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/ptrace.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>
53 #include <asm/pgtable.h>
54 #include <asm/pgalloc.h>
55 #include <asm/uaccess.h>
56 #include <asm/mmu_context.h>
57 #include <asm/cacheflush.h>
58 #include <asm/tlbflush.h>
61 * Protected counters by write_lock_irq(&tasklist_lock)
63 unsigned long total_forks
; /* Handle normal Linux uptimes. */
64 int nr_threads
; /* The idle threads do not count.. */
66 int max_threads
; /* tunable limit on nr_threads */
68 DEFINE_PER_CPU(unsigned long, process_counts
) = 0;
70 __cacheline_aligned
DEFINE_RWLOCK(tasklist_lock
); /* outer */
72 int nr_processes(void)
77 for_each_online_cpu(cpu
)
78 total
+= per_cpu(process_counts
, cpu
);
83 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
84 # define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL)
85 # define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk))
86 static struct kmem_cache
*task_struct_cachep
;
89 /* SLAB cache for signal_struct structures (tsk->signal) */
90 static struct kmem_cache
*signal_cachep
;
92 /* SLAB cache for sighand_struct structures (tsk->sighand) */
93 struct kmem_cache
*sighand_cachep
;
95 /* SLAB cache for files_struct structures (tsk->files) */
96 struct kmem_cache
*files_cachep
;
98 /* SLAB cache for fs_struct structures (tsk->fs) */
99 struct kmem_cache
*fs_cachep
;
101 /* SLAB cache for vm_area_struct structures */
102 struct kmem_cache
*vm_area_cachep
;
104 /* SLAB cache for mm_struct structures (tsk->mm) */
105 static struct kmem_cache
*mm_cachep
;
107 void free_task(struct task_struct
*tsk
)
109 free_thread_info(tsk
->thread_info
);
110 rt_mutex_debug_task_free(tsk
);
111 free_task_struct(tsk
);
113 EXPORT_SYMBOL(free_task
);
115 void __put_task_struct(struct task_struct
*tsk
)
117 WARN_ON(!(tsk
->exit_state
& (EXIT_DEAD
| EXIT_ZOMBIE
)));
118 WARN_ON(atomic_read(&tsk
->usage
));
119 WARN_ON(tsk
== current
);
121 security_task_free(tsk
);
123 put_group_info(tsk
->group_info
);
124 delayacct_tsk_free(tsk
);
126 if (!profile_handoff_task(tsk
))
130 void __init
fork_init(unsigned long mempages
)
132 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
133 #ifndef ARCH_MIN_TASKALIGN
134 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
136 /* create a slab on which task_structs can be allocated */
138 kmem_cache_create("task_struct", sizeof(struct task_struct
),
139 ARCH_MIN_TASKALIGN
, SLAB_PANIC
, NULL
, NULL
);
143 * The default maximum number of threads is set to a safe
144 * value: the thread structures can take up at most half
147 max_threads
= mempages
/ (8 * THREAD_SIZE
/ PAGE_SIZE
);
150 * we need to allow at least 20 threads to boot a system
155 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_cur
= max_threads
/2;
156 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_max
= max_threads
/2;
157 init_task
.signal
->rlim
[RLIMIT_SIGPENDING
] =
158 init_task
.signal
->rlim
[RLIMIT_NPROC
];
161 static struct task_struct
*dup_task_struct(struct task_struct
*orig
)
163 struct task_struct
*tsk
;
164 struct thread_info
*ti
;
166 prepare_to_copy(orig
);
168 tsk
= alloc_task_struct();
172 ti
= alloc_thread_info(tsk
);
174 free_task_struct(tsk
);
179 tsk
->thread_info
= ti
;
180 setup_thread_stack(tsk
, orig
);
182 #ifdef CONFIG_CC_STACKPROTECTOR
183 tsk
->stack_canary
= get_random_int();
186 /* One for us, one for whoever does the "release_task()" (usually parent) */
187 atomic_set(&tsk
->usage
,2);
188 atomic_set(&tsk
->fs_excl
, 0);
189 #ifdef CONFIG_BLK_DEV_IO_TRACE
192 tsk
->splice_pipe
= NULL
;
197 static inline int dup_mmap(struct mm_struct
*mm
, struct mm_struct
*oldmm
)
199 struct vm_area_struct
*mpnt
, *tmp
, **pprev
;
200 struct rb_node
**rb_link
, *rb_parent
;
202 unsigned long charge
;
203 struct mempolicy
*pol
;
205 down_write(&oldmm
->mmap_sem
);
206 flush_cache_mm(oldmm
);
208 * Not linked in yet - no deadlock potential:
210 down_write_nested(&mm
->mmap_sem
, SINGLE_DEPTH_NESTING
);
214 mm
->mmap_cache
= NULL
;
215 mm
->free_area_cache
= oldmm
->mmap_base
;
216 mm
->cached_hole_size
= ~0UL;
218 cpus_clear(mm
->cpu_vm_mask
);
220 rb_link
= &mm
->mm_rb
.rb_node
;
224 for (mpnt
= oldmm
->mmap
; mpnt
; mpnt
= mpnt
->vm_next
) {
227 if (mpnt
->vm_flags
& VM_DONTCOPY
) {
228 long pages
= vma_pages(mpnt
);
229 mm
->total_vm
-= pages
;
230 vm_stat_account(mm
, mpnt
->vm_flags
, mpnt
->vm_file
,
235 if (mpnt
->vm_flags
& VM_ACCOUNT
) {
236 unsigned int len
= (mpnt
->vm_end
- mpnt
->vm_start
) >> PAGE_SHIFT
;
237 if (security_vm_enough_memory(len
))
241 tmp
= kmem_cache_alloc(vm_area_cachep
, GFP_KERNEL
);
245 pol
= mpol_copy(vma_policy(mpnt
));
246 retval
= PTR_ERR(pol
);
248 goto fail_nomem_policy
;
249 vma_set_policy(tmp
, pol
);
250 tmp
->vm_flags
&= ~VM_LOCKED
;
256 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
258 if (tmp
->vm_flags
& VM_DENYWRITE
)
259 atomic_dec(&inode
->i_writecount
);
261 /* insert tmp into the share list, just after mpnt */
262 spin_lock(&file
->f_mapping
->i_mmap_lock
);
263 tmp
->vm_truncate_count
= mpnt
->vm_truncate_count
;
264 flush_dcache_mmap_lock(file
->f_mapping
);
265 vma_prio_tree_add(tmp
, mpnt
);
266 flush_dcache_mmap_unlock(file
->f_mapping
);
267 spin_unlock(&file
->f_mapping
->i_mmap_lock
);
271 * Link in the new vma and copy the page table entries.
274 pprev
= &tmp
->vm_next
;
276 __vma_link_rb(mm
, tmp
, rb_link
, rb_parent
);
277 rb_link
= &tmp
->vm_rb
.rb_right
;
278 rb_parent
= &tmp
->vm_rb
;
281 retval
= copy_page_range(mm
, oldmm
, mpnt
);
283 if (tmp
->vm_ops
&& tmp
->vm_ops
->open
)
284 tmp
->vm_ops
->open(tmp
);
291 up_write(&mm
->mmap_sem
);
293 up_write(&oldmm
->mmap_sem
);
296 kmem_cache_free(vm_area_cachep
, tmp
);
299 vm_unacct_memory(charge
);
303 static inline int mm_alloc_pgd(struct mm_struct
* mm
)
305 mm
->pgd
= pgd_alloc(mm
);
306 if (unlikely(!mm
->pgd
))
311 static inline void mm_free_pgd(struct mm_struct
* mm
)
316 #define dup_mmap(mm, oldmm) (0)
317 #define mm_alloc_pgd(mm) (0)
318 #define mm_free_pgd(mm)
319 #endif /* CONFIG_MMU */
321 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(mmlist_lock
);
323 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
324 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
326 #include <linux/init_task.h>
328 static struct mm_struct
* mm_init(struct mm_struct
* mm
)
330 atomic_set(&mm
->mm_users
, 1);
331 atomic_set(&mm
->mm_count
, 1);
332 init_rwsem(&mm
->mmap_sem
);
333 INIT_LIST_HEAD(&mm
->mmlist
);
334 mm
->core_waiters
= 0;
336 set_mm_counter(mm
, file_rss
, 0);
337 set_mm_counter(mm
, anon_rss
, 0);
338 spin_lock_init(&mm
->page_table_lock
);
339 rwlock_init(&mm
->ioctx_list_lock
);
340 mm
->ioctx_list
= NULL
;
341 mm
->free_area_cache
= TASK_UNMAPPED_BASE
;
342 mm
->cached_hole_size
= ~0UL;
344 if (likely(!mm_alloc_pgd(mm
))) {
353 * Allocate and initialize an mm_struct.
355 struct mm_struct
* mm_alloc(void)
357 struct mm_struct
* mm
;
361 memset(mm
, 0, sizeof(*mm
));
368 * Called when the last reference to the mm
369 * is dropped: either by a lazy thread or by
370 * mmput. Free the page directory and the mm.
372 void fastcall
__mmdrop(struct mm_struct
*mm
)
374 BUG_ON(mm
== &init_mm
);
381 * Decrement the use count and release all resources for an mm.
383 void mmput(struct mm_struct
*mm
)
387 if (atomic_dec_and_test(&mm
->mm_users
)) {
390 if (!list_empty(&mm
->mmlist
)) {
391 spin_lock(&mmlist_lock
);
392 list_del(&mm
->mmlist
);
393 spin_unlock(&mmlist_lock
);
399 EXPORT_SYMBOL_GPL(mmput
);
402 * get_task_mm - acquire a reference to the task's mm
404 * Returns %NULL if the task has no mm. Checks PF_BORROWED_MM (meaning
405 * this kernel workthread has transiently adopted a user mm with use_mm,
406 * to do its AIO) is not set and if so returns a reference to it, after
407 * bumping up the use count. User must release the mm via mmput()
408 * after use. Typically used by /proc and ptrace.
410 struct mm_struct
*get_task_mm(struct task_struct
*task
)
412 struct mm_struct
*mm
;
417 if (task
->flags
& PF_BORROWED_MM
)
420 atomic_inc(&mm
->mm_users
);
425 EXPORT_SYMBOL_GPL(get_task_mm
);
427 /* Please note the differences between mmput and mm_release.
428 * mmput is called whenever we stop holding onto a mm_struct,
429 * error success whatever.
431 * mm_release is called after a mm_struct has been removed
432 * from the current process.
434 * This difference is important for error handling, when we
435 * only half set up a mm_struct for a new process and need to restore
436 * the old one. Because we mmput the new mm_struct before
437 * restoring the old one. . .
438 * Eric Biederman 10 January 1998
440 void mm_release(struct task_struct
*tsk
, struct mm_struct
*mm
)
442 struct completion
*vfork_done
= tsk
->vfork_done
;
444 /* Get rid of any cached register state */
445 deactivate_mm(tsk
, mm
);
447 /* notify parent sleeping on vfork() */
449 tsk
->vfork_done
= NULL
;
450 complete(vfork_done
);
454 * If we're exiting normally, clear a user-space tid field if
455 * requested. We leave this alone when dying by signal, to leave
456 * the value intact in a core dump, and to save the unnecessary
457 * trouble otherwise. Userland only wants this done for a sys_exit.
459 if (tsk
->clear_child_tid
460 && !(tsk
->flags
& PF_SIGNALED
)
461 && atomic_read(&mm
->mm_users
) > 1) {
462 u32 __user
* tidptr
= tsk
->clear_child_tid
;
463 tsk
->clear_child_tid
= NULL
;
466 * We don't check the error code - if userspace has
467 * not set up a proper pointer then tough luck.
470 sys_futex(tidptr
, FUTEX_WAKE
, 1, NULL
, NULL
, 0);
475 * Allocate a new mm structure and copy contents from the
476 * mm structure of the passed in task structure.
478 static struct mm_struct
*dup_mm(struct task_struct
*tsk
)
480 struct mm_struct
*mm
, *oldmm
= current
->mm
;
490 memcpy(mm
, oldmm
, sizeof(*mm
));
492 /* Initializing for Swap token stuff */
493 mm
->token_priority
= 0;
494 mm
->last_interval
= 0;
499 if (init_new_context(tsk
, mm
))
502 err
= dup_mmap(mm
, oldmm
);
506 mm
->hiwater_rss
= get_mm_rss(mm
);
507 mm
->hiwater_vm
= mm
->total_vm
;
519 * If init_new_context() failed, we cannot use mmput() to free the mm
520 * because it calls destroy_context()
527 static int copy_mm(unsigned long clone_flags
, struct task_struct
* tsk
)
529 struct mm_struct
* mm
, *oldmm
;
532 tsk
->min_flt
= tsk
->maj_flt
= 0;
533 tsk
->nvcsw
= tsk
->nivcsw
= 0;
536 tsk
->active_mm
= NULL
;
539 * Are we cloning a kernel thread?
541 * We need to steal a active VM for that..
547 if (clone_flags
& CLONE_VM
) {
548 atomic_inc(&oldmm
->mm_users
);
559 /* Initializing for Swap token stuff */
560 mm
->token_priority
= 0;
561 mm
->last_interval
= 0;
571 static inline struct fs_struct
*__copy_fs_struct(struct fs_struct
*old
)
573 struct fs_struct
*fs
= kmem_cache_alloc(fs_cachep
, GFP_KERNEL
);
574 /* We don't need to lock fs - think why ;-) */
576 atomic_set(&fs
->count
, 1);
577 rwlock_init(&fs
->lock
);
578 fs
->umask
= old
->umask
;
579 read_lock(&old
->lock
);
580 fs
->rootmnt
= mntget(old
->rootmnt
);
581 fs
->root
= dget(old
->root
);
582 fs
->pwdmnt
= mntget(old
->pwdmnt
);
583 fs
->pwd
= dget(old
->pwd
);
585 fs
->altrootmnt
= mntget(old
->altrootmnt
);
586 fs
->altroot
= dget(old
->altroot
);
588 fs
->altrootmnt
= NULL
;
591 read_unlock(&old
->lock
);
596 struct fs_struct
*copy_fs_struct(struct fs_struct
*old
)
598 return __copy_fs_struct(old
);
601 EXPORT_SYMBOL_GPL(copy_fs_struct
);
603 static inline int copy_fs(unsigned long clone_flags
, struct task_struct
* tsk
)
605 if (clone_flags
& CLONE_FS
) {
606 atomic_inc(¤t
->fs
->count
);
609 tsk
->fs
= __copy_fs_struct(current
->fs
);
615 static int count_open_files(struct fdtable
*fdt
)
617 int size
= fdt
->max_fds
;
620 /* Find the last open fd */
621 for (i
= size
/(8*sizeof(long)); i
> 0; ) {
622 if (fdt
->open_fds
->fds_bits
[--i
])
625 i
= (i
+1) * 8 * sizeof(long);
629 static struct files_struct
*alloc_files(void)
631 struct files_struct
*newf
;
634 newf
= kmem_cache_alloc(files_cachep
, GFP_KERNEL
);
638 atomic_set(&newf
->count
, 1);
640 spin_lock_init(&newf
->file_lock
);
643 fdt
->max_fds
= NR_OPEN_DEFAULT
;
644 fdt
->close_on_exec
= (fd_set
*)&newf
->close_on_exec_init
;
645 fdt
->open_fds
= (fd_set
*)&newf
->open_fds_init
;
646 fdt
->fd
= &newf
->fd_array
[0];
647 INIT_RCU_HEAD(&fdt
->rcu
);
648 fdt
->free_files
= NULL
;
650 rcu_assign_pointer(newf
->fdt
, fdt
);
656 * Allocate a new files structure and copy contents from the
657 * passed in files structure.
658 * errorp will be valid only when the returned files_struct is NULL.
660 static struct files_struct
*dup_fd(struct files_struct
*oldf
, int *errorp
)
662 struct files_struct
*newf
;
663 struct file
**old_fds
, **new_fds
;
664 int open_files
, size
, i
;
665 struct fdtable
*old_fdt
, *new_fdt
;
668 newf
= alloc_files();
672 spin_lock(&oldf
->file_lock
);
673 old_fdt
= files_fdtable(oldf
);
674 new_fdt
= files_fdtable(newf
);
675 open_files
= count_open_files(old_fdt
);
678 * Check whether we need to allocate a larger fd array and fd set.
679 * Note: we're not a clone task, so the open count won't change.
681 if (open_files
> new_fdt
->max_fds
) {
682 new_fdt
->max_fds
= 0;
683 spin_unlock(&oldf
->file_lock
);
684 spin_lock(&newf
->file_lock
);
685 *errorp
= expand_files(newf
, open_files
-1);
686 spin_unlock(&newf
->file_lock
);
689 new_fdt
= files_fdtable(newf
);
691 * Reacquire the oldf lock and a pointer to its fd table
692 * who knows it may have a new bigger fd table. We need
693 * the latest pointer.
695 spin_lock(&oldf
->file_lock
);
696 old_fdt
= files_fdtable(oldf
);
699 old_fds
= old_fdt
->fd
;
700 new_fds
= new_fdt
->fd
;
702 memcpy(new_fdt
->open_fds
->fds_bits
,
703 old_fdt
->open_fds
->fds_bits
, open_files
/8);
704 memcpy(new_fdt
->close_on_exec
->fds_bits
,
705 old_fdt
->close_on_exec
->fds_bits
, open_files
/8);
707 for (i
= open_files
; i
!= 0; i
--) {
708 struct file
*f
= *old_fds
++;
713 * The fd may be claimed in the fd bitmap but not yet
714 * instantiated in the files array if a sibling thread
715 * is partway through open(). So make sure that this
716 * fd is available to the new process.
718 FD_CLR(open_files
- i
, new_fdt
->open_fds
);
720 rcu_assign_pointer(*new_fds
++, f
);
722 spin_unlock(&oldf
->file_lock
);
724 /* compute the remainder to be cleared */
725 size
= (new_fdt
->max_fds
- open_files
) * sizeof(struct file
*);
727 /* This is long word aligned thus could use a optimized version */
728 memset(new_fds
, 0, size
);
730 if (new_fdt
->max_fds
> open_files
) {
731 int left
= (new_fdt
->max_fds
-open_files
)/8;
732 int start
= open_files
/ (8 * sizeof(unsigned long));
734 memset(&new_fdt
->open_fds
->fds_bits
[start
], 0, left
);
735 memset(&new_fdt
->close_on_exec
->fds_bits
[start
], 0, left
);
741 kmem_cache_free(files_cachep
, newf
);
746 static int copy_files(unsigned long clone_flags
, struct task_struct
* tsk
)
748 struct files_struct
*oldf
, *newf
;
752 * A background process may not have any files ...
754 oldf
= current
->files
;
758 if (clone_flags
& CLONE_FILES
) {
759 atomic_inc(&oldf
->count
);
764 * Note: we may be using current for both targets (See exec.c)
765 * This works because we cache current->files (old) as oldf. Don't
769 newf
= dup_fd(oldf
, &error
);
780 * Helper to unshare the files of the current task.
781 * We don't want to expose copy_files internals to
782 * the exec layer of the kernel.
785 int unshare_files(void)
787 struct files_struct
*files
= current
->files
;
792 /* This can race but the race causes us to copy when we don't
793 need to and drop the copy */
794 if(atomic_read(&files
->count
) == 1)
796 atomic_inc(&files
->count
);
799 rc
= copy_files(0, current
);
801 current
->files
= files
;
805 EXPORT_SYMBOL(unshare_files
);
807 static inline int copy_sighand(unsigned long clone_flags
, struct task_struct
* tsk
)
809 struct sighand_struct
*sig
;
811 if (clone_flags
& (CLONE_SIGHAND
| CLONE_THREAD
)) {
812 atomic_inc(¤t
->sighand
->count
);
815 sig
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
816 rcu_assign_pointer(tsk
->sighand
, sig
);
819 atomic_set(&sig
->count
, 1);
820 memcpy(sig
->action
, current
->sighand
->action
, sizeof(sig
->action
));
824 void __cleanup_sighand(struct sighand_struct
*sighand
)
826 if (atomic_dec_and_test(&sighand
->count
))
827 kmem_cache_free(sighand_cachep
, sighand
);
830 static inline int copy_signal(unsigned long clone_flags
, struct task_struct
* tsk
)
832 struct signal_struct
*sig
;
835 if (clone_flags
& CLONE_THREAD
) {
836 atomic_inc(¤t
->signal
->count
);
837 atomic_inc(¤t
->signal
->live
);
840 sig
= kmem_cache_alloc(signal_cachep
, GFP_KERNEL
);
845 ret
= copy_thread_group_keys(tsk
);
847 kmem_cache_free(signal_cachep
, sig
);
851 atomic_set(&sig
->count
, 1);
852 atomic_set(&sig
->live
, 1);
853 init_waitqueue_head(&sig
->wait_chldexit
);
855 sig
->group_exit_code
= 0;
856 sig
->group_exit_task
= NULL
;
857 sig
->group_stop_count
= 0;
858 sig
->curr_target
= NULL
;
859 init_sigpending(&sig
->shared_pending
);
860 INIT_LIST_HEAD(&sig
->posix_timers
);
862 hrtimer_init(&sig
->real_timer
, CLOCK_MONOTONIC
, HRTIMER_REL
);
863 sig
->it_real_incr
.tv64
= 0;
864 sig
->real_timer
.function
= it_real_fn
;
867 sig
->it_virt_expires
= cputime_zero
;
868 sig
->it_virt_incr
= cputime_zero
;
869 sig
->it_prof_expires
= cputime_zero
;
870 sig
->it_prof_incr
= cputime_zero
;
872 sig
->leader
= 0; /* session leadership doesn't inherit */
873 sig
->tty_old_pgrp
= 0;
875 sig
->utime
= sig
->stime
= sig
->cutime
= sig
->cstime
= cputime_zero
;
876 sig
->nvcsw
= sig
->nivcsw
= sig
->cnvcsw
= sig
->cnivcsw
= 0;
877 sig
->min_flt
= sig
->maj_flt
= sig
->cmin_flt
= sig
->cmaj_flt
= 0;
879 INIT_LIST_HEAD(&sig
->cpu_timers
[0]);
880 INIT_LIST_HEAD(&sig
->cpu_timers
[1]);
881 INIT_LIST_HEAD(&sig
->cpu_timers
[2]);
882 taskstats_tgid_init(sig
);
884 task_lock(current
->group_leader
);
885 memcpy(sig
->rlim
, current
->signal
->rlim
, sizeof sig
->rlim
);
886 task_unlock(current
->group_leader
);
888 if (sig
->rlim
[RLIMIT_CPU
].rlim_cur
!= RLIM_INFINITY
) {
890 * New sole thread in the process gets an expiry time
891 * of the whole CPU time limit.
893 tsk
->it_prof_expires
=
894 secs_to_cputime(sig
->rlim
[RLIMIT_CPU
].rlim_cur
);
896 acct_init_pacct(&sig
->pacct
);
901 void __cleanup_signal(struct signal_struct
*sig
)
903 exit_thread_group_keys(sig
);
904 kmem_cache_free(signal_cachep
, sig
);
907 static inline void cleanup_signal(struct task_struct
*tsk
)
909 struct signal_struct
*sig
= tsk
->signal
;
911 atomic_dec(&sig
->live
);
913 if (atomic_dec_and_test(&sig
->count
))
914 __cleanup_signal(sig
);
917 static inline void copy_flags(unsigned long clone_flags
, struct task_struct
*p
)
919 unsigned long new_flags
= p
->flags
;
921 new_flags
&= ~(PF_SUPERPRIV
| PF_NOFREEZE
);
922 new_flags
|= PF_FORKNOEXEC
;
923 if (!(clone_flags
& CLONE_PTRACE
))
925 p
->flags
= new_flags
;
928 asmlinkage
long sys_set_tid_address(int __user
*tidptr
)
930 current
->clear_child_tid
= tidptr
;
935 static inline void rt_mutex_init_task(struct task_struct
*p
)
937 #ifdef CONFIG_RT_MUTEXES
938 spin_lock_init(&p
->pi_lock
);
939 plist_head_init(&p
->pi_waiters
, &p
->pi_lock
);
940 p
->pi_blocked_on
= NULL
;
945 * This creates a new process as a copy of the old one,
946 * but does not actually start it yet.
948 * It copies the registers, and all the appropriate
949 * parts of the process environment (as per the clone
950 * flags). The actual kick-off is left to the caller.
952 static struct task_struct
*copy_process(unsigned long clone_flags
,
953 unsigned long stack_start
,
954 struct pt_regs
*regs
,
955 unsigned long stack_size
,
956 int __user
*parent_tidptr
,
957 int __user
*child_tidptr
,
961 struct task_struct
*p
= NULL
;
963 if ((clone_flags
& (CLONE_NEWNS
|CLONE_FS
)) == (CLONE_NEWNS
|CLONE_FS
))
964 return ERR_PTR(-EINVAL
);
967 * Thread groups must share signals as well, and detached threads
968 * can only be started up within the thread group.
970 if ((clone_flags
& CLONE_THREAD
) && !(clone_flags
& CLONE_SIGHAND
))
971 return ERR_PTR(-EINVAL
);
974 * Shared signal handlers imply shared VM. By way of the above,
975 * thread groups also imply shared VM. Blocking this case allows
976 * for various simplifications in other code.
978 if ((clone_flags
& CLONE_SIGHAND
) && !(clone_flags
& CLONE_VM
))
979 return ERR_PTR(-EINVAL
);
981 retval
= security_task_create(clone_flags
);
986 p
= dup_task_struct(current
);
990 rt_mutex_init_task(p
);
992 #ifdef CONFIG_TRACE_IRQFLAGS
993 DEBUG_LOCKS_WARN_ON(!p
->hardirqs_enabled
);
994 DEBUG_LOCKS_WARN_ON(!p
->softirqs_enabled
);
997 if (atomic_read(&p
->user
->processes
) >=
998 p
->signal
->rlim
[RLIMIT_NPROC
].rlim_cur
) {
999 if (!capable(CAP_SYS_ADMIN
) && !capable(CAP_SYS_RESOURCE
) &&
1000 p
->user
!= &root_user
)
1004 atomic_inc(&p
->user
->__count
);
1005 atomic_inc(&p
->user
->processes
);
1006 get_group_info(p
->group_info
);
1009 * If multiple threads are within copy_process(), then this check
1010 * triggers too late. This doesn't hurt, the check is only there
1011 * to stop root fork bombs.
1013 if (nr_threads
>= max_threads
)
1014 goto bad_fork_cleanup_count
;
1016 if (!try_module_get(task_thread_info(p
)->exec_domain
->module
))
1017 goto bad_fork_cleanup_count
;
1019 if (p
->binfmt
&& !try_module_get(p
->binfmt
->module
))
1020 goto bad_fork_cleanup_put_domain
;
1023 delayacct_tsk_init(p
); /* Must remain after dup_task_struct() */
1024 copy_flags(clone_flags
, p
);
1027 if (clone_flags
& CLONE_PARENT_SETTID
)
1028 if (put_user(p
->pid
, parent_tidptr
))
1029 goto bad_fork_cleanup_delays_binfmt
;
1031 INIT_LIST_HEAD(&p
->children
);
1032 INIT_LIST_HEAD(&p
->sibling
);
1033 p
->vfork_done
= NULL
;
1034 spin_lock_init(&p
->alloc_lock
);
1036 clear_tsk_thread_flag(p
, TIF_SIGPENDING
);
1037 init_sigpending(&p
->pending
);
1039 p
->utime
= cputime_zero
;
1040 p
->stime
= cputime_zero
;
1042 p
->rchar
= 0; /* I/O counter: bytes read */
1043 p
->wchar
= 0; /* I/O counter: bytes written */
1044 p
->syscr
= 0; /* I/O counter: read syscalls */
1045 p
->syscw
= 0; /* I/O counter: write syscalls */
1046 task_io_accounting_init(p
);
1047 acct_clear_integrals(p
);
1049 p
->it_virt_expires
= cputime_zero
;
1050 p
->it_prof_expires
= cputime_zero
;
1051 p
->it_sched_expires
= 0;
1052 INIT_LIST_HEAD(&p
->cpu_timers
[0]);
1053 INIT_LIST_HEAD(&p
->cpu_timers
[1]);
1054 INIT_LIST_HEAD(&p
->cpu_timers
[2]);
1056 p
->lock_depth
= -1; /* -1 = no lock */
1057 do_posix_clock_monotonic_gettime(&p
->start_time
);
1059 p
->io_context
= NULL
;
1061 p
->audit_context
= NULL
;
1064 p
->mempolicy
= mpol_copy(p
->mempolicy
);
1065 if (IS_ERR(p
->mempolicy
)) {
1066 retval
= PTR_ERR(p
->mempolicy
);
1067 p
->mempolicy
= NULL
;
1068 goto bad_fork_cleanup_cpuset
;
1070 mpol_fix_fork_child_flag(p
);
1072 #ifdef CONFIG_TRACE_IRQFLAGS
1074 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1075 p
->hardirqs_enabled
= 1;
1077 p
->hardirqs_enabled
= 0;
1079 p
->hardirq_enable_ip
= 0;
1080 p
->hardirq_enable_event
= 0;
1081 p
->hardirq_disable_ip
= _THIS_IP_
;
1082 p
->hardirq_disable_event
= 0;
1083 p
->softirqs_enabled
= 1;
1084 p
->softirq_enable_ip
= _THIS_IP_
;
1085 p
->softirq_enable_event
= 0;
1086 p
->softirq_disable_ip
= 0;
1087 p
->softirq_disable_event
= 0;
1088 p
->hardirq_context
= 0;
1089 p
->softirq_context
= 0;
1091 #ifdef CONFIG_LOCKDEP
1092 p
->lockdep_depth
= 0; /* no locks held yet */
1093 p
->curr_chain_key
= 0;
1094 p
->lockdep_recursion
= 0;
1097 #ifdef CONFIG_DEBUG_MUTEXES
1098 p
->blocked_on
= NULL
; /* not blocked yet */
1102 if (clone_flags
& CLONE_THREAD
)
1103 p
->tgid
= current
->tgid
;
1105 if ((retval
= security_task_alloc(p
)))
1106 goto bad_fork_cleanup_policy
;
1107 if ((retval
= audit_alloc(p
)))
1108 goto bad_fork_cleanup_security
;
1109 /* copy all the process information */
1110 if ((retval
= copy_semundo(clone_flags
, p
)))
1111 goto bad_fork_cleanup_audit
;
1112 if ((retval
= copy_files(clone_flags
, p
)))
1113 goto bad_fork_cleanup_semundo
;
1114 if ((retval
= copy_fs(clone_flags
, p
)))
1115 goto bad_fork_cleanup_files
;
1116 if ((retval
= copy_sighand(clone_flags
, p
)))
1117 goto bad_fork_cleanup_fs
;
1118 if ((retval
= copy_signal(clone_flags
, p
)))
1119 goto bad_fork_cleanup_sighand
;
1120 if ((retval
= copy_mm(clone_flags
, p
)))
1121 goto bad_fork_cleanup_signal
;
1122 if ((retval
= copy_keys(clone_flags
, p
)))
1123 goto bad_fork_cleanup_mm
;
1124 if ((retval
= copy_namespaces(clone_flags
, p
)))
1125 goto bad_fork_cleanup_keys
;
1126 retval
= copy_thread(0, clone_flags
, stack_start
, stack_size
, p
, regs
);
1128 goto bad_fork_cleanup_namespaces
;
1130 p
->set_child_tid
= (clone_flags
& CLONE_CHILD_SETTID
) ? child_tidptr
: NULL
;
1132 * Clear TID on mm_release()?
1134 p
->clear_child_tid
= (clone_flags
& CLONE_CHILD_CLEARTID
) ? child_tidptr
: NULL
;
1135 p
->robust_list
= NULL
;
1136 #ifdef CONFIG_COMPAT
1137 p
->compat_robust_list
= NULL
;
1139 INIT_LIST_HEAD(&p
->pi_state_list
);
1140 p
->pi_state_cache
= NULL
;
1143 * sigaltstack should be cleared when sharing the same VM
1145 if ((clone_flags
& (CLONE_VM
|CLONE_VFORK
)) == CLONE_VM
)
1146 p
->sas_ss_sp
= p
->sas_ss_size
= 0;
1149 * Syscall tracing should be turned off in the child regardless
1152 clear_tsk_thread_flag(p
, TIF_SYSCALL_TRACE
);
1153 #ifdef TIF_SYSCALL_EMU
1154 clear_tsk_thread_flag(p
, TIF_SYSCALL_EMU
);
1157 /* Our parent execution domain becomes current domain
1158 These must match for thread signalling to apply */
1159 p
->parent_exec_id
= p
->self_exec_id
;
1161 /* ok, now we should be set up.. */
1162 p
->exit_signal
= (clone_flags
& CLONE_THREAD
) ? -1 : (clone_flags
& CSIGNAL
);
1163 p
->pdeath_signal
= 0;
1167 * Ok, make it visible to the rest of the system.
1168 * We dont wake it up yet.
1170 p
->group_leader
= p
;
1171 INIT_LIST_HEAD(&p
->thread_group
);
1172 INIT_LIST_HEAD(&p
->ptrace_children
);
1173 INIT_LIST_HEAD(&p
->ptrace_list
);
1175 /* Perform scheduler related setup. Assign this task to a CPU. */
1176 sched_fork(p
, clone_flags
);
1178 /* Need tasklist lock for parent etc handling! */
1179 write_lock_irq(&tasklist_lock
);
1181 /* for sys_ioprio_set(IOPRIO_WHO_PGRP) */
1182 p
->ioprio
= current
->ioprio
;
1185 * The task hasn't been attached yet, so its cpus_allowed mask will
1186 * not be changed, nor will its assigned CPU.
1188 * The cpus_allowed mask of the parent may have changed after it was
1189 * copied first time - so re-copy it here, then check the child's CPU
1190 * to ensure it is on a valid CPU (and if not, just force it back to
1191 * parent's CPU). This avoids alot of nasty races.
1193 p
->cpus_allowed
= current
->cpus_allowed
;
1194 if (unlikely(!cpu_isset(task_cpu(p
), p
->cpus_allowed
) ||
1195 !cpu_online(task_cpu(p
))))
1196 set_task_cpu(p
, smp_processor_id());
1198 /* CLONE_PARENT re-uses the old parent */
1199 if (clone_flags
& (CLONE_PARENT
|CLONE_THREAD
))
1200 p
->real_parent
= current
->real_parent
;
1202 p
->real_parent
= current
;
1203 p
->parent
= p
->real_parent
;
1205 spin_lock(¤t
->sighand
->siglock
);
1208 * Process group and session signals need to be delivered to just the
1209 * parent before the fork or both the parent and the child after the
1210 * fork. Restart if a signal comes in before we add the new process to
1211 * it's process group.
1212 * A fatal signal pending means that current will exit, so the new
1213 * thread can't slip out of an OOM kill (or normal SIGKILL).
1215 recalc_sigpending();
1216 if (signal_pending(current
)) {
1217 spin_unlock(¤t
->sighand
->siglock
);
1218 write_unlock_irq(&tasklist_lock
);
1219 retval
= -ERESTARTNOINTR
;
1220 goto bad_fork_cleanup_namespaces
;
1223 if (clone_flags
& CLONE_THREAD
) {
1224 p
->group_leader
= current
->group_leader
;
1225 list_add_tail_rcu(&p
->thread_group
, &p
->group_leader
->thread_group
);
1227 if (!cputime_eq(current
->signal
->it_virt_expires
,
1229 !cputime_eq(current
->signal
->it_prof_expires
,
1231 current
->signal
->rlim
[RLIMIT_CPU
].rlim_cur
!= RLIM_INFINITY
||
1232 !list_empty(¤t
->signal
->cpu_timers
[0]) ||
1233 !list_empty(¤t
->signal
->cpu_timers
[1]) ||
1234 !list_empty(¤t
->signal
->cpu_timers
[2])) {
1236 * Have child wake up on its first tick to check
1237 * for process CPU timers.
1239 p
->it_prof_expires
= jiffies_to_cputime(1);
1243 if (likely(p
->pid
)) {
1245 if (unlikely(p
->ptrace
& PT_PTRACED
))
1246 __ptrace_link(p
, current
->parent
);
1248 if (thread_group_leader(p
)) {
1249 p
->signal
->tty
= current
->signal
->tty
;
1250 p
->signal
->pgrp
= process_group(current
);
1251 set_signal_session(p
->signal
, process_session(current
));
1252 attach_pid(p
, PIDTYPE_PGID
, process_group(p
));
1253 attach_pid(p
, PIDTYPE_SID
, process_session(p
));
1255 list_add_tail_rcu(&p
->tasks
, &init_task
.tasks
);
1256 __get_cpu_var(process_counts
)++;
1258 attach_pid(p
, PIDTYPE_PID
, p
->pid
);
1263 spin_unlock(¤t
->sighand
->siglock
);
1264 write_unlock_irq(&tasklist_lock
);
1265 proc_fork_connector(p
);
1268 bad_fork_cleanup_namespaces
:
1269 exit_task_namespaces(p
);
1270 bad_fork_cleanup_keys
:
1272 bad_fork_cleanup_mm
:
1275 bad_fork_cleanup_signal
:
1277 bad_fork_cleanup_sighand
:
1278 __cleanup_sighand(p
->sighand
);
1279 bad_fork_cleanup_fs
:
1280 exit_fs(p
); /* blocking */
1281 bad_fork_cleanup_files
:
1282 exit_files(p
); /* blocking */
1283 bad_fork_cleanup_semundo
:
1285 bad_fork_cleanup_audit
:
1287 bad_fork_cleanup_security
:
1288 security_task_free(p
);
1289 bad_fork_cleanup_policy
:
1291 mpol_free(p
->mempolicy
);
1292 bad_fork_cleanup_cpuset
:
1295 bad_fork_cleanup_delays_binfmt
:
1296 delayacct_tsk_free(p
);
1298 module_put(p
->binfmt
->module
);
1299 bad_fork_cleanup_put_domain
:
1300 module_put(task_thread_info(p
)->exec_domain
->module
);
1301 bad_fork_cleanup_count
:
1302 put_group_info(p
->group_info
);
1303 atomic_dec(&p
->user
->processes
);
1308 return ERR_PTR(retval
);
1311 noinline
struct pt_regs
* __devinit
__attribute__((weak
)) idle_regs(struct pt_regs
*regs
)
1313 memset(regs
, 0, sizeof(struct pt_regs
));
1317 struct task_struct
* __devinit
fork_idle(int cpu
)
1319 struct task_struct
*task
;
1320 struct pt_regs regs
;
1322 task
= copy_process(CLONE_VM
, 0, idle_regs(®s
), 0, NULL
, NULL
, 0);
1324 init_idle(task
, cpu
);
1329 static inline int fork_traceflag (unsigned clone_flags
)
1331 if (clone_flags
& CLONE_UNTRACED
)
1333 else if (clone_flags
& CLONE_VFORK
) {
1334 if (current
->ptrace
& PT_TRACE_VFORK
)
1335 return PTRACE_EVENT_VFORK
;
1336 } else if ((clone_flags
& CSIGNAL
) != SIGCHLD
) {
1337 if (current
->ptrace
& PT_TRACE_CLONE
)
1338 return PTRACE_EVENT_CLONE
;
1339 } else if (current
->ptrace
& PT_TRACE_FORK
)
1340 return PTRACE_EVENT_FORK
;
1346 * Ok, this is the main fork-routine.
1348 * It copies the process, and if successful kick-starts
1349 * it and waits for it to finish using the VM if required.
1351 long do_fork(unsigned long clone_flags
,
1352 unsigned long stack_start
,
1353 struct pt_regs
*regs
,
1354 unsigned long stack_size
,
1355 int __user
*parent_tidptr
,
1356 int __user
*child_tidptr
)
1358 struct task_struct
*p
;
1360 struct pid
*pid
= alloc_pid();
1366 if (unlikely(current
->ptrace
)) {
1367 trace
= fork_traceflag (clone_flags
);
1369 clone_flags
|= CLONE_PTRACE
;
1372 p
= copy_process(clone_flags
, stack_start
, regs
, stack_size
, parent_tidptr
, child_tidptr
, nr
);
1374 * Do this prior waking up the new thread - the thread pointer
1375 * might get invalid after that point, if the thread exits quickly.
1378 struct completion vfork
;
1380 if (clone_flags
& CLONE_VFORK
) {
1381 p
->vfork_done
= &vfork
;
1382 init_completion(&vfork
);
1385 if ((p
->ptrace
& PT_PTRACED
) || (clone_flags
& CLONE_STOPPED
)) {
1387 * We'll start up with an immediate SIGSTOP.
1389 sigaddset(&p
->pending
.signal
, SIGSTOP
);
1390 set_tsk_thread_flag(p
, TIF_SIGPENDING
);
1393 if (!(clone_flags
& CLONE_STOPPED
))
1394 wake_up_new_task(p
, clone_flags
);
1396 p
->state
= TASK_STOPPED
;
1398 if (unlikely (trace
)) {
1399 current
->ptrace_message
= nr
;
1400 ptrace_notify ((trace
<< 8) | SIGTRAP
);
1403 if (clone_flags
& CLONE_VFORK
) {
1404 wait_for_completion(&vfork
);
1405 if (unlikely (current
->ptrace
& PT_TRACE_VFORK_DONE
)) {
1406 current
->ptrace_message
= nr
;
1407 ptrace_notify ((PTRACE_EVENT_VFORK_DONE
<< 8) | SIGTRAP
);
1417 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1418 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1421 static void sighand_ctor(void *data
, struct kmem_cache
*cachep
, unsigned long flags
)
1423 struct sighand_struct
*sighand
= data
;
1425 if ((flags
& (SLAB_CTOR_VERIFY
| SLAB_CTOR_CONSTRUCTOR
)) ==
1426 SLAB_CTOR_CONSTRUCTOR
)
1427 spin_lock_init(&sighand
->siglock
);
1430 void __init
proc_caches_init(void)
1432 sighand_cachep
= kmem_cache_create("sighand_cache",
1433 sizeof(struct sighand_struct
), 0,
1434 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_DESTROY_BY_RCU
,
1435 sighand_ctor
, NULL
);
1436 signal_cachep
= kmem_cache_create("signal_cache",
1437 sizeof(struct signal_struct
), 0,
1438 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
, NULL
);
1439 files_cachep
= kmem_cache_create("files_cache",
1440 sizeof(struct files_struct
), 0,
1441 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
, NULL
);
1442 fs_cachep
= kmem_cache_create("fs_cache",
1443 sizeof(struct fs_struct
), 0,
1444 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
, NULL
);
1445 vm_area_cachep
= kmem_cache_create("vm_area_struct",
1446 sizeof(struct vm_area_struct
), 0,
1447 SLAB_PANIC
, NULL
, NULL
);
1448 mm_cachep
= kmem_cache_create("mm_struct",
1449 sizeof(struct mm_struct
), ARCH_MIN_MMSTRUCT_ALIGN
,
1450 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
, NULL
);
1455 * Check constraints on flags passed to the unshare system call and
1456 * force unsharing of additional process context as appropriate.
1458 static inline void check_unshare_flags(unsigned long *flags_ptr
)
1461 * If unsharing a thread from a thread group, must also
1464 if (*flags_ptr
& CLONE_THREAD
)
1465 *flags_ptr
|= CLONE_VM
;
1468 * If unsharing vm, must also unshare signal handlers.
1470 if (*flags_ptr
& CLONE_VM
)
1471 *flags_ptr
|= CLONE_SIGHAND
;
1474 * If unsharing signal handlers and the task was created
1475 * using CLONE_THREAD, then must unshare the thread
1477 if ((*flags_ptr
& CLONE_SIGHAND
) &&
1478 (atomic_read(¤t
->signal
->count
) > 1))
1479 *flags_ptr
|= CLONE_THREAD
;
1482 * If unsharing namespace, must also unshare filesystem information.
1484 if (*flags_ptr
& CLONE_NEWNS
)
1485 *flags_ptr
|= CLONE_FS
;
1489 * Unsharing of tasks created with CLONE_THREAD is not supported yet
1491 static int unshare_thread(unsigned long unshare_flags
)
1493 if (unshare_flags
& CLONE_THREAD
)
1500 * Unshare the filesystem structure if it is being shared
1502 static int unshare_fs(unsigned long unshare_flags
, struct fs_struct
**new_fsp
)
1504 struct fs_struct
*fs
= current
->fs
;
1506 if ((unshare_flags
& CLONE_FS
) &&
1507 (fs
&& atomic_read(&fs
->count
) > 1)) {
1508 *new_fsp
= __copy_fs_struct(current
->fs
);
1517 * Unshare the mnt_namespace structure if it is being shared
1519 static int unshare_mnt_namespace(unsigned long unshare_flags
,
1520 struct mnt_namespace
**new_nsp
, struct fs_struct
*new_fs
)
1522 struct mnt_namespace
*ns
= current
->nsproxy
->mnt_ns
;
1524 if ((unshare_flags
& CLONE_NEWNS
) && ns
) {
1525 if (!capable(CAP_SYS_ADMIN
))
1528 *new_nsp
= dup_mnt_ns(current
, new_fs
? new_fs
: current
->fs
);
1537 * Unsharing of sighand is not supported yet
1539 static int unshare_sighand(unsigned long unshare_flags
, struct sighand_struct
**new_sighp
)
1541 struct sighand_struct
*sigh
= current
->sighand
;
1543 if ((unshare_flags
& CLONE_SIGHAND
) && atomic_read(&sigh
->count
) > 1)
1550 * Unshare vm if it is being shared
1552 static int unshare_vm(unsigned long unshare_flags
, struct mm_struct
**new_mmp
)
1554 struct mm_struct
*mm
= current
->mm
;
1556 if ((unshare_flags
& CLONE_VM
) &&
1557 (mm
&& atomic_read(&mm
->mm_users
) > 1)) {
1565 * Unshare file descriptor table if it is being shared
1567 static int unshare_fd(unsigned long unshare_flags
, struct files_struct
**new_fdp
)
1569 struct files_struct
*fd
= current
->files
;
1572 if ((unshare_flags
& CLONE_FILES
) &&
1573 (fd
&& atomic_read(&fd
->count
) > 1)) {
1574 *new_fdp
= dup_fd(fd
, &error
);
1583 * Unsharing of semundo for tasks created with CLONE_SYSVSEM is not
1586 static int unshare_semundo(unsigned long unshare_flags
, struct sem_undo_list
**new_ulistp
)
1588 if (unshare_flags
& CLONE_SYSVSEM
)
1594 #ifndef CONFIG_IPC_NS
1595 static inline int unshare_ipcs(unsigned long flags
, struct ipc_namespace
**ns
)
1597 if (flags
& CLONE_NEWIPC
)
1605 * unshare allows a process to 'unshare' part of the process
1606 * context which was originally shared using clone. copy_*
1607 * functions used by do_fork() cannot be used here directly
1608 * because they modify an inactive task_struct that is being
1609 * constructed. Here we are modifying the current, active,
1612 asmlinkage
long sys_unshare(unsigned long unshare_flags
)
1615 struct fs_struct
*fs
, *new_fs
= NULL
;
1616 struct mnt_namespace
*ns
, *new_ns
= NULL
;
1617 struct sighand_struct
*new_sigh
= NULL
;
1618 struct mm_struct
*mm
, *new_mm
= NULL
, *active_mm
= NULL
;
1619 struct files_struct
*fd
, *new_fd
= NULL
;
1620 struct sem_undo_list
*new_ulist
= NULL
;
1621 struct nsproxy
*new_nsproxy
= NULL
, *old_nsproxy
= NULL
;
1622 struct uts_namespace
*uts
, *new_uts
= NULL
;
1623 struct ipc_namespace
*ipc
, *new_ipc
= NULL
;
1625 check_unshare_flags(&unshare_flags
);
1627 /* Return -EINVAL for all unsupported flags */
1629 if (unshare_flags
& ~(CLONE_THREAD
|CLONE_FS
|CLONE_NEWNS
|CLONE_SIGHAND
|
1630 CLONE_VM
|CLONE_FILES
|CLONE_SYSVSEM
|
1631 CLONE_NEWUTS
|CLONE_NEWIPC
))
1632 goto bad_unshare_out
;
1634 if ((err
= unshare_thread(unshare_flags
)))
1635 goto bad_unshare_out
;
1636 if ((err
= unshare_fs(unshare_flags
, &new_fs
)))
1637 goto bad_unshare_cleanup_thread
;
1638 if ((err
= unshare_mnt_namespace(unshare_flags
, &new_ns
, new_fs
)))
1639 goto bad_unshare_cleanup_fs
;
1640 if ((err
= unshare_sighand(unshare_flags
, &new_sigh
)))
1641 goto bad_unshare_cleanup_ns
;
1642 if ((err
= unshare_vm(unshare_flags
, &new_mm
)))
1643 goto bad_unshare_cleanup_sigh
;
1644 if ((err
= unshare_fd(unshare_flags
, &new_fd
)))
1645 goto bad_unshare_cleanup_vm
;
1646 if ((err
= unshare_semundo(unshare_flags
, &new_ulist
)))
1647 goto bad_unshare_cleanup_fd
;
1648 if ((err
= unshare_utsname(unshare_flags
, &new_uts
)))
1649 goto bad_unshare_cleanup_semundo
;
1650 if ((err
= unshare_ipcs(unshare_flags
, &new_ipc
)))
1651 goto bad_unshare_cleanup_uts
;
1653 if (new_ns
|| new_uts
|| new_ipc
) {
1654 old_nsproxy
= current
->nsproxy
;
1655 new_nsproxy
= dup_namespaces(old_nsproxy
);
1658 goto bad_unshare_cleanup_ipc
;
1662 if (new_fs
|| new_ns
|| new_mm
|| new_fd
|| new_ulist
||
1663 new_uts
|| new_ipc
) {
1668 current
->nsproxy
= new_nsproxy
;
1669 new_nsproxy
= old_nsproxy
;
1674 current
->fs
= new_fs
;
1679 ns
= current
->nsproxy
->mnt_ns
;
1680 current
->nsproxy
->mnt_ns
= new_ns
;
1686 active_mm
= current
->active_mm
;
1687 current
->mm
= new_mm
;
1688 current
->active_mm
= new_mm
;
1689 activate_mm(active_mm
, new_mm
);
1694 fd
= current
->files
;
1695 current
->files
= new_fd
;
1700 uts
= current
->nsproxy
->uts_ns
;
1701 current
->nsproxy
->uts_ns
= new_uts
;
1706 ipc
= current
->nsproxy
->ipc_ns
;
1707 current
->nsproxy
->ipc_ns
= new_ipc
;
1711 task_unlock(current
);
1715 put_nsproxy(new_nsproxy
);
1717 bad_unshare_cleanup_ipc
:
1719 put_ipc_ns(new_ipc
);
1721 bad_unshare_cleanup_uts
:
1723 put_uts_ns(new_uts
);
1725 bad_unshare_cleanup_semundo
:
1726 bad_unshare_cleanup_fd
:
1728 put_files_struct(new_fd
);
1730 bad_unshare_cleanup_vm
:
1734 bad_unshare_cleanup_sigh
:
1736 if (atomic_dec_and_test(&new_sigh
->count
))
1737 kmem_cache_free(sighand_cachep
, new_sigh
);
1739 bad_unshare_cleanup_ns
:
1743 bad_unshare_cleanup_fs
:
1745 put_fs_struct(new_fs
);
1747 bad_unshare_cleanup_thread
: