unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
391 if (security_vm_enough_memory_mm(oldmm, len)) /* sic */
395 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
399 INIT_LIST_HEAD(&tmp->anon_vma_chain);
400 pol = mpol_dup(vma_policy(mpnt));
401 retval = PTR_ERR(pol);
403 goto fail_nomem_policy;
404 vma_set_policy(tmp, pol);
406 if (anon_vma_fork(tmp, mpnt))
407 goto fail_nomem_anon_vma_fork;
408 tmp->vm_flags &= ~VM_LOCKED;
409 tmp->vm_next = tmp->vm_prev = NULL;
412 struct inode *inode = file->f_path.dentry->d_inode;
413 struct address_space *mapping = file->f_mapping;
416 if (tmp->vm_flags & VM_DENYWRITE)
417 atomic_dec(&inode->i_writecount);
418 mutex_lock(&mapping->i_mmap_mutex);
419 if (tmp->vm_flags & VM_SHARED)
420 mapping->i_mmap_writable++;
421 flush_dcache_mmap_lock(mapping);
422 /* insert tmp into the share list, just after mpnt */
423 vma_prio_tree_add(tmp, mpnt);
424 flush_dcache_mmap_unlock(mapping);
425 mutex_unlock(&mapping->i_mmap_mutex);
429 * Clear hugetlb-related page reserves for children. This only
430 * affects MAP_PRIVATE mappings. Faults generated by the child
431 * are not guaranteed to succeed, even if read-only
433 if (is_vm_hugetlb_page(tmp))
434 reset_vma_resv_huge_pages(tmp);
437 * Link in the new vma and copy the page table entries.
440 pprev = &tmp->vm_next;
444 __vma_link_rb(mm, tmp, rb_link, rb_parent);
445 rb_link = &tmp->vm_rb.rb_right;
446 rb_parent = &tmp->vm_rb;
449 retval = copy_page_range(mm, oldmm, mpnt);
451 if (tmp->vm_ops && tmp->vm_ops->open)
452 tmp->vm_ops->open(tmp);
457 /* a new mm has just been created */
458 arch_dup_mmap(oldmm, mm);
461 up_write(&mm->mmap_sem);
463 up_write(&oldmm->mmap_sem);
465 fail_nomem_anon_vma_fork:
468 kmem_cache_free(vm_area_cachep, tmp);
471 vm_unacct_memory(charge);
475 static inline int mm_alloc_pgd(struct mm_struct *mm)
477 mm->pgd = pgd_alloc(mm);
478 if (unlikely(!mm->pgd))
483 static inline void mm_free_pgd(struct mm_struct *mm)
485 pgd_free(mm, mm->pgd);
488 #define dup_mmap(mm, oldmm) (0)
489 #define mm_alloc_pgd(mm) (0)
490 #define mm_free_pgd(mm)
491 #endif /* CONFIG_MMU */
493 __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
495 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
496 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
498 static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
500 static int __init coredump_filter_setup(char *s)
502 default_dump_filter =
503 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
504 MMF_DUMP_FILTER_MASK;
508 __setup("coredump_filter=", coredump_filter_setup);
510 #include <linux/init_task.h>
512 static void mm_init_aio(struct mm_struct *mm)
515 spin_lock_init(&mm->ioctx_lock);
516 INIT_HLIST_HEAD(&mm->ioctx_list);
520 static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p)
522 atomic_set(&mm->mm_users, 1);
523 atomic_set(&mm->mm_count, 1);
524 init_rwsem(&mm->mmap_sem);
525 INIT_LIST_HEAD(&mm->mmlist);
526 mm->flags = (current->mm) ?
527 (current->mm->flags & MMF_INIT_MASK) : default_dump_filter;
528 mm->core_state = NULL;
530 memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
531 spin_lock_init(&mm->page_table_lock);
532 mm->free_area_cache = TASK_UNMAPPED_BASE;
533 mm->cached_hole_size = ~0UL;
535 mm_init_owner(mm, p);
537 if (likely(!mm_alloc_pgd(mm))) {
539 mmu_notifier_mm_init(mm);
547 static void check_mm(struct mm_struct *mm)
551 for (i = 0; i < NR_MM_COUNTERS; i++) {
552 long x = atomic_long_read(&mm->rss_stat.count[i]);
555 printk(KERN_ALERT "BUG: Bad rss-counter state "
556 "mm:%p idx:%d val:%ld\n", mm, i, x);
559 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
560 VM_BUG_ON(mm->pmd_huge_pte);
565 * Allocate and initialize an mm_struct.
567 struct mm_struct *mm_alloc(void)
569 struct mm_struct *mm;
575 memset(mm, 0, sizeof(*mm));
577 return mm_init(mm, current);
581 * Called when the last reference to the mm
582 * is dropped: either by a lazy thread or by
583 * mmput. Free the page directory and the mm.
585 void __mmdrop(struct mm_struct *mm)
587 BUG_ON(mm == &init_mm);
590 mmu_notifier_mm_destroy(mm);
594 EXPORT_SYMBOL_GPL(__mmdrop);
597 * Decrement the use count and release all resources for an mm.
599 void mmput(struct mm_struct *mm)
603 if (atomic_dec_and_test(&mm->mm_users)) {
606 khugepaged_exit(mm); /* must run before exit_mmap */
608 set_mm_exe_file(mm, NULL);
609 if (!list_empty(&mm->mmlist)) {
610 spin_lock(&mmlist_lock);
611 list_del(&mm->mmlist);
612 spin_unlock(&mmlist_lock);
616 module_put(mm->binfmt->module);
620 EXPORT_SYMBOL_GPL(mmput);
623 * We added or removed a vma mapping the executable. The vmas are only mapped
624 * during exec and are not mapped with the mmap system call.
625 * Callers must hold down_write() on the mm's mmap_sem for these
627 void added_exe_file_vma(struct mm_struct *mm)
629 mm->num_exe_file_vmas++;
632 void removed_exe_file_vma(struct mm_struct *mm)
634 mm->num_exe_file_vmas--;
635 if ((mm->num_exe_file_vmas == 0) && mm->exe_file) {
642 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
645 get_file(new_exe_file);
648 mm->exe_file = new_exe_file;
649 mm->num_exe_file_vmas = 0;
652 struct file *get_mm_exe_file(struct mm_struct *mm)
654 struct file *exe_file;
656 /* We need mmap_sem to protect against races with removal of
657 * VM_EXECUTABLE vmas */
658 down_read(&mm->mmap_sem);
659 exe_file = mm->exe_file;
662 up_read(&mm->mmap_sem);
666 static void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
668 /* It's safe to write the exe_file pointer without exe_file_lock because
669 * this is called during fork when the task is not yet in /proc */
670 newmm->exe_file = get_mm_exe_file(oldmm);
674 * get_task_mm - acquire a reference to the task's mm
676 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
677 * this kernel workthread has transiently adopted a user mm with use_mm,
678 * to do its AIO) is not set and if so returns a reference to it, after
679 * bumping up the use count. User must release the mm via mmput()
680 * after use. Typically used by /proc and ptrace.
682 struct mm_struct *get_task_mm(struct task_struct *task)
684 struct mm_struct *mm;
689 if (task->flags & PF_KTHREAD)
692 atomic_inc(&mm->mm_users);
697 EXPORT_SYMBOL_GPL(get_task_mm);
699 struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
701 struct mm_struct *mm;
704 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
708 mm = get_task_mm(task);
709 if (mm && mm != current->mm &&
710 !ptrace_may_access(task, mode)) {
712 mm = ERR_PTR(-EACCES);
714 mutex_unlock(&task->signal->cred_guard_mutex);
719 static void complete_vfork_done(struct task_struct *tsk)
721 struct completion *vfork;
724 vfork = tsk->vfork_done;
726 tsk->vfork_done = NULL;
732 static int wait_for_vfork_done(struct task_struct *child,
733 struct completion *vfork)
737 freezer_do_not_count();
738 killed = wait_for_completion_killable(vfork);
743 child->vfork_done = NULL;
747 put_task_struct(child);
751 /* Please note the differences between mmput and mm_release.
752 * mmput is called whenever we stop holding onto a mm_struct,
753 * error success whatever.
755 * mm_release is called after a mm_struct has been removed
756 * from the current process.
758 * This difference is important for error handling, when we
759 * only half set up a mm_struct for a new process and need to restore
760 * the old one. Because we mmput the new mm_struct before
761 * restoring the old one. . .
762 * Eric Biederman 10 January 1998
764 void mm_release(struct task_struct *tsk, struct mm_struct *mm)
766 /* Get rid of any futexes when releasing the mm */
768 if (unlikely(tsk->robust_list)) {
769 exit_robust_list(tsk);
770 tsk->robust_list = NULL;
773 if (unlikely(tsk->compat_robust_list)) {
774 compat_exit_robust_list(tsk);
775 tsk->compat_robust_list = NULL;
778 if (unlikely(!list_empty(&tsk->pi_state_list)))
779 exit_pi_state_list(tsk);
782 /* Get rid of any cached register state */
783 deactivate_mm(tsk, mm);
786 complete_vfork_done(tsk);
789 * If we're exiting normally, clear a user-space tid field if
790 * requested. We leave this alone when dying by signal, to leave
791 * the value intact in a core dump, and to save the unnecessary
792 * trouble, say, a killed vfork parent shouldn't touch this mm.
793 * Userland only wants this done for a sys_exit.
795 if (tsk->clear_child_tid) {
796 if (!(tsk->flags & PF_SIGNALED) &&
797 atomic_read(&mm->mm_users) > 1) {
799 * We don't check the error code - if userspace has
800 * not set up a proper pointer then tough luck.
802 put_user(0, tsk->clear_child_tid);
803 sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
806 tsk->clear_child_tid = NULL;
811 * Allocate a new mm structure and copy contents from the
812 * mm structure of the passed in task structure.
814 struct mm_struct *dup_mm(struct task_struct *tsk)
816 struct mm_struct *mm, *oldmm = current->mm;
826 memcpy(mm, oldmm, sizeof(*mm));
829 /* Initializing for Swap token stuff */
830 mm->token_priority = 0;
831 mm->last_interval = 0;
833 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
834 mm->pmd_huge_pte = NULL;
837 if (!mm_init(mm, tsk))
840 if (init_new_context(tsk, mm))
843 dup_mm_exe_file(oldmm, mm);
845 err = dup_mmap(mm, oldmm);
849 mm->hiwater_rss = get_mm_rss(mm);
850 mm->hiwater_vm = mm->total_vm;
852 if (mm->binfmt && !try_module_get(mm->binfmt->module))
858 /* don't put binfmt in mmput, we haven't got module yet */
867 * If init_new_context() failed, we cannot use mmput() to free the mm
868 * because it calls destroy_context()
875 static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
877 struct mm_struct *mm, *oldmm;
880 tsk->min_flt = tsk->maj_flt = 0;
881 tsk->nvcsw = tsk->nivcsw = 0;
882 #ifdef CONFIG_DETECT_HUNG_TASK
883 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
887 tsk->active_mm = NULL;
890 * Are we cloning a kernel thread?
892 * We need to steal a active VM for that..
898 if (clone_flags & CLONE_VM) {
899 atomic_inc(&oldmm->mm_users);
910 /* Initializing for Swap token stuff */
911 mm->token_priority = 0;
912 mm->last_interval = 0;
922 static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
924 struct fs_struct *fs = current->fs;
925 if (clone_flags & CLONE_FS) {
926 /* tsk->fs is already what we want */
927 spin_lock(&fs->lock);
929 spin_unlock(&fs->lock);
933 spin_unlock(&fs->lock);
936 tsk->fs = copy_fs_struct(fs);
942 static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
944 struct files_struct *oldf, *newf;
948 * A background process may not have any files ...
950 oldf = current->files;
954 if (clone_flags & CLONE_FILES) {
955 atomic_inc(&oldf->count);
959 newf = dup_fd(oldf, &error);
969 static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
972 struct io_context *ioc = current->io_context;
973 struct io_context *new_ioc;
978 * Share io context with parent, if CLONE_IO is set
980 if (clone_flags & CLONE_IO) {
981 tsk->io_context = ioc_task_link(ioc);
982 if (unlikely(!tsk->io_context))
984 } else if (ioprio_valid(ioc->ioprio)) {
985 new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE);
986 if (unlikely(!new_ioc))
989 new_ioc->ioprio = ioc->ioprio;
990 put_io_context(new_ioc);
996 static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
998 struct sighand_struct *sig;
1000 if (clone_flags & CLONE_SIGHAND) {
1001 atomic_inc(¤t->sighand->count);
1004 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
1005 rcu_assign_pointer(tsk->sighand, sig);
1008 atomic_set(&sig->count, 1);
1009 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
1013 void __cleanup_sighand(struct sighand_struct *sighand)
1015 if (atomic_dec_and_test(&sighand->count)) {
1016 signalfd_cleanup(sighand);
1017 kmem_cache_free(sighand_cachep, sighand);
1023 * Initialize POSIX timer handling for a thread group.
1025 static void posix_cpu_timers_init_group(struct signal_struct *sig)
1027 unsigned long cpu_limit;
1029 /* Thread group counters. */
1030 thread_group_cputime_init(sig);
1032 cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
1033 if (cpu_limit != RLIM_INFINITY) {
1034 sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit);
1035 sig->cputimer.running = 1;
1038 /* The timer lists. */
1039 INIT_LIST_HEAD(&sig->cpu_timers[0]);
1040 INIT_LIST_HEAD(&sig->cpu_timers[1]);
1041 INIT_LIST_HEAD(&sig->cpu_timers[2]);
1044 static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
1046 struct signal_struct *sig;
1048 if (clone_flags & CLONE_THREAD)
1051 sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
1056 sig->nr_threads = 1;
1057 atomic_set(&sig->live, 1);
1058 atomic_set(&sig->sigcnt, 1);
1059 init_waitqueue_head(&sig->wait_chldexit);
1060 if (clone_flags & CLONE_NEWPID)
1061 sig->flags |= SIGNAL_UNKILLABLE;
1062 sig->curr_target = tsk;
1063 init_sigpending(&sig->shared_pending);
1064 INIT_LIST_HEAD(&sig->posix_timers);
1066 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
1067 sig->real_timer.function = it_real_fn;
1069 task_lock(current->group_leader);
1070 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
1071 task_unlock(current->group_leader);
1073 posix_cpu_timers_init_group(sig);
1075 tty_audit_fork(sig);
1076 sched_autogroup_fork(sig);
1078 #ifdef CONFIG_CGROUPS
1079 init_rwsem(&sig->group_rwsem);
1082 sig->oom_adj = current->signal->oom_adj;
1083 sig->oom_score_adj = current->signal->oom_score_adj;
1084 sig->oom_score_adj_min = current->signal->oom_score_adj_min;
1086 sig->has_child_subreaper = current->signal->has_child_subreaper ||
1087 current->signal->is_child_subreaper;
1089 mutex_init(&sig->cred_guard_mutex);
1094 static void copy_flags(unsigned long clone_flags, struct task_struct *p)
1096 unsigned long new_flags = p->flags;
1098 new_flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER);
1099 new_flags |= PF_FORKNOEXEC;
1100 p->flags = new_flags;
1103 SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
1105 current->clear_child_tid = tidptr;
1107 return task_pid_vnr(current);
1110 static void rt_mutex_init_task(struct task_struct *p)
1112 raw_spin_lock_init(&p->pi_lock);
1113 #ifdef CONFIG_RT_MUTEXES
1114 plist_head_init(&p->pi_waiters);
1115 p->pi_blocked_on = NULL;
1119 #ifdef CONFIG_MM_OWNER
1120 void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
1124 #endif /* CONFIG_MM_OWNER */
1127 * Initialize POSIX timer handling for a single task.
1129 static void posix_cpu_timers_init(struct task_struct *tsk)
1131 tsk->cputime_expires.prof_exp = 0;
1132 tsk->cputime_expires.virt_exp = 0;
1133 tsk->cputime_expires.sched_exp = 0;
1134 INIT_LIST_HEAD(&tsk->cpu_timers[0]);
1135 INIT_LIST_HEAD(&tsk->cpu_timers[1]);
1136 INIT_LIST_HEAD(&tsk->cpu_timers[2]);
1140 * This creates a new process as a copy of the old one,
1141 * but does not actually start it yet.
1143 * It copies the registers, and all the appropriate
1144 * parts of the process environment (as per the clone
1145 * flags). The actual kick-off is left to the caller.
1147 static struct task_struct *copy_process(unsigned long clone_flags,
1148 unsigned long stack_start,
1149 struct pt_regs *regs,
1150 unsigned long stack_size,
1151 int __user *child_tidptr,
1156 struct task_struct *p;
1157 int cgroup_callbacks_done = 0;
1159 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
1160 return ERR_PTR(-EINVAL);
1163 * Thread groups must share signals as well, and detached threads
1164 * can only be started up within the thread group.
1166 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1167 return ERR_PTR(-EINVAL);
1170 * Shared signal handlers imply shared VM. By way of the above,
1171 * thread groups also imply shared VM. Blocking this case allows
1172 * for various simplifications in other code.
1174 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1175 return ERR_PTR(-EINVAL);
1178 * Siblings of global init remain as zombies on exit since they are
1179 * not reaped by their parent (swapper). To solve this and to avoid
1180 * multi-rooted process trees, prevent global and container-inits
1181 * from creating siblings.
1183 if ((clone_flags & CLONE_PARENT) &&
1184 current->signal->flags & SIGNAL_UNKILLABLE)
1185 return ERR_PTR(-EINVAL);
1187 retval = security_task_create(clone_flags);
1192 p = dup_task_struct(current);
1196 ftrace_graph_init_task(p);
1197 get_seccomp_filter(p);
1199 rt_mutex_init_task(p);
1201 #ifdef CONFIG_PROVE_LOCKING
1202 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1203 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1206 if (atomic_read(&p->real_cred->user->processes) >=
1207 task_rlimit(p, RLIMIT_NPROC)) {
1208 if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
1209 p->real_cred->user != INIT_USER)
1212 current->flags &= ~PF_NPROC_EXCEEDED;
1214 retval = copy_creds(p, clone_flags);
1219 * If multiple threads are within copy_process(), then this check
1220 * triggers too late. This doesn't hurt, the check is only there
1221 * to stop root fork bombs.
1224 if (nr_threads >= max_threads)
1225 goto bad_fork_cleanup_count;
1227 if (!try_module_get(task_thread_info(p)->exec_domain->module))
1228 goto bad_fork_cleanup_count;
1231 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
1232 copy_flags(clone_flags, p);
1233 INIT_LIST_HEAD(&p->children);
1234 INIT_LIST_HEAD(&p->sibling);
1235 rcu_copy_process(p);
1236 p->vfork_done = NULL;
1237 spin_lock_init(&p->alloc_lock);
1239 init_sigpending(&p->pending);
1241 p->utime = p->stime = p->gtime = 0;
1242 p->utimescaled = p->stimescaled = 0;
1243 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
1244 p->prev_utime = p->prev_stime = 0;
1246 #if defined(SPLIT_RSS_COUNTING)
1247 memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1250 p->default_timer_slack_ns = current->timer_slack_ns;
1252 task_io_accounting_init(&p->ioac);
1253 acct_clear_integrals(p);
1255 posix_cpu_timers_init(p);
1257 do_posix_clock_monotonic_gettime(&p->start_time);
1258 p->real_start_time = p->start_time;
1259 monotonic_to_bootbased(&p->real_start_time);
1260 p->io_context = NULL;
1261 p->audit_context = NULL;
1262 if (clone_flags & CLONE_THREAD)
1263 threadgroup_change_begin(current);
1266 p->mempolicy = mpol_dup(p->mempolicy);
1267 if (IS_ERR(p->mempolicy)) {
1268 retval = PTR_ERR(p->mempolicy);
1269 p->mempolicy = NULL;
1270 goto bad_fork_cleanup_cgroup;
1272 mpol_fix_fork_child_flag(p);
1274 #ifdef CONFIG_CPUSETS
1275 p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
1276 p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
1277 seqcount_init(&p->mems_allowed_seq);
1279 #ifdef CONFIG_TRACE_IRQFLAGS
1281 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1282 p->hardirqs_enabled = 1;
1284 p->hardirqs_enabled = 0;
1286 p->hardirq_enable_ip = 0;
1287 p->hardirq_enable_event = 0;
1288 p->hardirq_disable_ip = _THIS_IP_;
1289 p->hardirq_disable_event = 0;
1290 p->softirqs_enabled = 1;
1291 p->softirq_enable_ip = _THIS_IP_;
1292 p->softirq_enable_event = 0;
1293 p->softirq_disable_ip = 0;
1294 p->softirq_disable_event = 0;
1295 p->hardirq_context = 0;
1296 p->softirq_context = 0;
1298 #ifdef CONFIG_LOCKDEP
1299 p->lockdep_depth = 0; /* no locks held yet */
1300 p->curr_chain_key = 0;
1301 p->lockdep_recursion = 0;
1304 #ifdef CONFIG_DEBUG_MUTEXES
1305 p->blocked_on = NULL; /* not blocked yet */
1307 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
1308 p->memcg_batch.do_batch = 0;
1309 p->memcg_batch.memcg = NULL;
1312 /* Perform scheduler related setup. Assign this task to a CPU. */
1315 retval = perf_event_init_task(p);
1317 goto bad_fork_cleanup_policy;
1318 retval = audit_alloc(p);
1320 goto bad_fork_cleanup_policy;
1321 /* copy all the process information */
1322 retval = copy_semundo(clone_flags, p);
1324 goto bad_fork_cleanup_audit;
1325 retval = copy_files(clone_flags, p);
1327 goto bad_fork_cleanup_semundo;
1328 retval = copy_fs(clone_flags, p);
1330 goto bad_fork_cleanup_files;
1331 retval = copy_sighand(clone_flags, p);
1333 goto bad_fork_cleanup_fs;
1334 retval = copy_signal(clone_flags, p);
1336 goto bad_fork_cleanup_sighand;
1337 retval = copy_mm(clone_flags, p);
1339 goto bad_fork_cleanup_signal;
1340 retval = copy_namespaces(clone_flags, p);
1342 goto bad_fork_cleanup_mm;
1343 retval = copy_io(clone_flags, p);
1345 goto bad_fork_cleanup_namespaces;
1346 retval = copy_thread(clone_flags, stack_start, stack_size, p, regs);
1348 goto bad_fork_cleanup_io;
1350 if (pid != &init_struct_pid) {
1352 pid = alloc_pid(p->nsproxy->pid_ns);
1354 goto bad_fork_cleanup_io;
1357 p->pid = pid_nr(pid);
1359 if (clone_flags & CLONE_THREAD)
1360 p->tgid = current->tgid;
1362 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1364 * Clear TID on mm_release()?
1366 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL;
1371 p->robust_list = NULL;
1372 #ifdef CONFIG_COMPAT
1373 p->compat_robust_list = NULL;
1375 INIT_LIST_HEAD(&p->pi_state_list);
1376 p->pi_state_cache = NULL;
1379 * sigaltstack should be cleared when sharing the same VM
1381 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1382 p->sas_ss_sp = p->sas_ss_size = 0;
1385 * Syscall tracing and stepping should be turned off in the
1386 * child regardless of CLONE_PTRACE.
1388 user_disable_single_step(p);
1389 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
1390 #ifdef TIF_SYSCALL_EMU
1391 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1393 clear_all_latency_tracing(p);
1395 /* ok, now we should be set up.. */
1396 if (clone_flags & CLONE_THREAD)
1397 p->exit_signal = -1;
1398 else if (clone_flags & CLONE_PARENT)
1399 p->exit_signal = current->group_leader->exit_signal;
1401 p->exit_signal = (clone_flags & CSIGNAL);
1403 p->pdeath_signal = 0;
1407 p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
1408 p->dirty_paused_when = 0;
1411 * Ok, make it visible to the rest of the system.
1412 * We dont wake it up yet.
1414 p->group_leader = p;
1415 INIT_LIST_HEAD(&p->thread_group);
1417 /* Now that the task is set up, run cgroup callbacks if
1418 * necessary. We need to run them before the task is visible
1419 * on the tasklist. */
1420 cgroup_fork_callbacks(p);
1421 cgroup_callbacks_done = 1;
1423 /* Need tasklist lock for parent etc handling! */
1424 write_lock_irq(&tasklist_lock);
1426 /* CLONE_PARENT re-uses the old parent */
1427 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
1428 p->real_parent = current->real_parent;
1429 p->parent_exec_id = current->parent_exec_id;
1431 p->real_parent = current;
1432 p->parent_exec_id = current->self_exec_id;
1435 spin_lock(¤t->sighand->siglock);
1438 * Process group and session signals need to be delivered to just the
1439 * parent before the fork or both the parent and the child after the
1440 * fork. Restart if a signal comes in before we add the new process to
1441 * it's process group.
1442 * A fatal signal pending means that current will exit, so the new
1443 * thread can't slip out of an OOM kill (or normal SIGKILL).
1445 recalc_sigpending();
1446 if (signal_pending(current)) {
1447 spin_unlock(¤t->sighand->siglock);
1448 write_unlock_irq(&tasklist_lock);
1449 retval = -ERESTARTNOINTR;
1450 goto bad_fork_free_pid;
1453 if (clone_flags & CLONE_THREAD) {
1454 current->signal->nr_threads++;
1455 atomic_inc(¤t->signal->live);
1456 atomic_inc(¤t->signal->sigcnt);
1457 p->group_leader = current->group_leader;
1458 list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
1461 if (likely(p->pid)) {
1462 ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
1464 if (thread_group_leader(p)) {
1465 if (is_child_reaper(pid))
1466 p->nsproxy->pid_ns->child_reaper = p;
1468 p->signal->leader_pid = pid;
1469 p->signal->tty = tty_kref_get(current->signal->tty);
1470 attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
1471 attach_pid(p, PIDTYPE_SID, task_session(current));
1472 list_add_tail(&p->sibling, &p->real_parent->children);
1473 list_add_tail_rcu(&p->tasks, &init_task.tasks);
1474 __this_cpu_inc(process_counts);
1476 attach_pid(p, PIDTYPE_PID, pid);
1481 spin_unlock(¤t->sighand->siglock);
1482 write_unlock_irq(&tasklist_lock);
1483 proc_fork_connector(p);
1484 cgroup_post_fork(p);
1485 if (clone_flags & CLONE_THREAD)
1486 threadgroup_change_end(current);
1489 trace_task_newtask(p, clone_flags);
1494 if (pid != &init_struct_pid)
1496 bad_fork_cleanup_io:
1499 bad_fork_cleanup_namespaces:
1500 if (unlikely(clone_flags & CLONE_NEWPID))
1501 pid_ns_release_proc(p->nsproxy->pid_ns);
1502 exit_task_namespaces(p);
1503 bad_fork_cleanup_mm:
1506 bad_fork_cleanup_signal:
1507 if (!(clone_flags & CLONE_THREAD))
1508 free_signal_struct(p->signal);
1509 bad_fork_cleanup_sighand:
1510 __cleanup_sighand(p->sighand);
1511 bad_fork_cleanup_fs:
1512 exit_fs(p); /* blocking */
1513 bad_fork_cleanup_files:
1514 exit_files(p); /* blocking */
1515 bad_fork_cleanup_semundo:
1517 bad_fork_cleanup_audit:
1519 bad_fork_cleanup_policy:
1520 perf_event_free_task(p);
1522 mpol_put(p->mempolicy);
1523 bad_fork_cleanup_cgroup:
1525 if (clone_flags & CLONE_THREAD)
1526 threadgroup_change_end(current);
1527 cgroup_exit(p, cgroup_callbacks_done);
1528 delayacct_tsk_free(p);
1529 module_put(task_thread_info(p)->exec_domain->module);
1530 bad_fork_cleanup_count:
1531 atomic_dec(&p->cred->user->processes);
1536 return ERR_PTR(retval);
1539 noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
1541 memset(regs, 0, sizeof(struct pt_regs));
1545 static inline void init_idle_pids(struct pid_link *links)
1549 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1550 INIT_HLIST_NODE(&links[type].node); /* not really needed */
1551 links[type].pid = &init_struct_pid;
1555 struct task_struct * __cpuinit fork_idle(int cpu)
1557 struct task_struct *task;
1558 struct pt_regs regs;
1560 task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL,
1561 &init_struct_pid, 0);
1562 if (!IS_ERR(task)) {
1563 init_idle_pids(task->pids);
1564 init_idle(task, cpu);
1571 * Ok, this is the main fork-routine.
1573 * It copies the process, and if successful kick-starts
1574 * it and waits for it to finish using the VM if required.
1576 long do_fork(unsigned long clone_flags,
1577 unsigned long stack_start,
1578 struct pt_regs *regs,
1579 unsigned long stack_size,
1580 int __user *parent_tidptr,
1581 int __user *child_tidptr)
1583 struct task_struct *p;
1588 * Do some preliminary argument and permissions checking before we
1589 * actually start allocating stuff
1591 if (clone_flags & CLONE_NEWUSER) {
1592 if (clone_flags & CLONE_THREAD)
1594 /* hopefully this check will go away when userns support is
1597 if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) ||
1598 !capable(CAP_SETGID))
1603 * Determine whether and which event to report to ptracer. When
1604 * called from kernel_thread or CLONE_UNTRACED is explicitly
1605 * requested, no event is reported; otherwise, report if the event
1606 * for the type of forking is enabled.
1608 if (likely(user_mode(regs)) && !(clone_flags & CLONE_UNTRACED)) {
1609 if (clone_flags & CLONE_VFORK)
1610 trace = PTRACE_EVENT_VFORK;
1611 else if ((clone_flags & CSIGNAL) != SIGCHLD)
1612 trace = PTRACE_EVENT_CLONE;
1614 trace = PTRACE_EVENT_FORK;
1616 if (likely(!ptrace_event_enabled(current, trace)))
1620 p = copy_process(clone_flags, stack_start, regs, stack_size,
1621 child_tidptr, NULL, trace);
1623 * Do this prior waking up the new thread - the thread pointer
1624 * might get invalid after that point, if the thread exits quickly.
1627 struct completion vfork;
1629 trace_sched_process_fork(current, p);
1631 nr = task_pid_vnr(p);
1633 if (clone_flags & CLONE_PARENT_SETTID)
1634 put_user(nr, parent_tidptr);
1636 if (clone_flags & CLONE_VFORK) {
1637 p->vfork_done = &vfork;
1638 init_completion(&vfork);
1642 wake_up_new_task(p);
1644 /* forking complete and child started to run, tell ptracer */
1645 if (unlikely(trace))
1646 ptrace_event(trace, nr);
1648 if (clone_flags & CLONE_VFORK) {
1649 if (!wait_for_vfork_done(p, &vfork))
1650 ptrace_event(PTRACE_EVENT_VFORK_DONE, nr);
1658 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1659 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1662 static void sighand_ctor(void *data)
1664 struct sighand_struct *sighand = data;
1666 spin_lock_init(&sighand->siglock);
1667 init_waitqueue_head(&sighand->signalfd_wqh);
1670 void __init proc_caches_init(void)
1672 sighand_cachep = kmem_cache_create("sighand_cache",
1673 sizeof(struct sighand_struct), 0,
1674 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
1675 SLAB_NOTRACK, sighand_ctor);
1676 signal_cachep = kmem_cache_create("signal_cache",
1677 sizeof(struct signal_struct), 0,
1678 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1679 files_cachep = kmem_cache_create("files_cache",
1680 sizeof(struct files_struct), 0,
1681 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1682 fs_cachep = kmem_cache_create("fs_cache",
1683 sizeof(struct fs_struct), 0,
1684 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1686 * FIXME! The "sizeof(struct mm_struct)" currently includes the
1687 * whole struct cpumask for the OFFSTACK case. We could change
1688 * this to *only* allocate as much of it as required by the
1689 * maximum number of CPU's we can ever have. The cpumask_allocation
1690 * is at the end of the structure, exactly for that reason.
1692 mm_cachep = kmem_cache_create("mm_struct",
1693 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
1694 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1695 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
1697 nsproxy_cache_init();
1701 * Check constraints on flags passed to the unshare system call.
1703 static int check_unshare_flags(unsigned long unshare_flags)
1705 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
1706 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
1707 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET))
1710 * Not implemented, but pretend it works if there is nothing to
1711 * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND
1712 * needs to unshare vm.
1714 if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
1715 /* FIXME: get_task_mm() increments ->mm_users */
1716 if (atomic_read(¤t->mm->mm_users) > 1)
1724 * Unshare the filesystem structure if it is being shared
1726 static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1728 struct fs_struct *fs = current->fs;
1730 if (!(unshare_flags & CLONE_FS) || !fs)
1733 /* don't need lock here; in the worst case we'll do useless copy */
1737 *new_fsp = copy_fs_struct(fs);
1745 * Unshare file descriptor table if it is being shared
1747 static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1749 struct files_struct *fd = current->files;
1752 if ((unshare_flags & CLONE_FILES) &&
1753 (fd && atomic_read(&fd->count) > 1)) {
1754 *new_fdp = dup_fd(fd, &error);
1763 * unshare allows a process to 'unshare' part of the process
1764 * context which was originally shared using clone. copy_*
1765 * functions used by do_fork() cannot be used here directly
1766 * because they modify an inactive task_struct that is being
1767 * constructed. Here we are modifying the current, active,
1770 SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
1772 struct fs_struct *fs, *new_fs = NULL;
1773 struct files_struct *fd, *new_fd = NULL;
1774 struct nsproxy *new_nsproxy = NULL;
1778 err = check_unshare_flags(unshare_flags);
1780 goto bad_unshare_out;
1783 * If unsharing namespace, must also unshare filesystem information.
1785 if (unshare_flags & CLONE_NEWNS)
1786 unshare_flags |= CLONE_FS;
1788 * CLONE_NEWIPC must also detach from the undolist: after switching
1789 * to a new ipc namespace, the semaphore arrays from the old
1790 * namespace are unreachable.
1792 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
1794 err = unshare_fs(unshare_flags, &new_fs);
1796 goto bad_unshare_out;
1797 err = unshare_fd(unshare_flags, &new_fd);
1799 goto bad_unshare_cleanup_fs;
1800 err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, new_fs);
1802 goto bad_unshare_cleanup_fd;
1804 if (new_fs || new_fd || do_sysvsem || new_nsproxy) {
1807 * CLONE_SYSVSEM is equivalent to sys_exit().
1813 switch_task_namespaces(current, new_nsproxy);
1821 spin_lock(&fs->lock);
1822 current->fs = new_fs;
1827 spin_unlock(&fs->lock);
1831 fd = current->files;
1832 current->files = new_fd;
1836 task_unlock(current);
1840 put_nsproxy(new_nsproxy);
1842 bad_unshare_cleanup_fd:
1844 put_files_struct(new_fd);
1846 bad_unshare_cleanup_fs:
1848 free_fs_struct(new_fs);
1855 * Helper to unshare the files of the current task.
1856 * We don't want to expose copy_files internals to
1857 * the exec layer of the kernel.
1860 int unshare_files(struct files_struct **displaced)
1862 struct task_struct *task = current;
1863 struct files_struct *copy = NULL;
1866 error = unshare_fd(CLONE_FILES, ©);
1867 if (error || !copy) {
1871 *displaced = task->files;