4 * Copyright (C) 1991, 1992 Linus Torvalds
8 * 'fork.c' contains the help-routines for the 'fork' system call
9 * (see also entry.S and others).
10 * Fork is rather simple, once you get the hang of it, but the memory
11 * management can be a bitch. See 'mm/memory.c': 'copy_page_range()'
14 #include <linux/slab.h>
15 #include <linux/init.h>
16 #include <linux/unistd.h>
17 #include <linux/module.h>
18 #include <linux/vmalloc.h>
19 #include <linux/completion.h>
20 #include <linux/personality.h>
21 #include <linux/mempolicy.h>
22 #include <linux/sem.h>
23 #include <linux/file.h>
24 #include <linux/fdtable.h>
25 #include <linux/iocontext.h>
26 #include <linux/key.h>
27 #include <linux/binfmts.h>
28 #include <linux/mman.h>
29 #include <linux/mmu_notifier.h>
31 #include <linux/nsproxy.h>
32 #include <linux/capability.h>
33 #include <linux/cpu.h>
34 #include <linux/cgroup.h>
35 #include <linux/security.h>
36 #include <linux/hugetlb.h>
37 #include <linux/swap.h>
38 #include <linux/syscalls.h>
39 #include <linux/jiffies.h>
40 #include <linux/futex.h>
41 #include <linux/compat.h>
42 #include <linux/kthread.h>
43 #include <linux/task_io_accounting_ops.h>
44 #include <linux/rcupdate.h>
45 #include <linux/ptrace.h>
46 #include <linux/mount.h>
47 #include <linux/audit.h>
48 #include <linux/memcontrol.h>
49 #include <linux/ftrace.h>
50 #include <linux/profile.h>
51 #include <linux/rmap.h>
52 #include <linux/ksm.h>
53 #include <linux/acct.h>
54 #include <linux/tsacct_kern.h>
55 #include <linux/cn_proc.h>
56 #include <linux/freezer.h>
57 #include <linux/delayacct.h>
58 #include <linux/taskstats_kern.h>
59 #include <linux/random.h>
60 #include <linux/tty.h>
61 #include <linux/blkdev.h>
62 #include <linux/fs_struct.h>
63 #include <linux/magic.h>
64 #include <linux/perf_event.h>
65 #include <linux/posix-timers.h>
66 #include <linux/user-return-notifier.h>
67 #include <linux/oom.h>
68 #include <linux/khugepaged.h>
69 #include <linux/signalfd.h>
71 #include <asm/pgtable.h>
72 #include <asm/pgalloc.h>
73 #include <asm/uaccess.h>
74 #include <asm/mmu_context.h>
75 #include <asm/cacheflush.h>
76 #include <asm/tlbflush.h>
78 #include <trace/events/sched.h>
80 #define CREATE_TRACE_POINTS
81 #include <trace/events/task.h>
84 * Protected counters by write_lock_irq(&tasklist_lock)
86 unsigned long total_forks; /* Handle normal Linux uptimes. */
87 int nr_threads; /* The idle threads do not count.. */
89 int max_threads; /* tunable limit on nr_threads */
91 DEFINE_PER_CPU(unsigned long, process_counts) = 0;
93 __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
95 #ifdef CONFIG_PROVE_RCU
96 int lockdep_tasklist_lock_is_held(void)
98 return lockdep_is_held(&tasklist_lock);
100 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
101 #endif /* #ifdef CONFIG_PROVE_RCU */
103 int nr_processes(void)
108 for_each_possible_cpu(cpu)
109 total += per_cpu(process_counts, cpu);
114 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
115 static struct kmem_cache *task_struct_cachep;
117 static inline struct task_struct *alloc_task_struct_node(int node)
119 return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node);
122 void __weak arch_release_task_struct(struct task_struct *tsk) { }
124 static inline void free_task_struct(struct task_struct *tsk)
126 arch_release_task_struct(tsk);
127 kmem_cache_free(task_struct_cachep, tsk);
131 #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
133 void __weak arch_release_thread_info(struct thread_info *ti) { }
135 static struct thread_info *alloc_thread_info_node(struct task_struct *tsk,
138 struct page *page = alloc_pages_node(node, THREADINFO_GFP,
141 return page ? page_address(page) : NULL;
144 static inline void free_thread_info(struct thread_info *ti)
146 arch_release_thread_info(ti);
147 free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
151 /* SLAB cache for signal_struct structures (tsk->signal) */
152 static struct kmem_cache *signal_cachep;
154 /* SLAB cache for sighand_struct structures (tsk->sighand) */
155 struct kmem_cache *sighand_cachep;
157 /* SLAB cache for files_struct structures (tsk->files) */
158 struct kmem_cache *files_cachep;
160 /* SLAB cache for fs_struct structures (tsk->fs) */
161 struct kmem_cache *fs_cachep;
163 /* SLAB cache for vm_area_struct structures */
164 struct kmem_cache *vm_area_cachep;
166 /* SLAB cache for mm_struct structures (tsk->mm) */
167 static struct kmem_cache *mm_cachep;
169 static void account_kernel_stack(struct thread_info *ti, int account)
171 struct zone *zone = page_zone(virt_to_page(ti));
173 mod_zone_page_state(zone, NR_KERNEL_STACK, account);
176 void free_task(struct task_struct *tsk)
178 account_kernel_stack(tsk->stack, -1);
179 free_thread_info(tsk->stack);
180 rt_mutex_debug_task_free(tsk);
181 ftrace_graph_exit_task(tsk);
182 free_task_struct(tsk);
184 EXPORT_SYMBOL(free_task);
186 static inline void free_signal_struct(struct signal_struct *sig)
188 taskstats_tgid_free(sig);
189 sched_autogroup_exit(sig);
190 kmem_cache_free(signal_cachep, sig);
193 static inline void put_signal_struct(struct signal_struct *sig)
195 if (atomic_dec_and_test(&sig->sigcnt))
196 free_signal_struct(sig);
199 void __put_task_struct(struct task_struct *tsk)
201 WARN_ON(!tsk->exit_state);
202 WARN_ON(atomic_read(&tsk->usage));
203 WARN_ON(tsk == current);
205 security_task_free(tsk);
207 delayacct_tsk_free(tsk);
208 put_signal_struct(tsk->signal);
210 if (!profile_handoff_task(tsk))
213 EXPORT_SYMBOL_GPL(__put_task_struct);
215 void __init __weak arch_task_cache_init(void) { }
217 void __init fork_init(unsigned long mempages)
219 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
220 #ifndef ARCH_MIN_TASKALIGN
221 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
223 /* create a slab on which task_structs can be allocated */
225 kmem_cache_create("task_struct", sizeof(struct task_struct),
226 ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL);
229 /* do the arch specific task caches init */
230 arch_task_cache_init();
233 * The default maximum number of threads is set to a safe
234 * value: the thread structures can take up at most half
237 max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
240 * we need to allow at least 20 threads to boot a system
242 if (max_threads < 20)
245 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
246 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
247 init_task.signal->rlim[RLIMIT_SIGPENDING] =
248 init_task.signal->rlim[RLIMIT_NPROC];
251 int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst,
252 struct task_struct *src)
258 static struct task_struct *dup_task_struct(struct task_struct *orig)
260 struct task_struct *tsk;
261 struct thread_info *ti;
262 unsigned long *stackend;
263 int node = tsk_fork_get_node(orig);
266 prepare_to_copy(orig);
268 tsk = alloc_task_struct_node(node);
272 ti = alloc_thread_info_node(tsk, node);
274 free_task_struct(tsk);
278 err = arch_dup_task_struct(tsk, orig);
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();
295 * One for us, one for whoever does the "release_task()" (usually
298 atomic_set(&tsk->usage, 2);
299 #ifdef CONFIG_BLK_DEV_IO_TRACE
302 tsk->splice_pipe = NULL;
304 account_kernel_stack(ti, 1);
309 free_thread_info(ti);
310 free_task_struct(tsk);
315 static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
317 struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
318 struct rb_node **rb_link, *rb_parent;
320 unsigned long charge;
321 struct mempolicy *pol;
323 down_write(&oldmm->mmap_sem);
324 flush_cache_dup_mm(oldmm);
326 * Not linked in yet - no deadlock potential:
328 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
332 mm->mmap_cache = NULL;
333 mm->free_area_cache = oldmm->mmap_base;
334 mm->cached_hole_size = ~0UL;
336 cpumask_clear(mm_cpumask(mm));
338 rb_link = &mm->mm_rb.rb_node;
341 retval = ksm_fork(mm, oldmm);
344 retval = khugepaged_fork(mm, oldmm);
349 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
352 if (mpnt->vm_flags & VM_DONTCOPY) {
353 long pages = vma_pages(mpnt);
354 mm->total_vm -= pages;
355 vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
360 if (mpnt->vm_flags & VM_ACCOUNT) {
361 unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
362 if (security_vm_enough_memory_mm(oldmm, len)) /* sic */
366 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
370 INIT_LIST_HEAD(&tmp->anon_vma_chain);
371 pol = mpol_dup(vma_policy(mpnt));
372 retval = PTR_ERR(pol);
374 goto fail_nomem_policy;
375 vma_set_policy(tmp, pol);
377 if (anon_vma_fork(tmp, mpnt))
378 goto fail_nomem_anon_vma_fork;
379 tmp->vm_flags &= ~VM_LOCKED;
380 tmp->vm_next = tmp->vm_prev = NULL;
383 struct inode *inode = file->f_path.dentry->d_inode;
384 struct address_space *mapping = file->f_mapping;
387 if (tmp->vm_flags & VM_DENYWRITE)
388 atomic_dec(&inode->i_writecount);
389 mutex_lock(&mapping->i_mmap_mutex);
390 if (tmp->vm_flags & VM_SHARED)
391 mapping->i_mmap_writable++;
392 flush_dcache_mmap_lock(mapping);
393 /* insert tmp into the share list, just after mpnt */
394 vma_prio_tree_add(tmp, mpnt);
395 flush_dcache_mmap_unlock(mapping);
396 mutex_unlock(&mapping->i_mmap_mutex);
400 * Clear hugetlb-related page reserves for children. This only
401 * affects MAP_PRIVATE mappings. Faults generated by the child
402 * are not guaranteed to succeed, even if read-only
404 if (is_vm_hugetlb_page(tmp))
405 reset_vma_resv_huge_pages(tmp);
408 * Link in the new vma and copy the page table entries.
411 pprev = &tmp->vm_next;
415 __vma_link_rb(mm, tmp, rb_link, rb_parent);
416 rb_link = &tmp->vm_rb.rb_right;
417 rb_parent = &tmp->vm_rb;
420 retval = copy_page_range(mm, oldmm, mpnt);
422 if (tmp->vm_ops && tmp->vm_ops->open)
423 tmp->vm_ops->open(tmp);
428 /* a new mm has just been created */
429 arch_dup_mmap(oldmm, mm);
432 up_write(&mm->mmap_sem);
434 up_write(&oldmm->mmap_sem);
436 fail_nomem_anon_vma_fork:
439 kmem_cache_free(vm_area_cachep, tmp);
442 vm_unacct_memory(charge);
446 static inline int mm_alloc_pgd(struct mm_struct *mm)
448 mm->pgd = pgd_alloc(mm);
449 if (unlikely(!mm->pgd))
454 static inline void mm_free_pgd(struct mm_struct *mm)
456 pgd_free(mm, mm->pgd);
459 #define dup_mmap(mm, oldmm) (0)
460 #define mm_alloc_pgd(mm) (0)
461 #define mm_free_pgd(mm)
462 #endif /* CONFIG_MMU */
464 __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
466 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
467 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
469 static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
471 static int __init coredump_filter_setup(char *s)
473 default_dump_filter =
474 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
475 MMF_DUMP_FILTER_MASK;
479 __setup("coredump_filter=", coredump_filter_setup);
481 #include <linux/init_task.h>
483 static void mm_init_aio(struct mm_struct *mm)
486 spin_lock_init(&mm->ioctx_lock);
487 INIT_HLIST_HEAD(&mm->ioctx_list);
491 static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p)
493 atomic_set(&mm->mm_users, 1);
494 atomic_set(&mm->mm_count, 1);
495 init_rwsem(&mm->mmap_sem);
496 INIT_LIST_HEAD(&mm->mmlist);
497 mm->flags = (current->mm) ?
498 (current->mm->flags & MMF_INIT_MASK) : default_dump_filter;
499 mm->core_state = NULL;
501 memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
502 spin_lock_init(&mm->page_table_lock);
503 mm->free_area_cache = TASK_UNMAPPED_BASE;
504 mm->cached_hole_size = ~0UL;
506 mm_init_owner(mm, p);
508 if (likely(!mm_alloc_pgd(mm))) {
510 mmu_notifier_mm_init(mm);
518 static void check_mm(struct mm_struct *mm)
522 for (i = 0; i < NR_MM_COUNTERS; i++) {
523 long x = atomic_long_read(&mm->rss_stat.count[i]);
526 printk(KERN_ALERT "BUG: Bad rss-counter state "
527 "mm:%p idx:%d val:%ld\n", mm, i, x);
530 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
531 VM_BUG_ON(mm->pmd_huge_pte);
536 * Allocate and initialize an mm_struct.
538 struct mm_struct *mm_alloc(void)
540 struct mm_struct *mm;
546 memset(mm, 0, sizeof(*mm));
548 return mm_init(mm, current);
552 * Called when the last reference to the mm
553 * is dropped: either by a lazy thread or by
554 * mmput. Free the page directory and the mm.
556 void __mmdrop(struct mm_struct *mm)
558 BUG_ON(mm == &init_mm);
561 mmu_notifier_mm_destroy(mm);
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 struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
672 struct mm_struct *mm;
675 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
679 mm = get_task_mm(task);
680 if (mm && mm != current->mm &&
681 !ptrace_may_access(task, mode)) {
683 mm = ERR_PTR(-EACCES);
685 mutex_unlock(&task->signal->cred_guard_mutex);
690 static void complete_vfork_done(struct task_struct *tsk)
692 struct completion *vfork;
695 vfork = tsk->vfork_done;
697 tsk->vfork_done = NULL;
703 static int wait_for_vfork_done(struct task_struct *child,
704 struct completion *vfork)
708 freezer_do_not_count();
709 killed = wait_for_completion_killable(vfork);
714 child->vfork_done = NULL;
718 put_task_struct(child);
722 /* Please note the differences between mmput and mm_release.
723 * mmput is called whenever we stop holding onto a mm_struct,
724 * error success whatever.
726 * mm_release is called after a mm_struct has been removed
727 * from the current process.
729 * This difference is important for error handling, when we
730 * only half set up a mm_struct for a new process and need to restore
731 * the old one. Because we mmput the new mm_struct before
732 * restoring the old one. . .
733 * Eric Biederman 10 January 1998
735 void mm_release(struct task_struct *tsk, struct mm_struct *mm)
737 /* Get rid of any futexes when releasing the mm */
739 if (unlikely(tsk->robust_list)) {
740 exit_robust_list(tsk);
741 tsk->robust_list = NULL;
744 if (unlikely(tsk->compat_robust_list)) {
745 compat_exit_robust_list(tsk);
746 tsk->compat_robust_list = NULL;
749 if (unlikely(!list_empty(&tsk->pi_state_list)))
750 exit_pi_state_list(tsk);
753 /* Get rid of any cached register state */
754 deactivate_mm(tsk, mm);
757 complete_vfork_done(tsk);
760 * If we're exiting normally, clear a user-space tid field if
761 * requested. We leave this alone when dying by signal, to leave
762 * the value intact in a core dump, and to save the unnecessary
763 * trouble, say, a killed vfork parent shouldn't touch this mm.
764 * Userland only wants this done for a sys_exit.
766 if (tsk->clear_child_tid) {
767 if (!(tsk->flags & PF_SIGNALED) &&
768 atomic_read(&mm->mm_users) > 1) {
770 * We don't check the error code - if userspace has
771 * not set up a proper pointer then tough luck.
773 put_user(0, tsk->clear_child_tid);
774 sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
777 tsk->clear_child_tid = NULL;
782 * Allocate a new mm structure and copy contents from the
783 * mm structure of the passed in task structure.
785 struct mm_struct *dup_mm(struct task_struct *tsk)
787 struct mm_struct *mm, *oldmm = current->mm;
797 memcpy(mm, oldmm, sizeof(*mm));
800 /* Initializing for Swap token stuff */
801 mm->token_priority = 0;
802 mm->last_interval = 0;
804 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
805 mm->pmd_huge_pte = NULL;
808 if (!mm_init(mm, tsk))
811 if (init_new_context(tsk, mm))
814 dup_mm_exe_file(oldmm, mm);
816 err = dup_mmap(mm, oldmm);
820 mm->hiwater_rss = get_mm_rss(mm);
821 mm->hiwater_vm = mm->total_vm;
823 if (mm->binfmt && !try_module_get(mm->binfmt->module))
829 /* don't put binfmt in mmput, we haven't got module yet */
838 * If init_new_context() failed, we cannot use mmput() to free the mm
839 * because it calls destroy_context()
846 static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
848 struct mm_struct *mm, *oldmm;
851 tsk->min_flt = tsk->maj_flt = 0;
852 tsk->nvcsw = tsk->nivcsw = 0;
853 #ifdef CONFIG_DETECT_HUNG_TASK
854 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
858 tsk->active_mm = NULL;
861 * Are we cloning a kernel thread?
863 * We need to steal a active VM for that..
869 if (clone_flags & CLONE_VM) {
870 atomic_inc(&oldmm->mm_users);
881 /* Initializing for Swap token stuff */
882 mm->token_priority = 0;
883 mm->last_interval = 0;
893 static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
895 struct fs_struct *fs = current->fs;
896 if (clone_flags & CLONE_FS) {
897 /* tsk->fs is already what we want */
898 spin_lock(&fs->lock);
900 spin_unlock(&fs->lock);
904 spin_unlock(&fs->lock);
907 tsk->fs = copy_fs_struct(fs);
913 static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
915 struct files_struct *oldf, *newf;
919 * A background process may not have any files ...
921 oldf = current->files;
925 if (clone_flags & CLONE_FILES) {
926 atomic_inc(&oldf->count);
930 newf = dup_fd(oldf, &error);
940 static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
943 struct io_context *ioc = current->io_context;
944 struct io_context *new_ioc;
949 * Share io context with parent, if CLONE_IO is set
951 if (clone_flags & CLONE_IO) {
952 tsk->io_context = ioc_task_link(ioc);
953 if (unlikely(!tsk->io_context))
955 } else if (ioprio_valid(ioc->ioprio)) {
956 new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE);
957 if (unlikely(!new_ioc))
960 new_ioc->ioprio = ioc->ioprio;
961 put_io_context(new_ioc);
967 static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
969 struct sighand_struct *sig;
971 if (clone_flags & CLONE_SIGHAND) {
972 atomic_inc(¤t->sighand->count);
975 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
976 rcu_assign_pointer(tsk->sighand, sig);
979 atomic_set(&sig->count, 1);
980 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
984 void __cleanup_sighand(struct sighand_struct *sighand)
986 if (atomic_dec_and_test(&sighand->count)) {
987 signalfd_cleanup(sighand);
988 kmem_cache_free(sighand_cachep, sighand);
994 * Initialize POSIX timer handling for a thread group.
996 static void posix_cpu_timers_init_group(struct signal_struct *sig)
998 unsigned long cpu_limit;
1000 /* Thread group counters. */
1001 thread_group_cputime_init(sig);
1003 cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
1004 if (cpu_limit != RLIM_INFINITY) {
1005 sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit);
1006 sig->cputimer.running = 1;
1009 /* The timer lists. */
1010 INIT_LIST_HEAD(&sig->cpu_timers[0]);
1011 INIT_LIST_HEAD(&sig->cpu_timers[1]);
1012 INIT_LIST_HEAD(&sig->cpu_timers[2]);
1015 static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
1017 struct signal_struct *sig;
1019 if (clone_flags & CLONE_THREAD)
1022 sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
1027 sig->nr_threads = 1;
1028 atomic_set(&sig->live, 1);
1029 atomic_set(&sig->sigcnt, 1);
1030 init_waitqueue_head(&sig->wait_chldexit);
1031 if (clone_flags & CLONE_NEWPID)
1032 sig->flags |= SIGNAL_UNKILLABLE;
1033 sig->curr_target = tsk;
1034 init_sigpending(&sig->shared_pending);
1035 INIT_LIST_HEAD(&sig->posix_timers);
1037 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
1038 sig->real_timer.function = it_real_fn;
1040 task_lock(current->group_leader);
1041 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
1042 task_unlock(current->group_leader);
1044 posix_cpu_timers_init_group(sig);
1046 tty_audit_fork(sig);
1047 sched_autogroup_fork(sig);
1049 #ifdef CONFIG_CGROUPS
1050 init_rwsem(&sig->group_rwsem);
1053 sig->oom_adj = current->signal->oom_adj;
1054 sig->oom_score_adj = current->signal->oom_score_adj;
1055 sig->oom_score_adj_min = current->signal->oom_score_adj_min;
1057 sig->has_child_subreaper = current->signal->has_child_subreaper ||
1058 current->signal->is_child_subreaper;
1060 mutex_init(&sig->cred_guard_mutex);
1065 static void copy_flags(unsigned long clone_flags, struct task_struct *p)
1067 unsigned long new_flags = p->flags;
1069 new_flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER);
1070 new_flags |= PF_FORKNOEXEC;
1071 p->flags = new_flags;
1074 SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
1076 current->clear_child_tid = tidptr;
1078 return task_pid_vnr(current);
1081 static void rt_mutex_init_task(struct task_struct *p)
1083 raw_spin_lock_init(&p->pi_lock);
1084 #ifdef CONFIG_RT_MUTEXES
1085 plist_head_init(&p->pi_waiters);
1086 p->pi_blocked_on = NULL;
1090 #ifdef CONFIG_MM_OWNER
1091 void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
1095 #endif /* CONFIG_MM_OWNER */
1098 * Initialize POSIX timer handling for a single task.
1100 static void posix_cpu_timers_init(struct task_struct *tsk)
1102 tsk->cputime_expires.prof_exp = 0;
1103 tsk->cputime_expires.virt_exp = 0;
1104 tsk->cputime_expires.sched_exp = 0;
1105 INIT_LIST_HEAD(&tsk->cpu_timers[0]);
1106 INIT_LIST_HEAD(&tsk->cpu_timers[1]);
1107 INIT_LIST_HEAD(&tsk->cpu_timers[2]);
1111 * This creates a new process as a copy of the old one,
1112 * but does not actually start it yet.
1114 * It copies the registers, and all the appropriate
1115 * parts of the process environment (as per the clone
1116 * flags). The actual kick-off is left to the caller.
1118 static struct task_struct *copy_process(unsigned long clone_flags,
1119 unsigned long stack_start,
1120 struct pt_regs *regs,
1121 unsigned long stack_size,
1122 int __user *child_tidptr,
1127 struct task_struct *p;
1128 int cgroup_callbacks_done = 0;
1130 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
1131 return ERR_PTR(-EINVAL);
1134 * Thread groups must share signals as well, and detached threads
1135 * can only be started up within the thread group.
1137 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1138 return ERR_PTR(-EINVAL);
1141 * Shared signal handlers imply shared VM. By way of the above,
1142 * thread groups also imply shared VM. Blocking this case allows
1143 * for various simplifications in other code.
1145 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1146 return ERR_PTR(-EINVAL);
1149 * Siblings of global init remain as zombies on exit since they are
1150 * not reaped by their parent (swapper). To solve this and to avoid
1151 * multi-rooted process trees, prevent global and container-inits
1152 * from creating siblings.
1154 if ((clone_flags & CLONE_PARENT) &&
1155 current->signal->flags & SIGNAL_UNKILLABLE)
1156 return ERR_PTR(-EINVAL);
1158 retval = security_task_create(clone_flags);
1163 p = dup_task_struct(current);
1167 ftrace_graph_init_task(p);
1169 rt_mutex_init_task(p);
1171 #ifdef CONFIG_PROVE_LOCKING
1172 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1173 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1176 if (atomic_read(&p->real_cred->user->processes) >=
1177 task_rlimit(p, RLIMIT_NPROC)) {
1178 if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
1179 p->real_cred->user != INIT_USER)
1182 current->flags &= ~PF_NPROC_EXCEEDED;
1184 retval = copy_creds(p, clone_flags);
1189 * If multiple threads are within copy_process(), then this check
1190 * triggers too late. This doesn't hurt, the check is only there
1191 * to stop root fork bombs.
1194 if (nr_threads >= max_threads)
1195 goto bad_fork_cleanup_count;
1197 if (!try_module_get(task_thread_info(p)->exec_domain->module))
1198 goto bad_fork_cleanup_count;
1201 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
1202 copy_flags(clone_flags, p);
1203 INIT_LIST_HEAD(&p->children);
1204 INIT_LIST_HEAD(&p->sibling);
1205 rcu_copy_process(p);
1206 p->vfork_done = NULL;
1207 spin_lock_init(&p->alloc_lock);
1209 init_sigpending(&p->pending);
1211 p->utime = p->stime = p->gtime = 0;
1212 p->utimescaled = p->stimescaled = 0;
1213 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
1214 p->prev_utime = p->prev_stime = 0;
1216 #if defined(SPLIT_RSS_COUNTING)
1217 memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1220 p->default_timer_slack_ns = current->timer_slack_ns;
1222 task_io_accounting_init(&p->ioac);
1223 acct_clear_integrals(p);
1225 posix_cpu_timers_init(p);
1227 do_posix_clock_monotonic_gettime(&p->start_time);
1228 p->real_start_time = p->start_time;
1229 monotonic_to_bootbased(&p->real_start_time);
1230 p->io_context = NULL;
1231 p->audit_context = NULL;
1232 if (clone_flags & CLONE_THREAD)
1233 threadgroup_change_begin(current);
1236 p->mempolicy = mpol_dup(p->mempolicy);
1237 if (IS_ERR(p->mempolicy)) {
1238 retval = PTR_ERR(p->mempolicy);
1239 p->mempolicy = NULL;
1240 goto bad_fork_cleanup_cgroup;
1242 mpol_fix_fork_child_flag(p);
1244 #ifdef CONFIG_CPUSETS
1245 p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
1246 p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
1247 seqcount_init(&p->mems_allowed_seq);
1249 #ifdef CONFIG_TRACE_IRQFLAGS
1251 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1252 p->hardirqs_enabled = 1;
1254 p->hardirqs_enabled = 0;
1256 p->hardirq_enable_ip = 0;
1257 p->hardirq_enable_event = 0;
1258 p->hardirq_disable_ip = _THIS_IP_;
1259 p->hardirq_disable_event = 0;
1260 p->softirqs_enabled = 1;
1261 p->softirq_enable_ip = _THIS_IP_;
1262 p->softirq_enable_event = 0;
1263 p->softirq_disable_ip = 0;
1264 p->softirq_disable_event = 0;
1265 p->hardirq_context = 0;
1266 p->softirq_context = 0;
1268 #ifdef CONFIG_LOCKDEP
1269 p->lockdep_depth = 0; /* no locks held yet */
1270 p->curr_chain_key = 0;
1271 p->lockdep_recursion = 0;
1274 #ifdef CONFIG_DEBUG_MUTEXES
1275 p->blocked_on = NULL; /* not blocked yet */
1277 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
1278 p->memcg_batch.do_batch = 0;
1279 p->memcg_batch.memcg = NULL;
1282 /* Perform scheduler related setup. Assign this task to a CPU. */
1285 retval = perf_event_init_task(p);
1287 goto bad_fork_cleanup_policy;
1288 retval = audit_alloc(p);
1290 goto bad_fork_cleanup_policy;
1291 /* copy all the process information */
1292 retval = copy_semundo(clone_flags, p);
1294 goto bad_fork_cleanup_audit;
1295 retval = copy_files(clone_flags, p);
1297 goto bad_fork_cleanup_semundo;
1298 retval = copy_fs(clone_flags, p);
1300 goto bad_fork_cleanup_files;
1301 retval = copy_sighand(clone_flags, p);
1303 goto bad_fork_cleanup_fs;
1304 retval = copy_signal(clone_flags, p);
1306 goto bad_fork_cleanup_sighand;
1307 retval = copy_mm(clone_flags, p);
1309 goto bad_fork_cleanup_signal;
1310 retval = copy_namespaces(clone_flags, p);
1312 goto bad_fork_cleanup_mm;
1313 retval = copy_io(clone_flags, p);
1315 goto bad_fork_cleanup_namespaces;
1316 retval = copy_thread(clone_flags, stack_start, stack_size, p, regs);
1318 goto bad_fork_cleanup_io;
1320 if (pid != &init_struct_pid) {
1322 pid = alloc_pid(p->nsproxy->pid_ns);
1324 goto bad_fork_cleanup_io;
1327 p->pid = pid_nr(pid);
1329 if (clone_flags & CLONE_THREAD)
1330 p->tgid = current->tgid;
1332 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1334 * Clear TID on mm_release()?
1336 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL;
1341 p->robust_list = NULL;
1342 #ifdef CONFIG_COMPAT
1343 p->compat_robust_list = NULL;
1345 INIT_LIST_HEAD(&p->pi_state_list);
1346 p->pi_state_cache = NULL;
1349 * sigaltstack should be cleared when sharing the same VM
1351 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1352 p->sas_ss_sp = p->sas_ss_size = 0;
1355 * Syscall tracing and stepping should be turned off in the
1356 * child regardless of CLONE_PTRACE.
1358 user_disable_single_step(p);
1359 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
1360 #ifdef TIF_SYSCALL_EMU
1361 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1363 clear_all_latency_tracing(p);
1365 /* ok, now we should be set up.. */
1366 if (clone_flags & CLONE_THREAD)
1367 p->exit_signal = -1;
1368 else if (clone_flags & CLONE_PARENT)
1369 p->exit_signal = current->group_leader->exit_signal;
1371 p->exit_signal = (clone_flags & CSIGNAL);
1373 p->pdeath_signal = 0;
1377 p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
1378 p->dirty_paused_when = 0;
1381 * Ok, make it visible to the rest of the system.
1382 * We dont wake it up yet.
1384 p->group_leader = p;
1385 INIT_LIST_HEAD(&p->thread_group);
1387 /* Now that the task is set up, run cgroup callbacks if
1388 * necessary. We need to run them before the task is visible
1389 * on the tasklist. */
1390 cgroup_fork_callbacks(p);
1391 cgroup_callbacks_done = 1;
1393 /* Need tasklist lock for parent etc handling! */
1394 write_lock_irq(&tasklist_lock);
1396 /* CLONE_PARENT re-uses the old parent */
1397 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
1398 p->real_parent = current->real_parent;
1399 p->parent_exec_id = current->parent_exec_id;
1401 p->real_parent = current;
1402 p->parent_exec_id = current->self_exec_id;
1405 spin_lock(¤t->sighand->siglock);
1408 * Process group and session signals need to be delivered to just the
1409 * parent before the fork or both the parent and the child after the
1410 * fork. Restart if a signal comes in before we add the new process to
1411 * it's process group.
1412 * A fatal signal pending means that current will exit, so the new
1413 * thread can't slip out of an OOM kill (or normal SIGKILL).
1415 recalc_sigpending();
1416 if (signal_pending(current)) {
1417 spin_unlock(¤t->sighand->siglock);
1418 write_unlock_irq(&tasklist_lock);
1419 retval = -ERESTARTNOINTR;
1420 goto bad_fork_free_pid;
1423 if (clone_flags & CLONE_THREAD) {
1424 current->signal->nr_threads++;
1425 atomic_inc(¤t->signal->live);
1426 atomic_inc(¤t->signal->sigcnt);
1427 p->group_leader = current->group_leader;
1428 list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
1431 if (likely(p->pid)) {
1432 ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
1434 if (thread_group_leader(p)) {
1435 if (is_child_reaper(pid))
1436 p->nsproxy->pid_ns->child_reaper = p;
1438 p->signal->leader_pid = pid;
1439 p->signal->tty = tty_kref_get(current->signal->tty);
1440 attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
1441 attach_pid(p, PIDTYPE_SID, task_session(current));
1442 list_add_tail(&p->sibling, &p->real_parent->children);
1443 list_add_tail_rcu(&p->tasks, &init_task.tasks);
1444 __this_cpu_inc(process_counts);
1446 attach_pid(p, PIDTYPE_PID, pid);
1451 spin_unlock(¤t->sighand->siglock);
1452 write_unlock_irq(&tasklist_lock);
1453 proc_fork_connector(p);
1454 cgroup_post_fork(p);
1455 if (clone_flags & CLONE_THREAD)
1456 threadgroup_change_end(current);
1459 trace_task_newtask(p, clone_flags);
1464 if (pid != &init_struct_pid)
1466 bad_fork_cleanup_io:
1469 bad_fork_cleanup_namespaces:
1470 exit_task_namespaces(p);
1471 bad_fork_cleanup_mm:
1474 bad_fork_cleanup_signal:
1475 if (!(clone_flags & CLONE_THREAD))
1476 free_signal_struct(p->signal);
1477 bad_fork_cleanup_sighand:
1478 __cleanup_sighand(p->sighand);
1479 bad_fork_cleanup_fs:
1480 exit_fs(p); /* blocking */
1481 bad_fork_cleanup_files:
1482 exit_files(p); /* blocking */
1483 bad_fork_cleanup_semundo:
1485 bad_fork_cleanup_audit:
1487 bad_fork_cleanup_policy:
1488 perf_event_free_task(p);
1490 mpol_put(p->mempolicy);
1491 bad_fork_cleanup_cgroup:
1493 if (clone_flags & CLONE_THREAD)
1494 threadgroup_change_end(current);
1495 cgroup_exit(p, cgroup_callbacks_done);
1496 delayacct_tsk_free(p);
1497 module_put(task_thread_info(p)->exec_domain->module);
1498 bad_fork_cleanup_count:
1499 atomic_dec(&p->cred->user->processes);
1504 return ERR_PTR(retval);
1507 noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
1509 memset(regs, 0, sizeof(struct pt_regs));
1513 static inline void init_idle_pids(struct pid_link *links)
1517 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1518 INIT_HLIST_NODE(&links[type].node); /* not really needed */
1519 links[type].pid = &init_struct_pid;
1523 struct task_struct * __cpuinit fork_idle(int cpu)
1525 struct task_struct *task;
1526 struct pt_regs regs;
1528 task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL,
1529 &init_struct_pid, 0);
1530 if (!IS_ERR(task)) {
1531 init_idle_pids(task->pids);
1532 init_idle(task, cpu);
1539 * Ok, this is the main fork-routine.
1541 * It copies the process, and if successful kick-starts
1542 * it and waits for it to finish using the VM if required.
1544 long do_fork(unsigned long clone_flags,
1545 unsigned long stack_start,
1546 struct pt_regs *regs,
1547 unsigned long stack_size,
1548 int __user *parent_tidptr,
1549 int __user *child_tidptr)
1551 struct task_struct *p;
1556 * Do some preliminary argument and permissions checking before we
1557 * actually start allocating stuff
1559 if (clone_flags & CLONE_NEWUSER) {
1560 if (clone_flags & CLONE_THREAD)
1562 /* hopefully this check will go away when userns support is
1565 if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) ||
1566 !capable(CAP_SETGID))
1571 * Determine whether and which event to report to ptracer. When
1572 * called from kernel_thread or CLONE_UNTRACED is explicitly
1573 * requested, no event is reported; otherwise, report if the event
1574 * for the type of forking is enabled.
1576 if (likely(user_mode(regs)) && !(clone_flags & CLONE_UNTRACED)) {
1577 if (clone_flags & CLONE_VFORK)
1578 trace = PTRACE_EVENT_VFORK;
1579 else if ((clone_flags & CSIGNAL) != SIGCHLD)
1580 trace = PTRACE_EVENT_CLONE;
1582 trace = PTRACE_EVENT_FORK;
1584 if (likely(!ptrace_event_enabled(current, trace)))
1588 p = copy_process(clone_flags, stack_start, regs, stack_size,
1589 child_tidptr, NULL, trace);
1591 * Do this prior waking up the new thread - the thread pointer
1592 * might get invalid after that point, if the thread exits quickly.
1595 struct completion vfork;
1597 trace_sched_process_fork(current, p);
1599 nr = task_pid_vnr(p);
1601 if (clone_flags & CLONE_PARENT_SETTID)
1602 put_user(nr, parent_tidptr);
1604 if (clone_flags & CLONE_VFORK) {
1605 p->vfork_done = &vfork;
1606 init_completion(&vfork);
1610 wake_up_new_task(p);
1612 /* forking complete and child started to run, tell ptracer */
1613 if (unlikely(trace))
1614 ptrace_event(trace, nr);
1616 if (clone_flags & CLONE_VFORK) {
1617 if (!wait_for_vfork_done(p, &vfork))
1618 ptrace_event(PTRACE_EVENT_VFORK_DONE, nr);
1626 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1627 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1630 static void sighand_ctor(void *data)
1632 struct sighand_struct *sighand = data;
1634 spin_lock_init(&sighand->siglock);
1635 init_waitqueue_head(&sighand->signalfd_wqh);
1638 void __init proc_caches_init(void)
1640 sighand_cachep = kmem_cache_create("sighand_cache",
1641 sizeof(struct sighand_struct), 0,
1642 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
1643 SLAB_NOTRACK, sighand_ctor);
1644 signal_cachep = kmem_cache_create("signal_cache",
1645 sizeof(struct signal_struct), 0,
1646 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1647 files_cachep = kmem_cache_create("files_cache",
1648 sizeof(struct files_struct), 0,
1649 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1650 fs_cachep = kmem_cache_create("fs_cache",
1651 sizeof(struct fs_struct), 0,
1652 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1654 * FIXME! The "sizeof(struct mm_struct)" currently includes the
1655 * whole struct cpumask for the OFFSTACK case. We could change
1656 * this to *only* allocate as much of it as required by the
1657 * maximum number of CPU's we can ever have. The cpumask_allocation
1658 * is at the end of the structure, exactly for that reason.
1660 mm_cachep = kmem_cache_create("mm_struct",
1661 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
1662 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1663 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
1665 nsproxy_cache_init();
1669 * Check constraints on flags passed to the unshare system call.
1671 static int check_unshare_flags(unsigned long unshare_flags)
1673 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
1674 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
1675 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET))
1678 * Not implemented, but pretend it works if there is nothing to
1679 * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND
1680 * needs to unshare vm.
1682 if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
1683 /* FIXME: get_task_mm() increments ->mm_users */
1684 if (atomic_read(¤t->mm->mm_users) > 1)
1692 * Unshare the filesystem structure if it is being shared
1694 static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1696 struct fs_struct *fs = current->fs;
1698 if (!(unshare_flags & CLONE_FS) || !fs)
1701 /* don't need lock here; in the worst case we'll do useless copy */
1705 *new_fsp = copy_fs_struct(fs);
1713 * Unshare file descriptor table if it is being shared
1715 static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1717 struct files_struct *fd = current->files;
1720 if ((unshare_flags & CLONE_FILES) &&
1721 (fd && atomic_read(&fd->count) > 1)) {
1722 *new_fdp = dup_fd(fd, &error);
1731 * unshare allows a process to 'unshare' part of the process
1732 * context which was originally shared using clone. copy_*
1733 * functions used by do_fork() cannot be used here directly
1734 * because they modify an inactive task_struct that is being
1735 * constructed. Here we are modifying the current, active,
1738 SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
1740 struct fs_struct *fs, *new_fs = NULL;
1741 struct files_struct *fd, *new_fd = NULL;
1742 struct nsproxy *new_nsproxy = NULL;
1746 err = check_unshare_flags(unshare_flags);
1748 goto bad_unshare_out;
1751 * If unsharing namespace, must also unshare filesystem information.
1753 if (unshare_flags & CLONE_NEWNS)
1754 unshare_flags |= CLONE_FS;
1756 * CLONE_NEWIPC must also detach from the undolist: after switching
1757 * to a new ipc namespace, the semaphore arrays from the old
1758 * namespace are unreachable.
1760 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
1762 err = unshare_fs(unshare_flags, &new_fs);
1764 goto bad_unshare_out;
1765 err = unshare_fd(unshare_flags, &new_fd);
1767 goto bad_unshare_cleanup_fs;
1768 err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, new_fs);
1770 goto bad_unshare_cleanup_fd;
1772 if (new_fs || new_fd || do_sysvsem || new_nsproxy) {
1775 * CLONE_SYSVSEM is equivalent to sys_exit().
1781 switch_task_namespaces(current, new_nsproxy);
1789 spin_lock(&fs->lock);
1790 current->fs = new_fs;
1795 spin_unlock(&fs->lock);
1799 fd = current->files;
1800 current->files = new_fd;
1804 task_unlock(current);
1808 put_nsproxy(new_nsproxy);
1810 bad_unshare_cleanup_fd:
1812 put_files_struct(new_fd);
1814 bad_unshare_cleanup_fs:
1816 free_fs_struct(new_fs);
1823 * Helper to unshare the files of the current task.
1824 * We don't want to expose copy_files internals to
1825 * the exec layer of the kernel.
1828 int unshare_files(struct files_struct **displaced)
1830 struct task_struct *task = current;
1831 struct files_struct *copy = NULL;
1834 error = unshare_fd(CLONE_FILES, ©);
1835 if (error || !copy) {
1839 *displaced = task->files;