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/mnt_namespace.h>
21 #include <linux/personality.h>
22 #include <linux/mempolicy.h>
23 #include <linux/sem.h>
24 #include <linux/file.h>
25 #include <linux/key.h>
26 #include <linux/binfmts.h>
27 #include <linux/mman.h>
29 #include <linux/nsproxy.h>
30 #include <linux/capability.h>
31 #include <linux/cpu.h>
32 #include <linux/cgroup.h>
33 #include <linux/security.h>
34 #include <linux/swap.h>
35 #include <linux/syscalls.h>
36 #include <linux/jiffies.h>
37 #include <linux/futex.h>
38 #include <linux/task_io_accounting_ops.h>
39 #include <linux/rcupdate.h>
40 #include <linux/ptrace.h>
41 #include <linux/mount.h>
42 #include <linux/audit.h>
43 #include <linux/memcontrol.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/freezer.h>
50 #include <linux/delayacct.h>
51 #include <linux/taskstats_kern.h>
52 #include <linux/random.h>
53 #include <linux/tty.h>
54 #include <linux/proc_fs.h>
55 #include <linux/blkdev.h>
57 #include <asm/pgtable.h>
58 #include <asm/pgalloc.h>
59 #include <asm/uaccess.h>
60 #include <asm/mmu_context.h>
61 #include <asm/cacheflush.h>
62 #include <asm/tlbflush.h>
65 * Protected counters by write_lock_irq(&tasklist_lock)
67 unsigned long total_forks
; /* Handle normal Linux uptimes. */
68 int nr_threads
; /* The idle threads do not count.. */
70 int max_threads
; /* tunable limit on nr_threads */
72 DEFINE_PER_CPU(unsigned long, process_counts
) = 0;
74 __cacheline_aligned
DEFINE_RWLOCK(tasklist_lock
); /* outer */
76 int nr_processes(void)
81 for_each_online_cpu(cpu
)
82 total
+= per_cpu(process_counts
, cpu
);
87 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
88 # define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL)
89 # define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk))
90 static struct kmem_cache
*task_struct_cachep
;
93 /* SLAB cache for signal_struct structures (tsk->signal) */
94 static struct kmem_cache
*signal_cachep
;
96 /* SLAB cache for sighand_struct structures (tsk->sighand) */
97 struct kmem_cache
*sighand_cachep
;
99 /* SLAB cache for files_struct structures (tsk->files) */
100 struct kmem_cache
*files_cachep
;
102 /* SLAB cache for fs_struct structures (tsk->fs) */
103 struct kmem_cache
*fs_cachep
;
105 /* SLAB cache for vm_area_struct structures */
106 struct kmem_cache
*vm_area_cachep
;
108 /* SLAB cache for mm_struct structures (tsk->mm) */
109 static struct kmem_cache
*mm_cachep
;
111 void free_task(struct task_struct
*tsk
)
113 prop_local_destroy_single(&tsk
->dirties
);
114 free_thread_info(tsk
->stack
);
115 rt_mutex_debug_task_free(tsk
);
116 free_task_struct(tsk
);
118 EXPORT_SYMBOL(free_task
);
120 void __put_task_struct(struct task_struct
*tsk
)
122 WARN_ON(!tsk
->exit_state
);
123 WARN_ON(atomic_read(&tsk
->usage
));
124 WARN_ON(tsk
== current
);
126 security_task_free(tsk
);
128 put_group_info(tsk
->group_info
);
129 delayacct_tsk_free(tsk
);
131 if (!profile_handoff_task(tsk
))
136 * macro override instead of weak attribute alias, to workaround
137 * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
139 #ifndef arch_task_cache_init
140 #define arch_task_cache_init()
143 void __init
fork_init(unsigned long mempages
)
145 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
146 #ifndef ARCH_MIN_TASKALIGN
147 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
149 /* create a slab on which task_structs can be allocated */
151 kmem_cache_create("task_struct", sizeof(struct task_struct
),
152 ARCH_MIN_TASKALIGN
, SLAB_PANIC
, NULL
);
155 /* do the arch specific task caches init */
156 arch_task_cache_init();
159 * The default maximum number of threads is set to a safe
160 * value: the thread structures can take up at most half
163 max_threads
= mempages
/ (8 * THREAD_SIZE
/ PAGE_SIZE
);
166 * we need to allow at least 20 threads to boot a system
171 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_cur
= max_threads
/2;
172 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_max
= max_threads
/2;
173 init_task
.signal
->rlim
[RLIMIT_SIGPENDING
] =
174 init_task
.signal
->rlim
[RLIMIT_NPROC
];
177 int __attribute__((weak
)) arch_dup_task_struct(struct task_struct
*dst
,
178 struct task_struct
*src
)
184 static struct task_struct
*dup_task_struct(struct task_struct
*orig
)
186 struct task_struct
*tsk
;
187 struct thread_info
*ti
;
190 prepare_to_copy(orig
);
192 tsk
= alloc_task_struct();
196 ti
= alloc_thread_info(tsk
);
198 free_task_struct(tsk
);
202 err
= arch_dup_task_struct(tsk
, orig
);
208 err
= prop_local_init_single(&tsk
->dirties
);
212 setup_thread_stack(tsk
, orig
);
214 #ifdef CONFIG_CC_STACKPROTECTOR
215 tsk
->stack_canary
= get_random_int();
218 /* One for us, one for whoever does the "release_task()" (usually parent) */
219 atomic_set(&tsk
->usage
,2);
220 atomic_set(&tsk
->fs_excl
, 0);
221 #ifdef CONFIG_BLK_DEV_IO_TRACE
224 tsk
->splice_pipe
= NULL
;
228 free_thread_info(ti
);
229 free_task_struct(tsk
);
234 static int dup_mmap(struct mm_struct
*mm
, struct mm_struct
*oldmm
)
236 struct vm_area_struct
*mpnt
, *tmp
, **pprev
;
237 struct rb_node
**rb_link
, *rb_parent
;
239 unsigned long charge
;
240 struct mempolicy
*pol
;
242 down_write(&oldmm
->mmap_sem
);
243 flush_cache_dup_mm(oldmm
);
245 * Not linked in yet - no deadlock potential:
247 down_write_nested(&mm
->mmap_sem
, SINGLE_DEPTH_NESTING
);
251 mm
->mmap_cache
= NULL
;
252 mm
->free_area_cache
= oldmm
->mmap_base
;
253 mm
->cached_hole_size
= ~0UL;
255 cpus_clear(mm
->cpu_vm_mask
);
257 rb_link
= &mm
->mm_rb
.rb_node
;
261 for (mpnt
= oldmm
->mmap
; mpnt
; mpnt
= mpnt
->vm_next
) {
264 if (mpnt
->vm_flags
& VM_DONTCOPY
) {
265 long pages
= vma_pages(mpnt
);
266 mm
->total_vm
-= pages
;
267 vm_stat_account(mm
, mpnt
->vm_flags
, mpnt
->vm_file
,
272 if (mpnt
->vm_flags
& VM_ACCOUNT
) {
273 unsigned int len
= (mpnt
->vm_end
- mpnt
->vm_start
) >> PAGE_SHIFT
;
274 if (security_vm_enough_memory(len
))
278 tmp
= kmem_cache_alloc(vm_area_cachep
, GFP_KERNEL
);
282 pol
= mpol_copy(vma_policy(mpnt
));
283 retval
= PTR_ERR(pol
);
285 goto fail_nomem_policy
;
286 vma_set_policy(tmp
, pol
);
287 tmp
->vm_flags
&= ~VM_LOCKED
;
293 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
295 if (tmp
->vm_flags
& VM_DENYWRITE
)
296 atomic_dec(&inode
->i_writecount
);
298 /* insert tmp into the share list, just after mpnt */
299 spin_lock(&file
->f_mapping
->i_mmap_lock
);
300 tmp
->vm_truncate_count
= mpnt
->vm_truncate_count
;
301 flush_dcache_mmap_lock(file
->f_mapping
);
302 vma_prio_tree_add(tmp
, mpnt
);
303 flush_dcache_mmap_unlock(file
->f_mapping
);
304 spin_unlock(&file
->f_mapping
->i_mmap_lock
);
308 * Link in the new vma and copy the page table entries.
311 pprev
= &tmp
->vm_next
;
313 __vma_link_rb(mm
, tmp
, rb_link
, rb_parent
);
314 rb_link
= &tmp
->vm_rb
.rb_right
;
315 rb_parent
= &tmp
->vm_rb
;
318 retval
= copy_page_range(mm
, oldmm
, mpnt
);
320 if (tmp
->vm_ops
&& tmp
->vm_ops
->open
)
321 tmp
->vm_ops
->open(tmp
);
326 /* a new mm has just been created */
327 arch_dup_mmap(oldmm
, mm
);
330 up_write(&mm
->mmap_sem
);
332 up_write(&oldmm
->mmap_sem
);
335 kmem_cache_free(vm_area_cachep
, tmp
);
338 vm_unacct_memory(charge
);
342 static inline int mm_alloc_pgd(struct mm_struct
* mm
)
344 mm
->pgd
= pgd_alloc(mm
);
345 if (unlikely(!mm
->pgd
))
350 static inline void mm_free_pgd(struct mm_struct
* mm
)
352 pgd_free(mm
, mm
->pgd
);
355 #define dup_mmap(mm, oldmm) (0)
356 #define mm_alloc_pgd(mm) (0)
357 #define mm_free_pgd(mm)
358 #endif /* CONFIG_MMU */
360 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(mmlist_lock
);
362 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
363 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
365 #include <linux/init_task.h>
367 static struct mm_struct
* mm_init(struct mm_struct
* mm
, struct task_struct
*p
)
369 atomic_set(&mm
->mm_users
, 1);
370 atomic_set(&mm
->mm_count
, 1);
371 init_rwsem(&mm
->mmap_sem
);
372 INIT_LIST_HEAD(&mm
->mmlist
);
373 mm
->flags
= (current
->mm
) ? current
->mm
->flags
374 : MMF_DUMP_FILTER_DEFAULT
;
375 mm
->core_waiters
= 0;
377 set_mm_counter(mm
, file_rss
, 0);
378 set_mm_counter(mm
, anon_rss
, 0);
379 spin_lock_init(&mm
->page_table_lock
);
380 rwlock_init(&mm
->ioctx_list_lock
);
381 mm
->ioctx_list
= NULL
;
382 mm
->free_area_cache
= TASK_UNMAPPED_BASE
;
383 mm
->cached_hole_size
= ~0UL;
384 mm_init_cgroup(mm
, p
);
386 if (likely(!mm_alloc_pgd(mm
))) {
397 * Allocate and initialize an mm_struct.
399 struct mm_struct
* mm_alloc(void)
401 struct mm_struct
* mm
;
405 memset(mm
, 0, sizeof(*mm
));
406 mm
= mm_init(mm
, current
);
412 * Called when the last reference to the mm
413 * is dropped: either by a lazy thread or by
414 * mmput. Free the page directory and the mm.
416 void __mmdrop(struct mm_struct
*mm
)
418 BUG_ON(mm
== &init_mm
);
423 EXPORT_SYMBOL_GPL(__mmdrop
);
426 * Decrement the use count and release all resources for an mm.
428 void mmput(struct mm_struct
*mm
)
432 if (atomic_dec_and_test(&mm
->mm_users
)) {
435 if (!list_empty(&mm
->mmlist
)) {
436 spin_lock(&mmlist_lock
);
437 list_del(&mm
->mmlist
);
438 spin_unlock(&mmlist_lock
);
445 EXPORT_SYMBOL_GPL(mmput
);
448 * get_task_mm - acquire a reference to the task's mm
450 * Returns %NULL if the task has no mm. Checks PF_BORROWED_MM (meaning
451 * this kernel workthread has transiently adopted a user mm with use_mm,
452 * to do its AIO) is not set and if so returns a reference to it, after
453 * bumping up the use count. User must release the mm via mmput()
454 * after use. Typically used by /proc and ptrace.
456 struct mm_struct
*get_task_mm(struct task_struct
*task
)
458 struct mm_struct
*mm
;
463 if (task
->flags
& PF_BORROWED_MM
)
466 atomic_inc(&mm
->mm_users
);
471 EXPORT_SYMBOL_GPL(get_task_mm
);
473 /* Please note the differences between mmput and mm_release.
474 * mmput is called whenever we stop holding onto a mm_struct,
475 * error success whatever.
477 * mm_release is called after a mm_struct has been removed
478 * from the current process.
480 * This difference is important for error handling, when we
481 * only half set up a mm_struct for a new process and need to restore
482 * the old one. Because we mmput the new mm_struct before
483 * restoring the old one. . .
484 * Eric Biederman 10 January 1998
486 void mm_release(struct task_struct
*tsk
, struct mm_struct
*mm
)
488 struct completion
*vfork_done
= tsk
->vfork_done
;
490 /* Get rid of any cached register state */
491 deactivate_mm(tsk
, mm
);
493 /* notify parent sleeping on vfork() */
495 tsk
->vfork_done
= NULL
;
496 complete(vfork_done
);
500 * If we're exiting normally, clear a user-space tid field if
501 * requested. We leave this alone when dying by signal, to leave
502 * the value intact in a core dump, and to save the unnecessary
503 * trouble otherwise. Userland only wants this done for a sys_exit.
505 if (tsk
->clear_child_tid
506 && !(tsk
->flags
& PF_SIGNALED
)
507 && atomic_read(&mm
->mm_users
) > 1) {
508 u32 __user
* tidptr
= tsk
->clear_child_tid
;
509 tsk
->clear_child_tid
= NULL
;
512 * We don't check the error code - if userspace has
513 * not set up a proper pointer then tough luck.
516 sys_futex(tidptr
, FUTEX_WAKE
, 1, NULL
, NULL
, 0);
521 * Allocate a new mm structure and copy contents from the
522 * mm structure of the passed in task structure.
524 static struct mm_struct
*dup_mm(struct task_struct
*tsk
)
526 struct mm_struct
*mm
, *oldmm
= current
->mm
;
536 memcpy(mm
, oldmm
, sizeof(*mm
));
538 /* Initializing for Swap token stuff */
539 mm
->token_priority
= 0;
540 mm
->last_interval
= 0;
542 if (!mm_init(mm
, tsk
))
545 if (init_new_context(tsk
, mm
))
548 err
= dup_mmap(mm
, oldmm
);
552 mm
->hiwater_rss
= get_mm_rss(mm
);
553 mm
->hiwater_vm
= mm
->total_vm
;
565 * If init_new_context() failed, we cannot use mmput() to free the mm
566 * because it calls destroy_context()
573 static int copy_mm(unsigned long clone_flags
, struct task_struct
* tsk
)
575 struct mm_struct
* mm
, *oldmm
;
578 tsk
->min_flt
= tsk
->maj_flt
= 0;
579 tsk
->nvcsw
= tsk
->nivcsw
= 0;
582 tsk
->active_mm
= NULL
;
585 * Are we cloning a kernel thread?
587 * We need to steal a active VM for that..
593 if (clone_flags
& CLONE_VM
) {
594 atomic_inc(&oldmm
->mm_users
);
605 /* Initializing for Swap token stuff */
606 mm
->token_priority
= 0;
607 mm
->last_interval
= 0;
617 static struct fs_struct
*__copy_fs_struct(struct fs_struct
*old
)
619 struct fs_struct
*fs
= kmem_cache_alloc(fs_cachep
, GFP_KERNEL
);
620 /* We don't need to lock fs - think why ;-) */
622 atomic_set(&fs
->count
, 1);
623 rwlock_init(&fs
->lock
);
624 fs
->umask
= old
->umask
;
625 read_lock(&old
->lock
);
626 fs
->root
= old
->root
;
627 path_get(&old
->root
);
630 if (old
->altroot
.dentry
) {
631 fs
->altroot
= old
->altroot
;
632 path_get(&old
->altroot
);
634 fs
->altroot
.mnt
= NULL
;
635 fs
->altroot
.dentry
= NULL
;
637 read_unlock(&old
->lock
);
642 struct fs_struct
*copy_fs_struct(struct fs_struct
*old
)
644 return __copy_fs_struct(old
);
647 EXPORT_SYMBOL_GPL(copy_fs_struct
);
649 static int copy_fs(unsigned long clone_flags
, struct task_struct
*tsk
)
651 if (clone_flags
& CLONE_FS
) {
652 atomic_inc(¤t
->fs
->count
);
655 tsk
->fs
= __copy_fs_struct(current
->fs
);
661 static int count_open_files(struct fdtable
*fdt
)
663 int size
= fdt
->max_fds
;
666 /* Find the last open fd */
667 for (i
= size
/(8*sizeof(long)); i
> 0; ) {
668 if (fdt
->open_fds
->fds_bits
[--i
])
671 i
= (i
+1) * 8 * sizeof(long);
675 static struct files_struct
*alloc_files(void)
677 struct files_struct
*newf
;
680 newf
= kmem_cache_alloc(files_cachep
, GFP_KERNEL
);
684 atomic_set(&newf
->count
, 1);
686 spin_lock_init(&newf
->file_lock
);
689 fdt
->max_fds
= NR_OPEN_DEFAULT
;
690 fdt
->close_on_exec
= (fd_set
*)&newf
->close_on_exec_init
;
691 fdt
->open_fds
= (fd_set
*)&newf
->open_fds_init
;
692 fdt
->fd
= &newf
->fd_array
[0];
693 INIT_RCU_HEAD(&fdt
->rcu
);
695 rcu_assign_pointer(newf
->fdt
, fdt
);
701 * Allocate a new files structure and copy contents from the
702 * passed in files structure.
703 * errorp will be valid only when the returned files_struct is NULL.
705 static struct files_struct
*dup_fd(struct files_struct
*oldf
, int *errorp
)
707 struct files_struct
*newf
;
708 struct file
**old_fds
, **new_fds
;
709 int open_files
, size
, i
;
710 struct fdtable
*old_fdt
, *new_fdt
;
713 newf
= alloc_files();
717 spin_lock(&oldf
->file_lock
);
718 old_fdt
= files_fdtable(oldf
);
719 new_fdt
= files_fdtable(newf
);
720 open_files
= count_open_files(old_fdt
);
723 * Check whether we need to allocate a larger fd array and fd set.
724 * Note: we're not a clone task, so the open count won't change.
726 if (open_files
> new_fdt
->max_fds
) {
727 new_fdt
->max_fds
= 0;
728 spin_unlock(&oldf
->file_lock
);
729 spin_lock(&newf
->file_lock
);
730 *errorp
= expand_files(newf
, open_files
-1);
731 spin_unlock(&newf
->file_lock
);
734 new_fdt
= files_fdtable(newf
);
736 * Reacquire the oldf lock and a pointer to its fd table
737 * who knows it may have a new bigger fd table. We need
738 * the latest pointer.
740 spin_lock(&oldf
->file_lock
);
741 old_fdt
= files_fdtable(oldf
);
744 old_fds
= old_fdt
->fd
;
745 new_fds
= new_fdt
->fd
;
747 memcpy(new_fdt
->open_fds
->fds_bits
,
748 old_fdt
->open_fds
->fds_bits
, open_files
/8);
749 memcpy(new_fdt
->close_on_exec
->fds_bits
,
750 old_fdt
->close_on_exec
->fds_bits
, open_files
/8);
752 for (i
= open_files
; i
!= 0; i
--) {
753 struct file
*f
= *old_fds
++;
758 * The fd may be claimed in the fd bitmap but not yet
759 * instantiated in the files array if a sibling thread
760 * is partway through open(). So make sure that this
761 * fd is available to the new process.
763 FD_CLR(open_files
- i
, new_fdt
->open_fds
);
765 rcu_assign_pointer(*new_fds
++, f
);
767 spin_unlock(&oldf
->file_lock
);
769 /* compute the remainder to be cleared */
770 size
= (new_fdt
->max_fds
- open_files
) * sizeof(struct file
*);
772 /* This is long word aligned thus could use a optimized version */
773 memset(new_fds
, 0, size
);
775 if (new_fdt
->max_fds
> open_files
) {
776 int left
= (new_fdt
->max_fds
-open_files
)/8;
777 int start
= open_files
/ (8 * sizeof(unsigned long));
779 memset(&new_fdt
->open_fds
->fds_bits
[start
], 0, left
);
780 memset(&new_fdt
->close_on_exec
->fds_bits
[start
], 0, left
);
786 kmem_cache_free(files_cachep
, newf
);
791 static int copy_files(unsigned long clone_flags
, struct task_struct
* tsk
)
793 struct files_struct
*oldf
, *newf
;
797 * A background process may not have any files ...
799 oldf
= current
->files
;
803 if (clone_flags
& CLONE_FILES
) {
804 atomic_inc(&oldf
->count
);
808 newf
= dup_fd(oldf
, &error
);
818 static int copy_io(unsigned long clone_flags
, struct task_struct
*tsk
)
821 struct io_context
*ioc
= current
->io_context
;
826 * Share io context with parent, if CLONE_IO is set
828 if (clone_flags
& CLONE_IO
) {
829 tsk
->io_context
= ioc_task_link(ioc
);
830 if (unlikely(!tsk
->io_context
))
832 } else if (ioprio_valid(ioc
->ioprio
)) {
833 tsk
->io_context
= alloc_io_context(GFP_KERNEL
, -1);
834 if (unlikely(!tsk
->io_context
))
837 tsk
->io_context
->ioprio
= ioc
->ioprio
;
844 * Helper to unshare the files of the current task.
845 * We don't want to expose copy_files internals to
846 * the exec layer of the kernel.
849 int unshare_files(void)
851 struct files_struct
*files
= current
->files
;
852 struct files_struct
*newf
;
857 /* This can race but the race causes us to copy when we don't
858 need to and drop the copy */
859 if(atomic_read(&files
->count
) == 1)
861 atomic_inc(&files
->count
);
864 newf
= dup_fd(files
, &error
);
867 current
->files
= newf
;
868 task_unlock(current
);
873 static int copy_sighand(unsigned long clone_flags
, struct task_struct
*tsk
)
875 struct sighand_struct
*sig
;
877 if (clone_flags
& (CLONE_SIGHAND
| CLONE_THREAD
)) {
878 atomic_inc(¤t
->sighand
->count
);
881 sig
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
882 rcu_assign_pointer(tsk
->sighand
, sig
);
885 atomic_set(&sig
->count
, 1);
886 memcpy(sig
->action
, current
->sighand
->action
, sizeof(sig
->action
));
890 void __cleanup_sighand(struct sighand_struct
*sighand
)
892 if (atomic_dec_and_test(&sighand
->count
))
893 kmem_cache_free(sighand_cachep
, sighand
);
896 static int copy_signal(unsigned long clone_flags
, struct task_struct
*tsk
)
898 struct signal_struct
*sig
;
901 if (clone_flags
& CLONE_THREAD
) {
902 atomic_inc(¤t
->signal
->count
);
903 atomic_inc(¤t
->signal
->live
);
906 sig
= kmem_cache_alloc(signal_cachep
, GFP_KERNEL
);
911 ret
= copy_thread_group_keys(tsk
);
913 kmem_cache_free(signal_cachep
, sig
);
917 atomic_set(&sig
->count
, 1);
918 atomic_set(&sig
->live
, 1);
919 init_waitqueue_head(&sig
->wait_chldexit
);
921 sig
->group_exit_code
= 0;
922 sig
->group_exit_task
= NULL
;
923 sig
->group_stop_count
= 0;
924 sig
->curr_target
= NULL
;
925 init_sigpending(&sig
->shared_pending
);
926 INIT_LIST_HEAD(&sig
->posix_timers
);
928 hrtimer_init(&sig
->real_timer
, CLOCK_MONOTONIC
, HRTIMER_MODE_REL
);
929 sig
->it_real_incr
.tv64
= 0;
930 sig
->real_timer
.function
= it_real_fn
;
932 sig
->it_virt_expires
= cputime_zero
;
933 sig
->it_virt_incr
= cputime_zero
;
934 sig
->it_prof_expires
= cputime_zero
;
935 sig
->it_prof_incr
= cputime_zero
;
937 sig
->leader
= 0; /* session leadership doesn't inherit */
938 sig
->tty_old_pgrp
= NULL
;
940 sig
->utime
= sig
->stime
= sig
->cutime
= sig
->cstime
= cputime_zero
;
941 sig
->gtime
= cputime_zero
;
942 sig
->cgtime
= cputime_zero
;
943 sig
->nvcsw
= sig
->nivcsw
= sig
->cnvcsw
= sig
->cnivcsw
= 0;
944 sig
->min_flt
= sig
->maj_flt
= sig
->cmin_flt
= sig
->cmaj_flt
= 0;
945 sig
->inblock
= sig
->oublock
= sig
->cinblock
= sig
->coublock
= 0;
946 sig
->sum_sched_runtime
= 0;
947 INIT_LIST_HEAD(&sig
->cpu_timers
[0]);
948 INIT_LIST_HEAD(&sig
->cpu_timers
[1]);
949 INIT_LIST_HEAD(&sig
->cpu_timers
[2]);
950 taskstats_tgid_init(sig
);
952 task_lock(current
->group_leader
);
953 memcpy(sig
->rlim
, current
->signal
->rlim
, sizeof sig
->rlim
);
954 task_unlock(current
->group_leader
);
956 if (sig
->rlim
[RLIMIT_CPU
].rlim_cur
!= RLIM_INFINITY
) {
958 * New sole thread in the process gets an expiry time
959 * of the whole CPU time limit.
961 tsk
->it_prof_expires
=
962 secs_to_cputime(sig
->rlim
[RLIMIT_CPU
].rlim_cur
);
964 acct_init_pacct(&sig
->pacct
);
971 void __cleanup_signal(struct signal_struct
*sig
)
973 exit_thread_group_keys(sig
);
974 kmem_cache_free(signal_cachep
, sig
);
977 static void cleanup_signal(struct task_struct
*tsk
)
979 struct signal_struct
*sig
= tsk
->signal
;
981 atomic_dec(&sig
->live
);
983 if (atomic_dec_and_test(&sig
->count
))
984 __cleanup_signal(sig
);
987 static void copy_flags(unsigned long clone_flags
, struct task_struct
*p
)
989 unsigned long new_flags
= p
->flags
;
991 new_flags
&= ~PF_SUPERPRIV
;
992 new_flags
|= PF_FORKNOEXEC
;
993 if (!(clone_flags
& CLONE_PTRACE
))
995 p
->flags
= new_flags
;
996 clear_freeze_flag(p
);
999 asmlinkage
long sys_set_tid_address(int __user
*tidptr
)
1001 current
->clear_child_tid
= tidptr
;
1003 return task_pid_vnr(current
);
1006 static void rt_mutex_init_task(struct task_struct
*p
)
1008 spin_lock_init(&p
->pi_lock
);
1009 #ifdef CONFIG_RT_MUTEXES
1010 plist_head_init(&p
->pi_waiters
, &p
->pi_lock
);
1011 p
->pi_blocked_on
= NULL
;
1016 * This creates a new process as a copy of the old one,
1017 * but does not actually start it yet.
1019 * It copies the registers, and all the appropriate
1020 * parts of the process environment (as per the clone
1021 * flags). The actual kick-off is left to the caller.
1023 static struct task_struct
*copy_process(unsigned long clone_flags
,
1024 unsigned long stack_start
,
1025 struct pt_regs
*regs
,
1026 unsigned long stack_size
,
1027 int __user
*child_tidptr
,
1031 struct task_struct
*p
;
1032 int cgroup_callbacks_done
= 0;
1034 if ((clone_flags
& (CLONE_NEWNS
|CLONE_FS
)) == (CLONE_NEWNS
|CLONE_FS
))
1035 return ERR_PTR(-EINVAL
);
1038 * Thread groups must share signals as well, and detached threads
1039 * can only be started up within the thread group.
1041 if ((clone_flags
& CLONE_THREAD
) && !(clone_flags
& CLONE_SIGHAND
))
1042 return ERR_PTR(-EINVAL
);
1045 * Shared signal handlers imply shared VM. By way of the above,
1046 * thread groups also imply shared VM. Blocking this case allows
1047 * for various simplifications in other code.
1049 if ((clone_flags
& CLONE_SIGHAND
) && !(clone_flags
& CLONE_VM
))
1050 return ERR_PTR(-EINVAL
);
1052 retval
= security_task_create(clone_flags
);
1057 p
= dup_task_struct(current
);
1061 rt_mutex_init_task(p
);
1063 #ifdef CONFIG_TRACE_IRQFLAGS
1064 DEBUG_LOCKS_WARN_ON(!p
->hardirqs_enabled
);
1065 DEBUG_LOCKS_WARN_ON(!p
->softirqs_enabled
);
1068 if (atomic_read(&p
->user
->processes
) >=
1069 p
->signal
->rlim
[RLIMIT_NPROC
].rlim_cur
) {
1070 if (!capable(CAP_SYS_ADMIN
) && !capable(CAP_SYS_RESOURCE
) &&
1071 p
->user
!= current
->nsproxy
->user_ns
->root_user
)
1075 atomic_inc(&p
->user
->__count
);
1076 atomic_inc(&p
->user
->processes
);
1077 get_group_info(p
->group_info
);
1080 * If multiple threads are within copy_process(), then this check
1081 * triggers too late. This doesn't hurt, the check is only there
1082 * to stop root fork bombs.
1084 if (nr_threads
>= max_threads
)
1085 goto bad_fork_cleanup_count
;
1087 if (!try_module_get(task_thread_info(p
)->exec_domain
->module
))
1088 goto bad_fork_cleanup_count
;
1090 if (p
->binfmt
&& !try_module_get(p
->binfmt
->module
))
1091 goto bad_fork_cleanup_put_domain
;
1094 delayacct_tsk_init(p
); /* Must remain after dup_task_struct() */
1095 copy_flags(clone_flags
, p
);
1096 INIT_LIST_HEAD(&p
->children
);
1097 INIT_LIST_HEAD(&p
->sibling
);
1098 #ifdef CONFIG_PREEMPT_RCU
1099 p
->rcu_read_lock_nesting
= 0;
1100 p
->rcu_flipctr_idx
= 0;
1101 #endif /* #ifdef CONFIG_PREEMPT_RCU */
1102 p
->vfork_done
= NULL
;
1103 spin_lock_init(&p
->alloc_lock
);
1105 clear_tsk_thread_flag(p
, TIF_SIGPENDING
);
1106 init_sigpending(&p
->pending
);
1108 p
->utime
= cputime_zero
;
1109 p
->stime
= cputime_zero
;
1110 p
->gtime
= cputime_zero
;
1111 p
->utimescaled
= cputime_zero
;
1112 p
->stimescaled
= cputime_zero
;
1113 p
->prev_utime
= cputime_zero
;
1114 p
->prev_stime
= cputime_zero
;
1116 #ifdef CONFIG_DETECT_SOFTLOCKUP
1117 p
->last_switch_count
= 0;
1118 p
->last_switch_timestamp
= 0;
1121 #ifdef CONFIG_TASK_XACCT
1122 p
->rchar
= 0; /* I/O counter: bytes read */
1123 p
->wchar
= 0; /* I/O counter: bytes written */
1124 p
->syscr
= 0; /* I/O counter: read syscalls */
1125 p
->syscw
= 0; /* I/O counter: write syscalls */
1127 task_io_accounting_init(p
);
1128 acct_clear_integrals(p
);
1130 p
->it_virt_expires
= cputime_zero
;
1131 p
->it_prof_expires
= cputime_zero
;
1132 p
->it_sched_expires
= 0;
1133 INIT_LIST_HEAD(&p
->cpu_timers
[0]);
1134 INIT_LIST_HEAD(&p
->cpu_timers
[1]);
1135 INIT_LIST_HEAD(&p
->cpu_timers
[2]);
1137 p
->lock_depth
= -1; /* -1 = no lock */
1138 do_posix_clock_monotonic_gettime(&p
->start_time
);
1139 p
->real_start_time
= p
->start_time
;
1140 monotonic_to_bootbased(&p
->real_start_time
);
1141 #ifdef CONFIG_SECURITY
1144 p
->cap_bset
= current
->cap_bset
;
1145 p
->io_context
= NULL
;
1146 p
->audit_context
= NULL
;
1149 p
->mempolicy
= mpol_copy(p
->mempolicy
);
1150 if (IS_ERR(p
->mempolicy
)) {
1151 retval
= PTR_ERR(p
->mempolicy
);
1152 p
->mempolicy
= NULL
;
1153 goto bad_fork_cleanup_cgroup
;
1155 mpol_fix_fork_child_flag(p
);
1157 #ifdef CONFIG_TRACE_IRQFLAGS
1159 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1160 p
->hardirqs_enabled
= 1;
1162 p
->hardirqs_enabled
= 0;
1164 p
->hardirq_enable_ip
= 0;
1165 p
->hardirq_enable_event
= 0;
1166 p
->hardirq_disable_ip
= _THIS_IP_
;
1167 p
->hardirq_disable_event
= 0;
1168 p
->softirqs_enabled
= 1;
1169 p
->softirq_enable_ip
= _THIS_IP_
;
1170 p
->softirq_enable_event
= 0;
1171 p
->softirq_disable_ip
= 0;
1172 p
->softirq_disable_event
= 0;
1173 p
->hardirq_context
= 0;
1174 p
->softirq_context
= 0;
1176 #ifdef CONFIG_LOCKDEP
1177 p
->lockdep_depth
= 0; /* no locks held yet */
1178 p
->curr_chain_key
= 0;
1179 p
->lockdep_recursion
= 0;
1182 #ifdef CONFIG_DEBUG_MUTEXES
1183 p
->blocked_on
= NULL
; /* not blocked yet */
1186 /* Perform scheduler related setup. Assign this task to a CPU. */
1187 sched_fork(p
, clone_flags
);
1189 if ((retval
= security_task_alloc(p
)))
1190 goto bad_fork_cleanup_policy
;
1191 if ((retval
= audit_alloc(p
)))
1192 goto bad_fork_cleanup_security
;
1193 /* copy all the process information */
1194 if ((retval
= copy_semundo(clone_flags
, p
)))
1195 goto bad_fork_cleanup_audit
;
1196 if ((retval
= copy_files(clone_flags
, p
)))
1197 goto bad_fork_cleanup_semundo
;
1198 if ((retval
= copy_fs(clone_flags
, p
)))
1199 goto bad_fork_cleanup_files
;
1200 if ((retval
= copy_sighand(clone_flags
, p
)))
1201 goto bad_fork_cleanup_fs
;
1202 if ((retval
= copy_signal(clone_flags
, p
)))
1203 goto bad_fork_cleanup_sighand
;
1204 if ((retval
= copy_mm(clone_flags
, p
)))
1205 goto bad_fork_cleanup_signal
;
1206 if ((retval
= copy_keys(clone_flags
, p
)))
1207 goto bad_fork_cleanup_mm
;
1208 if ((retval
= copy_namespaces(clone_flags
, p
)))
1209 goto bad_fork_cleanup_keys
;
1210 if ((retval
= copy_io(clone_flags
, p
)))
1211 goto bad_fork_cleanup_namespaces
;
1212 retval
= copy_thread(0, clone_flags
, stack_start
, stack_size
, p
, regs
);
1214 goto bad_fork_cleanup_io
;
1216 if (pid
!= &init_struct_pid
) {
1218 pid
= alloc_pid(task_active_pid_ns(p
));
1220 goto bad_fork_cleanup_io
;
1222 if (clone_flags
& CLONE_NEWPID
) {
1223 retval
= pid_ns_prepare_proc(task_active_pid_ns(p
));
1225 goto bad_fork_free_pid
;
1229 p
->pid
= pid_nr(pid
);
1231 if (clone_flags
& CLONE_THREAD
)
1232 p
->tgid
= current
->tgid
;
1234 p
->set_child_tid
= (clone_flags
& CLONE_CHILD_SETTID
) ? child_tidptr
: NULL
;
1236 * Clear TID on mm_release()?
1238 p
->clear_child_tid
= (clone_flags
& CLONE_CHILD_CLEARTID
) ? child_tidptr
: NULL
;
1240 p
->robust_list
= NULL
;
1241 #ifdef CONFIG_COMPAT
1242 p
->compat_robust_list
= NULL
;
1244 INIT_LIST_HEAD(&p
->pi_state_list
);
1245 p
->pi_state_cache
= NULL
;
1248 * sigaltstack should be cleared when sharing the same VM
1250 if ((clone_flags
& (CLONE_VM
|CLONE_VFORK
)) == CLONE_VM
)
1251 p
->sas_ss_sp
= p
->sas_ss_size
= 0;
1254 * Syscall tracing should be turned off in the child regardless
1257 clear_tsk_thread_flag(p
, TIF_SYSCALL_TRACE
);
1258 #ifdef TIF_SYSCALL_EMU
1259 clear_tsk_thread_flag(p
, TIF_SYSCALL_EMU
);
1261 clear_all_latency_tracing(p
);
1263 /* Our parent execution domain becomes current domain
1264 These must match for thread signalling to apply */
1265 p
->parent_exec_id
= p
->self_exec_id
;
1267 /* ok, now we should be set up.. */
1268 p
->exit_signal
= (clone_flags
& CLONE_THREAD
) ? -1 : (clone_flags
& CSIGNAL
);
1269 p
->pdeath_signal
= 0;
1273 * Ok, make it visible to the rest of the system.
1274 * We dont wake it up yet.
1276 p
->group_leader
= p
;
1277 INIT_LIST_HEAD(&p
->thread_group
);
1278 INIT_LIST_HEAD(&p
->ptrace_children
);
1279 INIT_LIST_HEAD(&p
->ptrace_list
);
1281 /* Now that the task is set up, run cgroup callbacks if
1282 * necessary. We need to run them before the task is visible
1283 * on the tasklist. */
1284 cgroup_fork_callbacks(p
);
1285 cgroup_callbacks_done
= 1;
1287 /* Need tasklist lock for parent etc handling! */
1288 write_lock_irq(&tasklist_lock
);
1291 * The task hasn't been attached yet, so its cpus_allowed mask will
1292 * not be changed, nor will its assigned CPU.
1294 * The cpus_allowed mask of the parent may have changed after it was
1295 * copied first time - so re-copy it here, then check the child's CPU
1296 * to ensure it is on a valid CPU (and if not, just force it back to
1297 * parent's CPU). This avoids alot of nasty races.
1299 p
->cpus_allowed
= current
->cpus_allowed
;
1300 p
->rt
.nr_cpus_allowed
= current
->rt
.nr_cpus_allowed
;
1301 if (unlikely(!cpu_isset(task_cpu(p
), p
->cpus_allowed
) ||
1302 !cpu_online(task_cpu(p
))))
1303 set_task_cpu(p
, smp_processor_id());
1305 /* CLONE_PARENT re-uses the old parent */
1306 if (clone_flags
& (CLONE_PARENT
|CLONE_THREAD
))
1307 p
->real_parent
= current
->real_parent
;
1309 p
->real_parent
= current
;
1310 p
->parent
= p
->real_parent
;
1312 spin_lock(¤t
->sighand
->siglock
);
1315 * Process group and session signals need to be delivered to just the
1316 * parent before the fork or both the parent and the child after the
1317 * fork. Restart if a signal comes in before we add the new process to
1318 * it's process group.
1319 * A fatal signal pending means that current will exit, so the new
1320 * thread can't slip out of an OOM kill (or normal SIGKILL).
1322 recalc_sigpending();
1323 if (signal_pending(current
)) {
1324 spin_unlock(¤t
->sighand
->siglock
);
1325 write_unlock_irq(&tasklist_lock
);
1326 retval
= -ERESTARTNOINTR
;
1327 goto bad_fork_free_pid
;
1330 if (clone_flags
& CLONE_THREAD
) {
1331 p
->group_leader
= current
->group_leader
;
1332 list_add_tail_rcu(&p
->thread_group
, &p
->group_leader
->thread_group
);
1334 if (!cputime_eq(current
->signal
->it_virt_expires
,
1336 !cputime_eq(current
->signal
->it_prof_expires
,
1338 current
->signal
->rlim
[RLIMIT_CPU
].rlim_cur
!= RLIM_INFINITY
||
1339 !list_empty(¤t
->signal
->cpu_timers
[0]) ||
1340 !list_empty(¤t
->signal
->cpu_timers
[1]) ||
1341 !list_empty(¤t
->signal
->cpu_timers
[2])) {
1343 * Have child wake up on its first tick to check
1344 * for process CPU timers.
1346 p
->it_prof_expires
= jiffies_to_cputime(1);
1350 if (likely(p
->pid
)) {
1352 if (unlikely(p
->ptrace
& PT_PTRACED
))
1353 __ptrace_link(p
, current
->parent
);
1355 if (thread_group_leader(p
)) {
1356 if (clone_flags
& CLONE_NEWPID
)
1357 p
->nsproxy
->pid_ns
->child_reaper
= p
;
1359 p
->signal
->leader_pid
= pid
;
1360 p
->signal
->tty
= current
->signal
->tty
;
1361 set_task_pgrp(p
, task_pgrp_nr(current
));
1362 set_task_session(p
, task_session_nr(current
));
1363 attach_pid(p
, PIDTYPE_PGID
, task_pgrp(current
));
1364 attach_pid(p
, PIDTYPE_SID
, task_session(current
));
1365 list_add_tail_rcu(&p
->tasks
, &init_task
.tasks
);
1366 __get_cpu_var(process_counts
)++;
1368 attach_pid(p
, PIDTYPE_PID
, pid
);
1373 spin_unlock(¤t
->sighand
->siglock
);
1374 write_unlock_irq(&tasklist_lock
);
1375 proc_fork_connector(p
);
1376 cgroup_post_fork(p
);
1380 if (pid
!= &init_struct_pid
)
1382 bad_fork_cleanup_io
:
1383 put_io_context(p
->io_context
);
1384 bad_fork_cleanup_namespaces
:
1385 exit_task_namespaces(p
);
1386 bad_fork_cleanup_keys
:
1388 bad_fork_cleanup_mm
:
1391 bad_fork_cleanup_signal
:
1393 bad_fork_cleanup_sighand
:
1394 __cleanup_sighand(p
->sighand
);
1395 bad_fork_cleanup_fs
:
1396 exit_fs(p
); /* blocking */
1397 bad_fork_cleanup_files
:
1398 exit_files(p
); /* blocking */
1399 bad_fork_cleanup_semundo
:
1401 bad_fork_cleanup_audit
:
1403 bad_fork_cleanup_security
:
1404 security_task_free(p
);
1405 bad_fork_cleanup_policy
:
1407 mpol_free(p
->mempolicy
);
1408 bad_fork_cleanup_cgroup
:
1410 cgroup_exit(p
, cgroup_callbacks_done
);
1411 delayacct_tsk_free(p
);
1413 module_put(p
->binfmt
->module
);
1414 bad_fork_cleanup_put_domain
:
1415 module_put(task_thread_info(p
)->exec_domain
->module
);
1416 bad_fork_cleanup_count
:
1417 put_group_info(p
->group_info
);
1418 atomic_dec(&p
->user
->processes
);
1423 return ERR_PTR(retval
);
1426 noinline
struct pt_regs
* __cpuinit
__attribute__((weak
)) idle_regs(struct pt_regs
*regs
)
1428 memset(regs
, 0, sizeof(struct pt_regs
));
1432 struct task_struct
* __cpuinit
fork_idle(int cpu
)
1434 struct task_struct
*task
;
1435 struct pt_regs regs
;
1437 task
= copy_process(CLONE_VM
, 0, idle_regs(®s
), 0, NULL
,
1440 init_idle(task
, cpu
);
1445 static int fork_traceflag(unsigned clone_flags
)
1447 if (clone_flags
& CLONE_UNTRACED
)
1449 else if (clone_flags
& CLONE_VFORK
) {
1450 if (current
->ptrace
& PT_TRACE_VFORK
)
1451 return PTRACE_EVENT_VFORK
;
1452 } else if ((clone_flags
& CSIGNAL
) != SIGCHLD
) {
1453 if (current
->ptrace
& PT_TRACE_CLONE
)
1454 return PTRACE_EVENT_CLONE
;
1455 } else if (current
->ptrace
& PT_TRACE_FORK
)
1456 return PTRACE_EVENT_FORK
;
1462 * Ok, this is the main fork-routine.
1464 * It copies the process, and if successful kick-starts
1465 * it and waits for it to finish using the VM if required.
1467 long do_fork(unsigned long clone_flags
,
1468 unsigned long stack_start
,
1469 struct pt_regs
*regs
,
1470 unsigned long stack_size
,
1471 int __user
*parent_tidptr
,
1472 int __user
*child_tidptr
)
1474 struct task_struct
*p
;
1479 * We hope to recycle these flags after 2.6.26
1481 if (unlikely(clone_flags
& CLONE_STOPPED
)) {
1482 static int __read_mostly count
= 100;
1484 if (count
> 0 && printk_ratelimit()) {
1485 char comm
[TASK_COMM_LEN
];
1488 printk(KERN_INFO
"fork(): process `%s' used deprecated "
1489 "clone flags 0x%lx\n",
1490 get_task_comm(comm
, current
),
1491 clone_flags
& CLONE_STOPPED
);
1495 if (unlikely(current
->ptrace
)) {
1496 trace
= fork_traceflag (clone_flags
);
1498 clone_flags
|= CLONE_PTRACE
;
1501 p
= copy_process(clone_flags
, stack_start
, regs
, stack_size
,
1502 child_tidptr
, NULL
);
1504 * Do this prior waking up the new thread - the thread pointer
1505 * might get invalid after that point, if the thread exits quickly.
1508 struct completion vfork
;
1510 nr
= task_pid_vnr(p
);
1512 if (clone_flags
& CLONE_PARENT_SETTID
)
1513 put_user(nr
, parent_tidptr
);
1515 if (clone_flags
& CLONE_VFORK
) {
1516 p
->vfork_done
= &vfork
;
1517 init_completion(&vfork
);
1520 if ((p
->ptrace
& PT_PTRACED
) || (clone_flags
& CLONE_STOPPED
)) {
1522 * We'll start up with an immediate SIGSTOP.
1524 sigaddset(&p
->pending
.signal
, SIGSTOP
);
1525 set_tsk_thread_flag(p
, TIF_SIGPENDING
);
1528 if (!(clone_flags
& CLONE_STOPPED
))
1529 wake_up_new_task(p
, clone_flags
);
1531 __set_task_state(p
, TASK_STOPPED
);
1533 if (unlikely (trace
)) {
1534 current
->ptrace_message
= nr
;
1535 ptrace_notify ((trace
<< 8) | SIGTRAP
);
1538 if (clone_flags
& CLONE_VFORK
) {
1539 freezer_do_not_count();
1540 wait_for_completion(&vfork
);
1542 if (unlikely (current
->ptrace
& PT_TRACE_VFORK_DONE
)) {
1543 current
->ptrace_message
= nr
;
1544 ptrace_notify ((PTRACE_EVENT_VFORK_DONE
<< 8) | SIGTRAP
);
1553 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1554 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1557 static void sighand_ctor(struct kmem_cache
*cachep
, void *data
)
1559 struct sighand_struct
*sighand
= data
;
1561 spin_lock_init(&sighand
->siglock
);
1562 init_waitqueue_head(&sighand
->signalfd_wqh
);
1565 void __init
proc_caches_init(void)
1567 sighand_cachep
= kmem_cache_create("sighand_cache",
1568 sizeof(struct sighand_struct
), 0,
1569 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_DESTROY_BY_RCU
,
1571 signal_cachep
= kmem_cache_create("signal_cache",
1572 sizeof(struct signal_struct
), 0,
1573 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
);
1574 files_cachep
= kmem_cache_create("files_cache",
1575 sizeof(struct files_struct
), 0,
1576 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
);
1577 fs_cachep
= kmem_cache_create("fs_cache",
1578 sizeof(struct fs_struct
), 0,
1579 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
);
1580 vm_area_cachep
= kmem_cache_create("vm_area_struct",
1581 sizeof(struct vm_area_struct
), 0,
1583 mm_cachep
= kmem_cache_create("mm_struct",
1584 sizeof(struct mm_struct
), ARCH_MIN_MMSTRUCT_ALIGN
,
1585 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
);
1589 * Check constraints on flags passed to the unshare system call and
1590 * force unsharing of additional process context as appropriate.
1592 static void check_unshare_flags(unsigned long *flags_ptr
)
1595 * If unsharing a thread from a thread group, must also
1598 if (*flags_ptr
& CLONE_THREAD
)
1599 *flags_ptr
|= CLONE_VM
;
1602 * If unsharing vm, must also unshare signal handlers.
1604 if (*flags_ptr
& CLONE_VM
)
1605 *flags_ptr
|= CLONE_SIGHAND
;
1608 * If unsharing signal handlers and the task was created
1609 * using CLONE_THREAD, then must unshare the thread
1611 if ((*flags_ptr
& CLONE_SIGHAND
) &&
1612 (atomic_read(¤t
->signal
->count
) > 1))
1613 *flags_ptr
|= CLONE_THREAD
;
1616 * If unsharing namespace, must also unshare filesystem information.
1618 if (*flags_ptr
& CLONE_NEWNS
)
1619 *flags_ptr
|= CLONE_FS
;
1623 * Unsharing of tasks created with CLONE_THREAD is not supported yet
1625 static int unshare_thread(unsigned long unshare_flags
)
1627 if (unshare_flags
& CLONE_THREAD
)
1634 * Unshare the filesystem structure if it is being shared
1636 static int unshare_fs(unsigned long unshare_flags
, struct fs_struct
**new_fsp
)
1638 struct fs_struct
*fs
= current
->fs
;
1640 if ((unshare_flags
& CLONE_FS
) &&
1641 (fs
&& atomic_read(&fs
->count
) > 1)) {
1642 *new_fsp
= __copy_fs_struct(current
->fs
);
1651 * Unsharing of sighand is not supported yet
1653 static int unshare_sighand(unsigned long unshare_flags
, struct sighand_struct
**new_sighp
)
1655 struct sighand_struct
*sigh
= current
->sighand
;
1657 if ((unshare_flags
& CLONE_SIGHAND
) && atomic_read(&sigh
->count
) > 1)
1664 * Unshare vm if it is being shared
1666 static int unshare_vm(unsigned long unshare_flags
, struct mm_struct
**new_mmp
)
1668 struct mm_struct
*mm
= current
->mm
;
1670 if ((unshare_flags
& CLONE_VM
) &&
1671 (mm
&& atomic_read(&mm
->mm_users
) > 1)) {
1679 * Unshare file descriptor table if it is being shared
1681 static int unshare_fd(unsigned long unshare_flags
, struct files_struct
**new_fdp
)
1683 struct files_struct
*fd
= current
->files
;
1686 if ((unshare_flags
& CLONE_FILES
) &&
1687 (fd
&& atomic_read(&fd
->count
) > 1)) {
1688 *new_fdp
= dup_fd(fd
, &error
);
1697 * Unsharing of semundo for tasks created with CLONE_SYSVSEM is not
1700 static int unshare_semundo(unsigned long unshare_flags
, struct sem_undo_list
**new_ulistp
)
1702 if (unshare_flags
& CLONE_SYSVSEM
)
1709 * unshare allows a process to 'unshare' part of the process
1710 * context which was originally shared using clone. copy_*
1711 * functions used by do_fork() cannot be used here directly
1712 * because they modify an inactive task_struct that is being
1713 * constructed. Here we are modifying the current, active,
1716 asmlinkage
long sys_unshare(unsigned long unshare_flags
)
1719 struct fs_struct
*fs
, *new_fs
= NULL
;
1720 struct sighand_struct
*new_sigh
= NULL
;
1721 struct mm_struct
*mm
, *new_mm
= NULL
, *active_mm
= NULL
;
1722 struct files_struct
*fd
, *new_fd
= NULL
;
1723 struct sem_undo_list
*new_ulist
= NULL
;
1724 struct nsproxy
*new_nsproxy
= NULL
;
1726 check_unshare_flags(&unshare_flags
);
1728 /* Return -EINVAL for all unsupported flags */
1730 if (unshare_flags
& ~(CLONE_THREAD
|CLONE_FS
|CLONE_NEWNS
|CLONE_SIGHAND
|
1731 CLONE_VM
|CLONE_FILES
|CLONE_SYSVSEM
|
1732 CLONE_NEWUTS
|CLONE_NEWIPC
|CLONE_NEWUSER
|
1734 goto bad_unshare_out
;
1736 if ((err
= unshare_thread(unshare_flags
)))
1737 goto bad_unshare_out
;
1738 if ((err
= unshare_fs(unshare_flags
, &new_fs
)))
1739 goto bad_unshare_cleanup_thread
;
1740 if ((err
= unshare_sighand(unshare_flags
, &new_sigh
)))
1741 goto bad_unshare_cleanup_fs
;
1742 if ((err
= unshare_vm(unshare_flags
, &new_mm
)))
1743 goto bad_unshare_cleanup_sigh
;
1744 if ((err
= unshare_fd(unshare_flags
, &new_fd
)))
1745 goto bad_unshare_cleanup_vm
;
1746 if ((err
= unshare_semundo(unshare_flags
, &new_ulist
)))
1747 goto bad_unshare_cleanup_fd
;
1748 if ((err
= unshare_nsproxy_namespaces(unshare_flags
, &new_nsproxy
,
1750 goto bad_unshare_cleanup_semundo
;
1752 if (new_fs
|| new_mm
|| new_fd
|| new_ulist
|| new_nsproxy
) {
1755 switch_task_namespaces(current
, new_nsproxy
);
1763 current
->fs
= new_fs
;
1769 active_mm
= current
->active_mm
;
1770 current
->mm
= new_mm
;
1771 current
->active_mm
= new_mm
;
1772 activate_mm(active_mm
, new_mm
);
1777 fd
= current
->files
;
1778 current
->files
= new_fd
;
1782 task_unlock(current
);
1786 put_nsproxy(new_nsproxy
);
1788 bad_unshare_cleanup_semundo
:
1789 bad_unshare_cleanup_fd
:
1791 put_files_struct(new_fd
);
1793 bad_unshare_cleanup_vm
:
1797 bad_unshare_cleanup_sigh
:
1799 if (atomic_dec_and_test(&new_sigh
->count
))
1800 kmem_cache_free(sighand_cachep
, new_sigh
);
1802 bad_unshare_cleanup_fs
:
1804 put_fs_struct(new_fs
);
1806 bad_unshare_cleanup_thread
: