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/smp_lock.h>
18 #include <linux/module.h>
19 #include <linux/vmalloc.h>
20 #include <linux/completion.h>
21 #include <linux/mnt_namespace.h>
22 #include <linux/personality.h>
23 #include <linux/mempolicy.h>
24 #include <linux/sem.h>
25 #include <linux/file.h>
26 #include <linux/key.h>
27 #include <linux/binfmts.h>
28 #include <linux/mman.h>
30 #include <linux/nsproxy.h>
31 #include <linux/capability.h>
32 #include <linux/cpu.h>
33 #include <linux/cpuset.h>
34 #include <linux/security.h>
35 #include <linux/swap.h>
36 #include <linux/syscalls.h>
37 #include <linux/jiffies.h>
38 #include <linux/futex.h>
39 #include <linux/task_io_accounting_ops.h>
40 #include <linux/rcupdate.h>
41 #include <linux/tracehook.h>
42 #include <linux/mount.h>
43 #include <linux/audit.h>
44 #include <linux/profile.h>
45 #include <linux/rmap.h>
46 #include <linux/acct.h>
47 #include <linux/tsacct_kern.h>
48 #include <linux/cn_proc.h>
49 #include <linux/delayacct.h>
50 #include <linux/taskstats_kern.h>
51 #include <linux/random.h>
52 #include <linux/ptrace.h>
53 #include <linux/vs_context.h>
54 #include <linux/vs_network.h>
55 #include <linux/vs_limit.h>
56 #include <linux/vs_memory.h>
57 #include <linux/vserver/global.h>
59 #include <asm/pgtable.h>
60 #include <asm/pgalloc.h>
61 #include <asm/uaccess.h>
62 #include <asm/mmu_context.h>
63 #include <asm/cacheflush.h>
64 #include <asm/tlbflush.h>
67 * Protected counters by write_lock_irq(&tasklist_lock)
69 unsigned long total_forks; /* Handle normal Linux uptimes. */
70 int nr_threads; /* The idle threads do not count.. */
72 int max_threads; /* tunable limit on nr_threads */
74 DEFINE_PER_CPU(unsigned long, process_counts) = 0;
76 __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
78 int nr_processes(void)
83 for_each_online_cpu(cpu)
84 total += per_cpu(process_counts, cpu);
89 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
90 # define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL)
91 # define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk))
92 static struct kmem_cache *task_struct_cachep;
95 /* SLAB cache for signal_struct structures (tsk->signal) */
96 static struct kmem_cache *signal_cachep;
98 /* SLAB cache for sighand_struct structures (tsk->sighand) */
99 struct kmem_cache *sighand_cachep;
101 /* SLAB cache for files_struct structures (tsk->files) */
102 struct kmem_cache *files_cachep;
104 /* SLAB cache for fs_struct structures (tsk->fs) */
105 struct kmem_cache *fs_cachep;
107 /* SLAB cache for vm_area_struct structures */
108 struct kmem_cache *vm_area_cachep;
110 /* SLAB cache for mm_struct structures (tsk->mm) */
111 static struct kmem_cache *mm_cachep;
113 void free_task(struct task_struct *tsk)
115 free_thread_info(tsk->thread_info);
116 rt_mutex_debug_task_free(tsk);
117 clr_vx_info(&tsk->vx_info);
118 clr_nx_info(&tsk->nx_info);
119 free_task_struct(tsk);
121 EXPORT_SYMBOL(free_task);
123 void __put_task_struct(struct task_struct *tsk)
125 WARN_ON(!(tsk->exit_state & (EXIT_DEAD | EXIT_ZOMBIE)));
126 WARN_ON(atomic_read(&tsk->usage));
127 WARN_ON(tsk == current);
129 security_task_free(tsk);
131 put_group_info(tsk->group_info);
132 delayacct_tsk_free(tsk);
134 if (!profile_handoff_task(tsk))
138 void __init fork_init(unsigned long mempages)
140 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
141 #ifndef ARCH_MIN_TASKALIGN
142 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
144 /* create a slab on which task_structs can be allocated */
146 kmem_cache_create("task_struct", sizeof(struct task_struct),
147 ARCH_MIN_TASKALIGN, SLAB_PANIC, NULL, NULL);
151 * The default maximum number of threads is set to a safe
152 * value: the thread structures can take up at most half
155 max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
158 * we need to allow at least 20 threads to boot a system
163 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
164 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
165 init_task.signal->rlim[RLIMIT_SIGPENDING] =
166 init_task.signal->rlim[RLIMIT_NPROC];
169 static struct task_struct *dup_task_struct(struct task_struct *orig)
171 struct task_struct *tsk;
172 struct thread_info *ti;
174 prepare_to_copy(orig);
176 tsk = alloc_task_struct();
180 ti = alloc_thread_info(tsk);
182 free_task_struct(tsk);
187 tsk->thread_info = ti;
188 setup_thread_stack(tsk, orig);
190 #ifdef CONFIG_CC_STACKPROTECTOR
191 tsk->stack_canary = get_random_int();
194 /* One for us, one for whoever does the "release_task()" (usually parent) */
195 atomic_set(&tsk->usage,2);
196 atomic_set(&tsk->fs_excl, 0);
197 #ifdef CONFIG_BLK_DEV_IO_TRACE
200 tsk->splice_pipe = NULL;
205 static inline int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
207 struct vm_area_struct *mpnt, *tmp, **pprev;
208 struct rb_node **rb_link, *rb_parent;
210 unsigned long charge;
211 struct mempolicy *pol;
213 down_write(&oldmm->mmap_sem);
214 flush_cache_dup_mm(oldmm);
216 * Not linked in yet - no deadlock potential:
218 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
222 mm->mmap_cache = NULL;
223 mm->free_area_cache = oldmm->mmap_base;
224 mm->cached_hole_size = ~0UL;
226 __set_mm_counter(mm, file_rss, 0);
227 __set_mm_counter(mm, anon_rss, 0);
228 cpus_clear(mm->cpu_vm_mask);
230 rb_link = &mm->mm_rb.rb_node;
234 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
237 if (mpnt->vm_flags & VM_DONTCOPY) {
238 long pages = vma_pages(mpnt);
239 vx_vmpages_sub(mm, pages);
240 vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
245 if (mpnt->vm_flags & VM_ACCOUNT) {
246 unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
247 if (security_vm_enough_memory(len))
251 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
255 pol = mpol_copy(vma_policy(mpnt));
256 retval = PTR_ERR(pol);
258 goto fail_nomem_policy;
259 vma_set_policy(tmp, pol);
260 tmp->vm_flags &= ~VM_LOCKED;
266 struct inode *inode = file->f_path.dentry->d_inode;
268 if (tmp->vm_flags & VM_DENYWRITE)
269 atomic_dec(&inode->i_writecount);
271 /* insert tmp into the share list, just after mpnt */
272 spin_lock(&file->f_mapping->i_mmap_lock);
273 tmp->vm_truncate_count = mpnt->vm_truncate_count;
274 flush_dcache_mmap_lock(file->f_mapping);
275 vma_prio_tree_add(tmp, mpnt);
276 flush_dcache_mmap_unlock(file->f_mapping);
277 spin_unlock(&file->f_mapping->i_mmap_lock);
281 * Link in the new vma and copy the page table entries.
284 pprev = &tmp->vm_next;
286 __vma_link_rb(mm, tmp, rb_link, rb_parent);
287 rb_link = &tmp->vm_rb.rb_right;
288 rb_parent = &tmp->vm_rb;
291 retval = copy_page_range(mm, oldmm, mpnt);
293 if (tmp->vm_ops && tmp->vm_ops->open)
294 tmp->vm_ops->open(tmp);
300 arch_dup_mmap(mm, oldmm);
304 up_write(&mm->mmap_sem);
306 up_write(&oldmm->mmap_sem);
309 kmem_cache_free(vm_area_cachep, tmp);
312 vm_unacct_memory(charge);
316 static inline int mm_alloc_pgd(struct mm_struct * mm)
318 mm->pgd = pgd_alloc(mm);
319 if (unlikely(!mm->pgd))
324 static inline void mm_free_pgd(struct mm_struct * mm)
329 #define dup_mmap(mm, oldmm) (0)
330 #define mm_alloc_pgd(mm) (0)
331 #define mm_free_pgd(mm)
332 #endif /* CONFIG_MMU */
334 __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
336 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
337 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
339 #include <linux/init_task.h>
341 static struct mm_struct * mm_init(struct mm_struct * mm)
343 atomic_set(&mm->mm_users, 1);
344 atomic_set(&mm->mm_count, 1);
345 init_rwsem(&mm->mmap_sem);
346 INIT_LIST_HEAD(&mm->mmlist);
347 mm->core_waiters = 0;
349 __set_mm_counter(mm, file_rss, 0);
350 __set_mm_counter(mm, anon_rss, 0);
351 spin_lock_init(&mm->page_table_lock);
352 rwlock_init(&mm->ioctx_list_lock);
353 mm->ioctx_list = NULL;
354 mm->free_area_cache = TASK_UNMAPPED_BASE;
355 mm->cached_hole_size = ~0UL;
357 if (likely(!mm_alloc_pgd(mm))) {
359 set_vx_info(&mm->mm_vx_info, current->vx_info);
367 * Allocate and initialize an mm_struct.
369 struct mm_struct * mm_alloc(void)
371 struct mm_struct * mm;
375 memset(mm, 0, sizeof(*mm));
382 * Called when the last reference to the mm
383 * is dropped: either by a lazy thread or by
384 * mmput. Free the page directory and the mm.
386 void fastcall __mmdrop(struct mm_struct *mm)
388 BUG_ON(mm == &init_mm);
391 clr_vx_info(&mm->mm_vx_info);
396 * Decrement the use count and release all resources for an mm.
398 void mmput(struct mm_struct *mm)
402 if (atomic_dec_and_test(&mm->mm_users)) {
405 if (!list_empty(&mm->mmlist)) {
406 spin_lock(&mmlist_lock);
407 list_del(&mm->mmlist);
408 spin_unlock(&mmlist_lock);
414 EXPORT_SYMBOL_GPL(mmput);
417 * get_task_mm - acquire a reference to the task's mm
419 * Returns %NULL if the task has no mm. Checks PF_BORROWED_MM (meaning
420 * this kernel workthread has transiently adopted a user mm with use_mm,
421 * to do its AIO) is not set and if so returns a reference to it, after
422 * bumping up the use count. User must release the mm via mmput()
423 * after use. Typically used by /proc and ptrace.
425 struct mm_struct *get_task_mm(struct task_struct *task)
427 struct mm_struct *mm;
432 if (task->flags & PF_BORROWED_MM)
435 atomic_inc(&mm->mm_users);
440 EXPORT_SYMBOL_GPL(get_task_mm);
442 /* Please note the differences between mmput and mm_release.
443 * mmput is called whenever we stop holding onto a mm_struct,
444 * error success whatever.
446 * mm_release is called after a mm_struct has been removed
447 * from the current process.
449 * This difference is important for error handling, when we
450 * only half set up a mm_struct for a new process and need to restore
451 * the old one. Because we mmput the new mm_struct before
452 * restoring the old one. . .
453 * Eric Biederman 10 January 1998
455 void mm_release(struct task_struct *tsk, struct mm_struct *mm)
457 struct completion *vfork_done = tsk->vfork_done;
459 /* Get rid of any cached register state */
460 deactivate_mm(tsk, mm);
462 /* notify parent sleeping on vfork() */
464 tsk->vfork_done = NULL;
465 complete(vfork_done);
469 * If we're exiting normally, clear a user-space tid field if
470 * requested. We leave this alone when dying by signal, to leave
471 * the value intact in a core dump, and to save the unnecessary
472 * trouble otherwise. Userland only wants this done for a sys_exit.
474 if (tsk->clear_child_tid
475 && !(tsk->flags & PF_SIGNALED)
476 && atomic_read(&mm->mm_users) > 1) {
477 u32 __user * tidptr = tsk->clear_child_tid;
478 tsk->clear_child_tid = NULL;
481 * We don't check the error code - if userspace has
482 * not set up a proper pointer then tough luck.
485 sys_futex(tidptr, FUTEX_WAKE, 1, NULL, NULL, 0);
490 * Allocate a new mm structure and copy contents from the
491 * mm structure of the passed in task structure.
493 static struct mm_struct *dup_mm(struct task_struct *tsk)
495 struct mm_struct *mm, *oldmm = current->mm;
505 memcpy(mm, oldmm, sizeof(*mm));
506 mm->mm_vx_info = NULL;
508 /* Initializing for Swap token stuff */
509 mm->token_priority = 0;
510 mm->last_interval = 0;
515 if (init_new_context(tsk, mm))
518 err = dup_mmap(mm, oldmm);
522 mm->hiwater_rss = get_mm_rss(mm);
523 mm->hiwater_vm = mm->total_vm;
535 * If init_new_context() failed, we cannot use mmput() to free the mm
536 * because it calls destroy_context()
538 clr_vx_info(&mm->mm_vx_info);
544 static int copy_mm(unsigned long clone_flags, struct task_struct * tsk)
546 struct mm_struct * mm, *oldmm;
549 tsk->min_flt = tsk->maj_flt = 0;
550 tsk->nvcsw = tsk->nivcsw = 0;
553 tsk->active_mm = NULL;
556 * Are we cloning a kernel thread?
558 * We need to steal a active VM for that..
564 if (clone_flags & CLONE_VM) {
565 atomic_inc(&oldmm->mm_users);
576 /* Initializing for Swap token stuff */
577 mm->token_priority = 0;
578 mm->last_interval = 0;
588 static inline struct fs_struct *__copy_fs_struct(struct fs_struct *old)
590 struct fs_struct *fs = kmem_cache_alloc(fs_cachep, GFP_KERNEL);
591 /* We don't need to lock fs - think why ;-) */
593 atomic_set(&fs->count, 1);
594 rwlock_init(&fs->lock);
595 fs->umask = old->umask;
596 read_lock(&old->lock);
597 fs->rootmnt = mntget(old->rootmnt);
598 fs->root = dget(old->root);
599 fs->pwdmnt = mntget(old->pwdmnt);
600 fs->pwd = dget(old->pwd);
602 fs->altrootmnt = mntget(old->altrootmnt);
603 fs->altroot = dget(old->altroot);
605 fs->altrootmnt = NULL;
608 read_unlock(&old->lock);
609 atomic_inc(&vs_global_fs);
614 struct fs_struct *copy_fs_struct(struct fs_struct *old)
616 return __copy_fs_struct(old);
619 EXPORT_SYMBOL_GPL(copy_fs_struct);
621 static inline int copy_fs(unsigned long clone_flags, struct task_struct * tsk)
623 if (clone_flags & CLONE_FS) {
624 atomic_inc(¤t->fs->count);
627 tsk->fs = __copy_fs_struct(current->fs);
633 static int count_open_files(struct fdtable *fdt)
635 int size = fdt->max_fds;
638 /* Find the last open fd */
639 for (i = size/(8*sizeof(long)); i > 0; ) {
640 if (fdt->open_fds->fds_bits[--i])
643 i = (i+1) * 8 * sizeof(long);
647 static struct files_struct *alloc_files(void)
649 struct files_struct *newf;
652 newf = kmem_cache_alloc(files_cachep, GFP_KERNEL);
656 atomic_set(&newf->count, 1);
658 spin_lock_init(&newf->file_lock);
661 fdt->max_fds = NR_OPEN_DEFAULT;
662 fdt->close_on_exec = (fd_set *)&newf->close_on_exec_init;
663 fdt->open_fds = (fd_set *)&newf->open_fds_init;
664 fdt->fd = &newf->fd_array[0];
665 INIT_RCU_HEAD(&fdt->rcu);
667 rcu_assign_pointer(newf->fdt, fdt);
673 * Allocate a new files structure and copy contents from the
674 * passed in files structure.
675 * errorp will be valid only when the returned files_struct is NULL.
677 static struct files_struct *dup_fd(struct files_struct *oldf, int *errorp)
679 struct files_struct *newf;
680 struct file **old_fds, **new_fds;
681 int open_files, size, i;
682 struct fdtable *old_fdt, *new_fdt;
685 newf = alloc_files();
689 spin_lock(&oldf->file_lock);
690 old_fdt = files_fdtable(oldf);
691 new_fdt = files_fdtable(newf);
692 open_files = count_open_files(old_fdt);
695 * Check whether we need to allocate a larger fd array and fd set.
696 * Note: we're not a clone task, so the open count won't change.
698 if (open_files > new_fdt->max_fds) {
699 new_fdt->max_fds = 0;
700 spin_unlock(&oldf->file_lock);
701 spin_lock(&newf->file_lock);
702 *errorp = expand_files(newf, open_files-1);
703 spin_unlock(&newf->file_lock);
706 new_fdt = files_fdtable(newf);
708 * Reacquire the oldf lock and a pointer to its fd table
709 * who knows it may have a new bigger fd table. We need
710 * the latest pointer.
712 spin_lock(&oldf->file_lock);
713 old_fdt = files_fdtable(oldf);
716 old_fds = old_fdt->fd;
717 new_fds = new_fdt->fd;
719 memcpy(new_fdt->open_fds->fds_bits,
720 old_fdt->open_fds->fds_bits, open_files/8);
721 memcpy(new_fdt->close_on_exec->fds_bits,
722 old_fdt->close_on_exec->fds_bits, open_files/8);
724 for (i = open_files; i != 0; i--) {
725 struct file *f = *old_fds++;
728 /* TODO: sum it first for check and performance */
729 vx_openfd_inc(open_files - i);
732 * The fd may be claimed in the fd bitmap but not yet
733 * instantiated in the files array if a sibling thread
734 * is partway through open(). So make sure that this
735 * fd is available to the new process.
737 FD_CLR(open_files - i, new_fdt->open_fds);
739 rcu_assign_pointer(*new_fds++, f);
741 spin_unlock(&oldf->file_lock);
743 /* compute the remainder to be cleared */
744 size = (new_fdt->max_fds - open_files) * sizeof(struct file *);
746 /* This is long word aligned thus could use a optimized version */
747 memset(new_fds, 0, size);
749 if (new_fdt->max_fds > open_files) {
750 int left = (new_fdt->max_fds-open_files)/8;
751 int start = open_files / (8 * sizeof(unsigned long));
753 memset(&new_fdt->open_fds->fds_bits[start], 0, left);
754 memset(&new_fdt->close_on_exec->fds_bits[start], 0, left);
760 kmem_cache_free(files_cachep, newf);
765 static int copy_files(unsigned long clone_flags, struct task_struct * tsk)
767 struct files_struct *oldf, *newf;
771 * A background process may not have any files ...
773 oldf = current->files;
777 if (clone_flags & CLONE_FILES) {
778 atomic_inc(&oldf->count);
783 * Note: we may be using current for both targets (See exec.c)
784 * This works because we cache current->files (old) as oldf. Don't
788 newf = dup_fd(oldf, &error);
799 * Helper to unshare the files of the current task.
800 * We don't want to expose copy_files internals to
801 * the exec layer of the kernel.
804 int unshare_files(void)
806 struct files_struct *files = current->files;
811 /* This can race but the race causes us to copy when we don't
812 need to and drop the copy */
813 if(atomic_read(&files->count) == 1)
815 atomic_inc(&files->count);
818 rc = copy_files(0, current);
820 current->files = files;
824 EXPORT_SYMBOL(unshare_files);
826 static inline int copy_sighand(unsigned long clone_flags, struct task_struct * tsk)
828 struct sighand_struct *sig;
830 if (clone_flags & (CLONE_SIGHAND | CLONE_THREAD)) {
831 atomic_inc(¤t->sighand->count);
834 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
835 rcu_assign_pointer(tsk->sighand, sig);
838 atomic_set(&sig->count, 1);
839 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
843 void __cleanup_sighand(struct sighand_struct *sighand)
845 if (atomic_dec_and_test(&sighand->count))
846 kmem_cache_free(sighand_cachep, sighand);
849 static inline int copy_signal(unsigned long clone_flags, struct task_struct * tsk)
851 struct signal_struct *sig;
854 if (clone_flags & CLONE_THREAD) {
855 atomic_inc(¤t->signal->count);
856 atomic_inc(¤t->signal->live);
859 sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL);
864 ret = copy_thread_group_keys(tsk);
866 kmem_cache_free(signal_cachep, sig);
870 atomic_set(&sig->count, 1);
871 atomic_set(&sig->live, 1);
872 init_waitqueue_head(&sig->wait_chldexit);
874 sig->group_exit_code = 0;
875 sig->group_exit_task = NULL;
876 sig->group_stop_count = 0;
877 sig->curr_target = NULL;
878 init_sigpending(&sig->shared_pending);
879 INIT_LIST_HEAD(&sig->posix_timers);
881 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_REL);
882 sig->it_real_incr.tv64 = 0;
883 sig->real_timer.function = it_real_fn;
886 sig->it_virt_expires = cputime_zero;
887 sig->it_virt_incr = cputime_zero;
888 sig->it_prof_expires = cputime_zero;
889 sig->it_prof_incr = cputime_zero;
891 sig->leader = 0; /* session leadership doesn't inherit */
892 sig->tty_old_pgrp = 0;
894 sig->utime = sig->stime = sig->cutime = sig->cstime = cputime_zero;
895 sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0;
896 sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0;
898 INIT_LIST_HEAD(&sig->cpu_timers[0]);
899 INIT_LIST_HEAD(&sig->cpu_timers[1]);
900 INIT_LIST_HEAD(&sig->cpu_timers[2]);
901 taskstats_tgid_init(sig);
903 task_lock(current->group_leader);
904 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
905 task_unlock(current->group_leader);
907 if (sig->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) {
909 * New sole thread in the process gets an expiry time
910 * of the whole CPU time limit.
912 tsk->it_prof_expires =
913 secs_to_cputime(sig->rlim[RLIMIT_CPU].rlim_cur);
915 acct_init_pacct(&sig->pacct);
920 void __cleanup_signal(struct signal_struct *sig)
922 exit_thread_group_keys(sig);
923 kmem_cache_free(signal_cachep, sig);
926 static inline void cleanup_signal(struct task_struct *tsk)
928 struct signal_struct *sig = tsk->signal;
930 atomic_dec(&sig->live);
932 if (atomic_dec_and_test(&sig->count))
933 __cleanup_signal(sig);
936 static inline void copy_flags(unsigned long clone_flags, struct task_struct *p)
938 unsigned long new_flags = p->flags;
940 new_flags &= ~(PF_SUPERPRIV | PF_NOFREEZE);
941 new_flags |= PF_FORKNOEXEC;
942 new_flags |= PF_STARTING;
943 p->flags = new_flags;
946 asmlinkage long sys_set_tid_address(int __user *tidptr)
948 current->clear_child_tid = tidptr;
953 static inline void rt_mutex_init_task(struct task_struct *p)
955 spin_lock_init(&p->pi_lock);
956 #ifdef CONFIG_RT_MUTEXES
957 plist_head_init(&p->pi_waiters, &p->pi_lock);
958 p->pi_blocked_on = NULL;
963 * This creates a new process as a copy of the old one,
964 * but does not actually start it yet.
966 * It copies the registers, and all the appropriate
967 * parts of the process environment (as per the clone
968 * flags). The actual kick-off is left to the caller.
970 static struct task_struct *copy_process(unsigned long clone_flags,
971 unsigned long stack_start,
972 struct pt_regs *regs,
973 unsigned long stack_size,
974 int __user *parent_tidptr,
975 int __user *child_tidptr,
979 struct task_struct *p = NULL;
983 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
984 return ERR_PTR(-EINVAL);
987 * Thread groups must share signals as well, and detached threads
988 * can only be started up within the thread group.
990 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
991 return ERR_PTR(-EINVAL);
994 * Shared signal handlers imply shared VM. By way of the above,
995 * thread groups also imply shared VM. Blocking this case allows
996 * for various simplifications in other code.
998 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
999 return ERR_PTR(-EINVAL);
1001 retval = security_task_create(clone_flags);
1006 p = dup_task_struct(current);
1010 rt_mutex_init_task(p);
1012 #ifdef CONFIG_TRACE_IRQFLAGS
1013 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1014 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1016 init_vx_info(&p->vx_info, current->vx_info);
1017 init_nx_info(&p->nx_info, current->nx_info);
1019 /* check vserver memory */
1020 if (p->mm && !(clone_flags & CLONE_VM)) {
1021 if (vx_vmpages_avail(p->mm, p->mm->total_vm))
1022 vx_pages_add(p->vx_info, RLIMIT_AS, p->mm->total_vm);
1026 if (p->mm && vx_flags(VXF_FORK_RSS, 0)) {
1027 if (!vx_rss_avail(p->mm, get_mm_counter(p->mm, file_rss)))
1028 goto bad_fork_cleanup_vm;
1032 if (!vx_nproc_avail(1))
1033 goto bad_fork_cleanup_vm;
1035 if (atomic_read(&p->user->processes) >=
1036 p->signal->rlim[RLIMIT_NPROC].rlim_cur) {
1037 if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
1038 p->user != &root_user)
1039 goto bad_fork_cleanup_vm;
1042 atomic_inc(&p->user->__count);
1043 atomic_inc(&p->user->processes);
1044 get_group_info(p->group_info);
1047 * If multiple threads are within copy_process(), then this check
1048 * triggers too late. This doesn't hurt, the check is only there
1049 * to stop root fork bombs.
1051 if (nr_threads >= max_threads)
1052 goto bad_fork_cleanup_count;
1054 if (!try_module_get(task_thread_info(p)->exec_domain->module))
1055 goto bad_fork_cleanup_count;
1057 if (p->binfmt && !try_module_get(p->binfmt->module))
1058 goto bad_fork_cleanup_put_domain;
1061 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
1062 copy_flags(clone_flags, p);
1065 if (clone_flags & CLONE_PARENT_SETTID)
1066 if (put_user(p->pid, parent_tidptr))
1067 goto bad_fork_cleanup_delays_binfmt;
1069 INIT_LIST_HEAD(&p->children);
1070 INIT_LIST_HEAD(&p->sibling);
1071 p->vfork_done = NULL;
1072 spin_lock_init(&p->alloc_lock);
1073 ptrace_init_task(p);
1075 clear_tsk_thread_flag(p, TIF_SIGPENDING);
1076 init_sigpending(&p->pending);
1078 p->utime = cputime_zero;
1079 p->stime = cputime_zero;
1081 p->rchar = 0; /* I/O counter: bytes read */
1082 p->wchar = 0; /* I/O counter: bytes written */
1083 p->syscr = 0; /* I/O counter: read syscalls */
1084 p->syscw = 0; /* I/O counter: write syscalls */
1085 task_io_accounting_init(p);
1086 acct_clear_integrals(p);
1088 p->it_virt_expires = cputime_zero;
1089 p->it_prof_expires = cputime_zero;
1090 p->it_sched_expires = 0;
1091 INIT_LIST_HEAD(&p->cpu_timers[0]);
1092 INIT_LIST_HEAD(&p->cpu_timers[1]);
1093 INIT_LIST_HEAD(&p->cpu_timers[2]);
1095 p->lock_depth = -1; /* -1 = no lock */
1096 do_posix_clock_monotonic_gettime(&p->start_time);
1098 p->io_context = NULL;
1100 p->audit_context = NULL;
1103 p->mempolicy = mpol_copy(p->mempolicy);
1104 if (IS_ERR(p->mempolicy)) {
1105 retval = PTR_ERR(p->mempolicy);
1106 p->mempolicy = NULL;
1107 goto bad_fork_cleanup_cpuset;
1109 mpol_fix_fork_child_flag(p);
1111 #ifdef CONFIG_TRACE_IRQFLAGS
1113 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1114 p->hardirqs_enabled = 1;
1116 p->hardirqs_enabled = 0;
1118 p->hardirq_enable_ip = 0;
1119 p->hardirq_enable_event = 0;
1120 p->hardirq_disable_ip = _THIS_IP_;
1121 p->hardirq_disable_event = 0;
1122 p->softirqs_enabled = 1;
1123 p->softirq_enable_ip = _THIS_IP_;
1124 p->softirq_enable_event = 0;
1125 p->softirq_disable_ip = 0;
1126 p->softirq_disable_event = 0;
1127 p->hardirq_context = 0;
1128 p->softirq_context = 0;
1130 #ifdef CONFIG_LOCKDEP
1131 p->lockdep_depth = 0; /* no locks held yet */
1132 p->curr_chain_key = 0;
1133 p->lockdep_recursion = 0;
1136 #ifdef CONFIG_DEBUG_MUTEXES
1137 p->blocked_on = NULL; /* not blocked yet */
1141 if (clone_flags & CLONE_THREAD)
1142 p->tgid = current->tgid;
1144 if ((retval = security_task_alloc(p)))
1145 goto bad_fork_cleanup_policy;
1146 if ((retval = audit_alloc(p)))
1147 goto bad_fork_cleanup_security;
1148 /* copy all the process information */
1149 if ((retval = copy_semundo(clone_flags, p)))
1150 goto bad_fork_cleanup_audit;
1151 if ((retval = copy_files(clone_flags, p)))
1152 goto bad_fork_cleanup_semundo;
1153 if ((retval = copy_fs(clone_flags, p)))
1154 goto bad_fork_cleanup_files;
1155 if ((retval = copy_sighand(clone_flags, p)))
1156 goto bad_fork_cleanup_fs;
1157 if ((retval = copy_signal(clone_flags, p)))
1158 goto bad_fork_cleanup_sighand;
1159 if ((retval = copy_mm(clone_flags, p)))
1160 goto bad_fork_cleanup_signal;
1161 if ((retval = copy_keys(clone_flags, p)))
1162 goto bad_fork_cleanup_mm;
1163 if ((retval = copy_namespaces(clone_flags, p)))
1164 goto bad_fork_cleanup_keys;
1165 retval = copy_thread(0, clone_flags, stack_start, stack_size, p, regs);
1167 goto bad_fork_cleanup_namespaces;
1169 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1171 * Clear TID on mm_release()?
1173 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL;
1174 p->robust_list = NULL;
1175 #ifdef CONFIG_COMPAT
1176 p->compat_robust_list = NULL;
1178 INIT_LIST_HEAD(&p->pi_state_list);
1179 p->pi_state_cache = NULL;
1182 * sigaltstack should be cleared when sharing the same VM
1184 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1185 p->sas_ss_sp = p->sas_ss_size = 0;
1188 * Syscall tracing should be turned off in the child regardless
1191 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
1192 #ifdef TIF_SYSCALL_EMU
1193 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1196 /* Our parent execution domain becomes current domain
1197 These must match for thread signalling to apply */
1198 p->parent_exec_id = p->self_exec_id;
1200 /* ok, now we should be set up.. */
1201 p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
1202 p->pdeath_signal = 0;
1206 * Ok, make it visible to the rest of the system.
1207 * We dont wake it up yet.
1209 p->group_leader = p;
1210 INIT_LIST_HEAD(&p->thread_group);
1212 /* Perform scheduler related setup. Assign this task to a CPU. */
1213 sched_fork(p, clone_flags);
1215 /* Need tasklist lock for parent etc handling! */
1216 write_lock_irq(&tasklist_lock);
1218 /* for sys_ioprio_set(IOPRIO_WHO_PGRP) */
1219 p->ioprio = current->ioprio;
1222 * The task hasn't been attached yet, so its cpus_allowed mask will
1223 * not be changed, nor will its assigned CPU.
1225 * The cpus_allowed mask of the parent may have changed after it was
1226 * copied first time - so re-copy it here, then check the child's CPU
1227 * to ensure it is on a valid CPU (and if not, just force it back to
1228 * parent's CPU). This avoids alot of nasty races.
1230 p->cpus_allowed = current->cpus_allowed;
1231 if (unlikely(!cpu_isset(task_cpu(p), p->cpus_allowed) ||
1232 !cpu_online(task_cpu(p))))
1233 set_task_cpu(p, smp_processor_id());
1235 /* CLONE_PARENT re-uses the old parent */
1236 if (clone_flags & (CLONE_PARENT|CLONE_THREAD))
1237 p->parent = current->parent;
1239 p->parent = current;
1241 spin_lock(¤t->sighand->siglock);
1244 * Process group and session signals need to be delivered to just the
1245 * parent before the fork or both the parent and the child after the
1246 * fork. Restart if a signal comes in before we add the new process to
1247 * it's process group.
1248 * A fatal signal pending means that current will exit, so the new
1249 * thread can't slip out of an OOM kill (or normal SIGKILL).
1251 recalc_sigpending();
1252 if (signal_pending(current)) {
1253 spin_unlock(¤t->sighand->siglock);
1254 write_unlock_irq(&tasklist_lock);
1255 retval = -ERESTARTNOINTR;
1256 goto bad_fork_cleanup_namespaces;
1259 if (clone_flags & CLONE_THREAD) {
1260 p->group_leader = current->group_leader;
1261 list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
1263 if (!cputime_eq(current->signal->it_virt_expires,
1265 !cputime_eq(current->signal->it_prof_expires,
1267 current->signal->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY ||
1268 !list_empty(¤t->signal->cpu_timers[0]) ||
1269 !list_empty(¤t->signal->cpu_timers[1]) ||
1270 !list_empty(¤t->signal->cpu_timers[2])) {
1272 * Have child wake up on its first tick to check
1273 * for process CPU timers.
1275 p->it_prof_expires = jiffies_to_cputime(1);
1279 if (likely(p->pid)) {
1281 tracehook_init_task(p);
1283 if (thread_group_leader(p)) {
1284 p->signal->tty = current->signal->tty;
1285 p->signal->pgrp = process_group(current);
1286 set_signal_session(p->signal, process_session(current));
1287 attach_pid(p, PIDTYPE_PGID, process_group(p));
1288 attach_pid(p, PIDTYPE_SID, process_session(p));
1290 list_add_tail_rcu(&p->tasks, &init_task.tasks);
1291 __get_cpu_var(process_counts)++;
1293 attach_pid(p, PIDTYPE_PID, p->pid);
1298 spin_unlock(¤t->sighand->siglock);
1300 /* p is copy of current */
1303 claim_vx_info(vxi, p);
1304 atomic_inc(&vxi->cvirt.nr_threads);
1305 atomic_inc(&vxi->cvirt.total_forks);
1310 claim_nx_info(nxi, p);
1311 write_unlock_irq(&tasklist_lock);
1312 proc_fork_connector(p);
1315 bad_fork_cleanup_namespaces:
1316 exit_task_namespaces(p);
1317 bad_fork_cleanup_keys:
1319 bad_fork_cleanup_mm:
1322 bad_fork_cleanup_signal:
1324 bad_fork_cleanup_sighand:
1325 __cleanup_sighand(p->sighand);
1326 bad_fork_cleanup_fs:
1327 exit_fs(p); /* blocking */
1328 bad_fork_cleanup_files:
1329 exit_files(p); /* blocking */
1330 bad_fork_cleanup_semundo:
1332 bad_fork_cleanup_audit:
1334 bad_fork_cleanup_security:
1335 security_task_free(p);
1336 bad_fork_cleanup_policy:
1338 mpol_free(p->mempolicy);
1339 bad_fork_cleanup_cpuset:
1342 bad_fork_cleanup_delays_binfmt:
1343 delayacct_tsk_free(p);
1345 module_put(p->binfmt->module);
1346 bad_fork_cleanup_put_domain:
1347 module_put(task_thread_info(p)->exec_domain->module);
1348 bad_fork_cleanup_count:
1349 put_group_info(p->group_info);
1350 atomic_dec(&p->user->processes);
1352 bad_fork_cleanup_vm:
1353 if (p->mm && !(clone_flags & CLONE_VM))
1354 vx_pages_sub(p->vx_info, RLIMIT_AS, p->mm->total_vm);
1358 return ERR_PTR(retval);
1361 noinline struct pt_regs * __devinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
1363 memset(regs, 0, sizeof(struct pt_regs));
1367 struct task_struct * __cpuinit fork_idle(int cpu)
1369 struct task_struct *task;
1370 struct pt_regs regs;
1372 task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL, NULL, 0);
1374 init_idle(task, cpu);
1380 * Ok, this is the main fork-routine.
1382 * It copies the process, and if successful kick-starts
1383 * it and waits for it to finish using the VM if required.
1385 long do_fork(unsigned long clone_flags,
1386 unsigned long stack_start,
1387 struct pt_regs *regs,
1388 unsigned long stack_size,
1389 int __user *parent_tidptr,
1390 int __user *child_tidptr)
1392 struct task_struct *p;
1393 struct pid *pid = alloc_pid();
1399 /* kernel threads are host only */
1400 if ((clone_flags & CLONE_KTHREAD) && !vx_check(0, VS_ADMIN)) {
1401 vxwprintk(1, "xid=%d tried to spawn a kernel thread.",
1409 p = copy_process(clone_flags, stack_start, regs, stack_size, parent_tidptr, child_tidptr, nr);
1411 * Do this prior waking up the new thread - the thread pointer
1412 * might get invalid after that point, if the thread exits quickly.
1416 * When called from kernel_thread, don't do user tracing stuff.
1418 int is_user = likely(user_mode(regs));
1419 struct completion vfork;
1421 if (clone_flags & CLONE_VFORK) {
1422 p->vfork_done = &vfork;
1423 init_completion(&vfork);
1426 if (likely(is_user))
1427 tracehook_report_clone(clone_flags, p);
1429 p->flags &= ~PF_STARTING;
1431 if (clone_flags & CLONE_STOPPED) {
1433 * We'll start up with an immediate SIGSTOP.
1435 sigaddset(&p->pending.signal, SIGSTOP);
1436 set_tsk_thread_flag(p, TIF_SIGPENDING);
1437 p->state = TASK_STOPPED;
1440 wake_up_new_task(p, clone_flags);
1442 if (likely(is_user))
1443 tracehook_report_clone_complete(clone_flags, nr, p);
1445 if (clone_flags & CLONE_VFORK) {
1446 wait_for_completion(&vfork);
1447 if (likely(is_user))
1448 tracehook_report_vfork_done(p, nr);
1457 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1458 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1461 static void sighand_ctor(void *data, struct kmem_cache *cachep, unsigned long flags)
1463 struct sighand_struct *sighand = data;
1465 if ((flags & (SLAB_CTOR_VERIFY | SLAB_CTOR_CONSTRUCTOR)) ==
1466 SLAB_CTOR_CONSTRUCTOR)
1467 spin_lock_init(&sighand->siglock);
1470 void __init proc_caches_init(void)
1472 sighand_cachep = kmem_cache_create("sighand_cache",
1473 sizeof(struct sighand_struct), 0,
1474 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU,
1475 sighand_ctor, NULL);
1476 signal_cachep = kmem_cache_create("signal_cache",
1477 sizeof(struct signal_struct), 0,
1478 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
1479 files_cachep = kmem_cache_create("files_cache",
1480 sizeof(struct files_struct), 0,
1481 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
1482 fs_cachep = kmem_cache_create("fs_cache",
1483 sizeof(struct fs_struct), 0,
1484 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
1485 vm_area_cachep = kmem_cache_create("vm_area_struct",
1486 sizeof(struct vm_area_struct), 0,
1487 SLAB_PANIC, NULL, NULL);
1488 mm_cachep = kmem_cache_create("mm_struct",
1489 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
1490 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
1495 * Check constraints on flags passed to the unshare system call and
1496 * force unsharing of additional process context as appropriate.
1498 static inline void check_unshare_flags(unsigned long *flags_ptr)
1501 * If unsharing a thread from a thread group, must also
1504 if (*flags_ptr & CLONE_THREAD)
1505 *flags_ptr |= CLONE_VM;
1508 * If unsharing vm, must also unshare signal handlers.
1510 if (*flags_ptr & CLONE_VM)
1511 *flags_ptr |= CLONE_SIGHAND;
1514 * If unsharing signal handlers and the task was created
1515 * using CLONE_THREAD, then must unshare the thread
1517 if ((*flags_ptr & CLONE_SIGHAND) &&
1518 (atomic_read(¤t->signal->count) > 1))
1519 *flags_ptr |= CLONE_THREAD;
1522 * If unsharing namespace, must also unshare filesystem information.
1524 if (*flags_ptr & CLONE_NEWNS)
1525 *flags_ptr |= CLONE_FS;
1529 * Unsharing of tasks created with CLONE_THREAD is not supported yet
1531 static int unshare_thread(unsigned long unshare_flags)
1533 if (unshare_flags & CLONE_THREAD)
1540 * Unshare the filesystem structure if it is being shared
1542 static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1544 struct fs_struct *fs = current->fs;
1546 if ((unshare_flags & CLONE_FS) &&
1547 (fs && atomic_read(&fs->count) > 1)) {
1548 *new_fsp = __copy_fs_struct(current->fs);
1557 * Unshare the mnt_namespace structure if it is being shared
1559 static int unshare_mnt_namespace(unsigned long unshare_flags,
1560 struct mnt_namespace **new_nsp, struct fs_struct *new_fs)
1562 struct mnt_namespace *ns = current->nsproxy->mnt_ns;
1564 if ((unshare_flags & CLONE_NEWNS) && ns) {
1565 if (!capable(CAP_SYS_ADMIN))
1568 *new_nsp = dup_mnt_ns(current, new_fs ? new_fs : current->fs);
1577 * Unsharing of sighand is not supported yet
1579 static int unshare_sighand(unsigned long unshare_flags, struct sighand_struct **new_sighp)
1581 struct sighand_struct *sigh = current->sighand;
1583 if ((unshare_flags & CLONE_SIGHAND) && atomic_read(&sigh->count) > 1)
1590 * Unshare vm if it is being shared
1592 static int unshare_vm(unsigned long unshare_flags, struct mm_struct **new_mmp)
1594 struct mm_struct *mm = current->mm;
1596 if ((unshare_flags & CLONE_VM) &&
1597 (mm && atomic_read(&mm->mm_users) > 1)) {
1605 * Unshare file descriptor table if it is being shared
1607 static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1609 struct files_struct *fd = current->files;
1612 if ((unshare_flags & CLONE_FILES) &&
1613 (fd && atomic_read(&fd->count) > 1)) {
1614 *new_fdp = dup_fd(fd, &error);
1623 * Unsharing of semundo for tasks created with CLONE_SYSVSEM is not
1626 static int unshare_semundo(unsigned long unshare_flags, struct sem_undo_list **new_ulistp)
1628 if (unshare_flags & CLONE_SYSVSEM)
1634 #ifndef CONFIG_IPC_NS
1635 static inline int unshare_ipcs(unsigned long flags, struct ipc_namespace **ns)
1637 if (flags & CLONE_NEWIPC)
1645 * unshare allows a process to 'unshare' part of the process
1646 * context which was originally shared using clone. copy_*
1647 * functions used by do_fork() cannot be used here directly
1648 * because they modify an inactive task_struct that is being
1649 * constructed. Here we are modifying the current, active,
1652 asmlinkage long sys_unshare(unsigned long unshare_flags)
1655 struct fs_struct *fs, *new_fs = NULL;
1656 struct mnt_namespace *ns, *new_ns = NULL;
1657 struct sighand_struct *new_sigh = NULL;
1658 struct mm_struct *mm, *new_mm = NULL, *active_mm = NULL;
1659 struct files_struct *fd, *new_fd = NULL;
1660 struct sem_undo_list *new_ulist = NULL;
1661 struct nsproxy *new_nsproxy = NULL, *old_nsproxy = NULL;
1662 struct uts_namespace *uts, *new_uts = NULL;
1663 struct ipc_namespace *ipc, *new_ipc = NULL;
1665 check_unshare_flags(&unshare_flags);
1667 /* Return -EINVAL for all unsupported flags */
1669 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
1670 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
1671 CLONE_NEWUTS|CLONE_NEWIPC))
1672 goto bad_unshare_out;
1674 if ((err = unshare_thread(unshare_flags)))
1675 goto bad_unshare_out;
1676 if ((err = unshare_fs(unshare_flags, &new_fs)))
1677 goto bad_unshare_cleanup_thread;
1678 if ((err = unshare_mnt_namespace(unshare_flags, &new_ns, new_fs)))
1679 goto bad_unshare_cleanup_fs;
1680 if ((err = unshare_sighand(unshare_flags, &new_sigh)))
1681 goto bad_unshare_cleanup_ns;
1682 if ((err = unshare_vm(unshare_flags, &new_mm)))
1683 goto bad_unshare_cleanup_sigh;
1684 if ((err = unshare_fd(unshare_flags, &new_fd)))
1685 goto bad_unshare_cleanup_vm;
1686 if ((err = unshare_semundo(unshare_flags, &new_ulist)))
1687 goto bad_unshare_cleanup_fd;
1688 if ((err = unshare_utsname(unshare_flags, &new_uts)))
1689 goto bad_unshare_cleanup_semundo;
1690 if ((err = unshare_ipcs(unshare_flags, &new_ipc)))
1691 goto bad_unshare_cleanup_uts;
1693 if (new_ns || new_uts || new_ipc) {
1694 old_nsproxy = current->nsproxy;
1695 new_nsproxy = dup_namespaces(old_nsproxy);
1698 goto bad_unshare_cleanup_ipc;
1702 if (new_fs || new_ns || new_mm || new_fd || new_ulist ||
1703 new_uts || new_ipc) {
1708 current->nsproxy = new_nsproxy;
1709 new_nsproxy = old_nsproxy;
1714 current->fs = new_fs;
1719 ns = current->nsproxy->mnt_ns;
1720 current->nsproxy->mnt_ns = new_ns;
1726 active_mm = current->active_mm;
1727 current->mm = new_mm;
1728 current->active_mm = new_mm;
1729 activate_mm(active_mm, new_mm);
1734 fd = current->files;
1735 current->files = new_fd;
1740 uts = current->nsproxy->uts_ns;
1741 current->nsproxy->uts_ns = new_uts;
1746 ipc = current->nsproxy->ipc_ns;
1747 current->nsproxy->ipc_ns = new_ipc;
1751 task_unlock(current);
1755 put_nsproxy(new_nsproxy);
1757 bad_unshare_cleanup_ipc:
1759 put_ipc_ns(new_ipc);
1761 bad_unshare_cleanup_uts:
1763 put_uts_ns(new_uts);
1765 bad_unshare_cleanup_semundo:
1766 bad_unshare_cleanup_fd:
1768 put_files_struct(new_fd);
1770 bad_unshare_cleanup_vm:
1774 bad_unshare_cleanup_sigh:
1776 if (atomic_dec_and_test(&new_sigh->count))
1777 kmem_cache_free(sighand_cachep, new_sigh);
1779 bad_unshare_cleanup_ns:
1783 bad_unshare_cleanup_fs:
1785 put_fs_struct(new_fs);
1787 bad_unshare_cleanup_thread: