4 * Copyright (C) 1991, 1992 Linus Torvalds
8 * 'fork.c' contains the help-routines for the 'fork' system call
9 * (see also entry.S and others).
10 * Fork is rather simple, once you get the hang of it, but the memory
11 * management can be a bitch. See 'mm/memory.c': 'copy_page_range()'
14 #include <linux/slab.h>
15 #include <linux/init.h>
16 #include <linux/unistd.h>
17 #include <linux/module.h>
18 #include <linux/vmalloc.h>
19 #include <linux/completion.h>
20 #include <linux/personality.h>
21 #include <linux/mempolicy.h>
22 #include <linux/sem.h>
23 #include <linux/file.h>
24 #include <linux/fdtable.h>
25 #include <linux/iocontext.h>
26 #include <linux/key.h>
27 #include <linux/binfmts.h>
28 #include <linux/mman.h>
29 #include <linux/mmu_notifier.h>
31 #include <linux/nsproxy.h>
32 #include <linux/capability.h>
33 #include <linux/cpu.h>
34 #include <linux/cgroup.h>
35 #include <linux/security.h>
36 #include <linux/hugetlb.h>
37 #include <linux/swap.h>
38 #include <linux/syscalls.h>
39 #include <linux/jiffies.h>
40 #include <linux/tracehook.h>
41 #include <linux/futex.h>
42 #include <linux/compat.h>
43 #include <linux/task_io_accounting_ops.h>
44 #include <linux/rcupdate.h>
45 #include <linux/ptrace.h>
46 #include <linux/mount.h>
47 #include <linux/audit.h>
48 #include <linux/memcontrol.h>
49 #include <linux/ftrace.h>
50 #include <linux/profile.h>
51 #include <linux/rmap.h>
52 #include <linux/ksm.h>
53 #include <linux/acct.h>
54 #include <linux/tsacct_kern.h>
55 #include <linux/cn_proc.h>
56 #include <linux/freezer.h>
57 #include <linux/delayacct.h>
58 #include <linux/taskstats_kern.h>
59 #include <linux/random.h>
60 #include <linux/tty.h>
61 #include <linux/proc_fs.h>
62 #include <linux/blkdev.h>
63 #include <linux/fs_struct.h>
64 #include <linux/magic.h>
65 #include <linux/perf_event.h>
66 #include <linux/posix-timers.h>
68 #include <asm/pgtable.h>
69 #include <asm/pgalloc.h>
70 #include <asm/uaccess.h>
71 #include <asm/mmu_context.h>
72 #include <asm/cacheflush.h>
73 #include <asm/tlbflush.h>
75 #include <trace/events/sched.h>
78 * Protected counters by write_lock_irq(&tasklist_lock)
80 unsigned long total_forks
; /* Handle normal Linux uptimes. */
81 int nr_threads
; /* The idle threads do not count.. */
83 int max_threads
; /* tunable limit on nr_threads */
85 DEFINE_PER_CPU(unsigned long, process_counts
) = 0;
87 __cacheline_aligned
DEFINE_RWLOCK(tasklist_lock
); /* outer */
89 int nr_processes(void)
94 for_each_possible_cpu(cpu
)
95 total
+= per_cpu(process_counts
, cpu
);
100 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
101 # define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL)
102 # define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk))
103 static struct kmem_cache
*task_struct_cachep
;
106 #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
107 static inline struct thread_info
*alloc_thread_info(struct task_struct
*tsk
)
109 #ifdef CONFIG_DEBUG_STACK_USAGE
110 gfp_t mask
= GFP_KERNEL
| __GFP_ZERO
;
112 gfp_t mask
= GFP_KERNEL
;
114 return (struct thread_info
*)__get_free_pages(mask
, THREAD_SIZE_ORDER
);
117 static inline void free_thread_info(struct thread_info
*ti
)
119 free_pages((unsigned long)ti
, THREAD_SIZE_ORDER
);
123 /* SLAB cache for signal_struct structures (tsk->signal) */
124 static struct kmem_cache
*signal_cachep
;
126 /* SLAB cache for sighand_struct structures (tsk->sighand) */
127 struct kmem_cache
*sighand_cachep
;
129 /* SLAB cache for files_struct structures (tsk->files) */
130 struct kmem_cache
*files_cachep
;
132 /* SLAB cache for fs_struct structures (tsk->fs) */
133 struct kmem_cache
*fs_cachep
;
135 /* SLAB cache for vm_area_struct structures */
136 struct kmem_cache
*vm_area_cachep
;
138 /* SLAB cache for mm_struct structures (tsk->mm) */
139 static struct kmem_cache
*mm_cachep
;
141 static void account_kernel_stack(struct thread_info
*ti
, int account
)
143 struct zone
*zone
= page_zone(virt_to_page(ti
));
145 mod_zone_page_state(zone
, NR_KERNEL_STACK
, account
);
148 void free_task(struct task_struct
*tsk
)
150 prop_local_destroy_single(&tsk
->dirties
);
151 account_kernel_stack(tsk
->stack
, -1);
152 free_thread_info(tsk
->stack
);
153 rt_mutex_debug_task_free(tsk
);
154 ftrace_graph_exit_task(tsk
);
155 free_task_struct(tsk
);
157 EXPORT_SYMBOL(free_task
);
159 void __put_task_struct(struct task_struct
*tsk
)
161 WARN_ON(!tsk
->exit_state
);
162 WARN_ON(atomic_read(&tsk
->usage
));
163 WARN_ON(tsk
== current
);
166 delayacct_tsk_free(tsk
);
168 if (!profile_handoff_task(tsk
))
173 * macro override instead of weak attribute alias, to workaround
174 * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
176 #ifndef arch_task_cache_init
177 #define arch_task_cache_init()
180 void __init
fork_init(unsigned long mempages
)
182 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
183 #ifndef ARCH_MIN_TASKALIGN
184 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
186 /* create a slab on which task_structs can be allocated */
188 kmem_cache_create("task_struct", sizeof(struct task_struct
),
189 ARCH_MIN_TASKALIGN
, SLAB_PANIC
| SLAB_NOTRACK
, NULL
);
192 /* do the arch specific task caches init */
193 arch_task_cache_init();
196 * The default maximum number of threads is set to a safe
197 * value: the thread structures can take up at most half
200 max_threads
= mempages
/ (8 * THREAD_SIZE
/ PAGE_SIZE
);
203 * we need to allow at least 20 threads to boot a system
208 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_cur
= max_threads
/2;
209 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_max
= max_threads
/2;
210 init_task
.signal
->rlim
[RLIMIT_SIGPENDING
] =
211 init_task
.signal
->rlim
[RLIMIT_NPROC
];
214 int __attribute__((weak
)) arch_dup_task_struct(struct task_struct
*dst
,
215 struct task_struct
*src
)
221 static struct task_struct
*dup_task_struct(struct task_struct
*orig
)
223 struct task_struct
*tsk
;
224 struct thread_info
*ti
;
225 unsigned long *stackend
;
229 prepare_to_copy(orig
);
231 tsk
= alloc_task_struct();
235 ti
= alloc_thread_info(tsk
);
237 free_task_struct(tsk
);
241 err
= arch_dup_task_struct(tsk
, orig
);
247 err
= prop_local_init_single(&tsk
->dirties
);
251 setup_thread_stack(tsk
, orig
);
252 stackend
= end_of_stack(tsk
);
253 *stackend
= STACK_END_MAGIC
; /* for overflow detection */
255 #ifdef CONFIG_CC_STACKPROTECTOR
256 tsk
->stack_canary
= get_random_int();
259 /* One for us, one for whoever does the "release_task()" (usually parent) */
260 atomic_set(&tsk
->usage
,2);
261 atomic_set(&tsk
->fs_excl
, 0);
262 #ifdef CONFIG_BLK_DEV_IO_TRACE
265 tsk
->splice_pipe
= NULL
;
267 account_kernel_stack(ti
, 1);
272 free_thread_info(ti
);
273 free_task_struct(tsk
);
278 static int dup_mmap(struct mm_struct
*mm
, struct mm_struct
*oldmm
)
280 struct vm_area_struct
*mpnt
, *tmp
, **pprev
;
281 struct rb_node
**rb_link
, *rb_parent
;
283 unsigned long charge
;
284 struct mempolicy
*pol
;
286 down_write(&oldmm
->mmap_sem
);
287 flush_cache_dup_mm(oldmm
);
289 * Not linked in yet - no deadlock potential:
291 down_write_nested(&mm
->mmap_sem
, SINGLE_DEPTH_NESTING
);
295 mm
->mmap_cache
= NULL
;
296 mm
->free_area_cache
= oldmm
->mmap_base
;
297 mm
->cached_hole_size
= ~0UL;
299 cpumask_clear(mm_cpumask(mm
));
301 rb_link
= &mm
->mm_rb
.rb_node
;
304 retval
= ksm_fork(mm
, oldmm
);
308 for (mpnt
= oldmm
->mmap
; mpnt
; mpnt
= mpnt
->vm_next
) {
311 if (mpnt
->vm_flags
& VM_DONTCOPY
) {
312 long pages
= vma_pages(mpnt
);
313 mm
->total_vm
-= pages
;
314 vm_stat_account(mm
, mpnt
->vm_flags
, mpnt
->vm_file
,
319 if (mpnt
->vm_flags
& VM_ACCOUNT
) {
320 unsigned int len
= (mpnt
->vm_end
- mpnt
->vm_start
) >> PAGE_SHIFT
;
321 if (security_vm_enough_memory(len
))
325 tmp
= kmem_cache_alloc(vm_area_cachep
, GFP_KERNEL
);
329 pol
= mpol_dup(vma_policy(mpnt
));
330 retval
= PTR_ERR(pol
);
332 goto fail_nomem_policy
;
333 vma_set_policy(tmp
, pol
);
334 tmp
->vm_flags
&= ~VM_LOCKED
;
340 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
341 struct address_space
*mapping
= file
->f_mapping
;
344 if (tmp
->vm_flags
& VM_DENYWRITE
)
345 atomic_dec(&inode
->i_writecount
);
346 spin_lock(&mapping
->i_mmap_lock
);
347 if (tmp
->vm_flags
& VM_SHARED
)
348 mapping
->i_mmap_writable
++;
349 tmp
->vm_truncate_count
= mpnt
->vm_truncate_count
;
350 flush_dcache_mmap_lock(mapping
);
351 /* insert tmp into the share list, just after mpnt */
352 vma_prio_tree_add(tmp
, mpnt
);
353 flush_dcache_mmap_unlock(mapping
);
354 spin_unlock(&mapping
->i_mmap_lock
);
358 * Clear hugetlb-related page reserves for children. This only
359 * affects MAP_PRIVATE mappings. Faults generated by the child
360 * are not guaranteed to succeed, even if read-only
362 if (is_vm_hugetlb_page(tmp
))
363 reset_vma_resv_huge_pages(tmp
);
366 * Link in the new vma and copy the page table entries.
369 pprev
= &tmp
->vm_next
;
371 __vma_link_rb(mm
, tmp
, rb_link
, rb_parent
);
372 rb_link
= &tmp
->vm_rb
.rb_right
;
373 rb_parent
= &tmp
->vm_rb
;
376 retval
= copy_page_range(mm
, oldmm
, mpnt
);
378 if (tmp
->vm_ops
&& tmp
->vm_ops
->open
)
379 tmp
->vm_ops
->open(tmp
);
384 /* a new mm has just been created */
385 arch_dup_mmap(oldmm
, mm
);
388 up_write(&mm
->mmap_sem
);
390 up_write(&oldmm
->mmap_sem
);
393 kmem_cache_free(vm_area_cachep
, tmp
);
396 vm_unacct_memory(charge
);
400 static inline int mm_alloc_pgd(struct mm_struct
* mm
)
402 mm
->pgd
= pgd_alloc(mm
);
403 if (unlikely(!mm
->pgd
))
408 static inline void mm_free_pgd(struct mm_struct
* mm
)
410 pgd_free(mm
, mm
->pgd
);
413 #define dup_mmap(mm, oldmm) (0)
414 #define mm_alloc_pgd(mm) (0)
415 #define mm_free_pgd(mm)
416 #endif /* CONFIG_MMU */
418 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(mmlist_lock
);
420 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
421 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
423 static unsigned long default_dump_filter
= MMF_DUMP_FILTER_DEFAULT
;
425 static int __init
coredump_filter_setup(char *s
)
427 default_dump_filter
=
428 (simple_strtoul(s
, NULL
, 0) << MMF_DUMP_FILTER_SHIFT
) &
429 MMF_DUMP_FILTER_MASK
;
433 __setup("coredump_filter=", coredump_filter_setup
);
435 #include <linux/init_task.h>
437 static void mm_init_aio(struct mm_struct
*mm
)
440 spin_lock_init(&mm
->ioctx_lock
);
441 INIT_HLIST_HEAD(&mm
->ioctx_list
);
445 static struct mm_struct
* mm_init(struct mm_struct
* mm
, struct task_struct
*p
)
447 atomic_set(&mm
->mm_users
, 1);
448 atomic_set(&mm
->mm_count
, 1);
449 init_rwsem(&mm
->mmap_sem
);
450 INIT_LIST_HEAD(&mm
->mmlist
);
451 mm
->flags
= (current
->mm
) ?
452 (current
->mm
->flags
& MMF_INIT_MASK
) : default_dump_filter
;
453 mm
->core_state
= NULL
;
455 set_mm_counter(mm
, file_rss
, 0);
456 set_mm_counter(mm
, anon_rss
, 0);
457 spin_lock_init(&mm
->page_table_lock
);
458 mm
->free_area_cache
= TASK_UNMAPPED_BASE
;
459 mm
->cached_hole_size
= ~0UL;
461 mm_init_owner(mm
, p
);
463 if (likely(!mm_alloc_pgd(mm
))) {
465 mmu_notifier_mm_init(mm
);
474 * Allocate and initialize an mm_struct.
476 struct mm_struct
* mm_alloc(void)
478 struct mm_struct
* mm
;
482 memset(mm
, 0, sizeof(*mm
));
483 mm
= mm_init(mm
, current
);
489 * Called when the last reference to the mm
490 * is dropped: either by a lazy thread or by
491 * mmput. Free the page directory and the mm.
493 void __mmdrop(struct mm_struct
*mm
)
495 BUG_ON(mm
== &init_mm
);
498 mmu_notifier_mm_destroy(mm
);
501 EXPORT_SYMBOL_GPL(__mmdrop
);
504 * Decrement the use count and release all resources for an mm.
506 void mmput(struct mm_struct
*mm
)
510 if (atomic_dec_and_test(&mm
->mm_users
)) {
514 set_mm_exe_file(mm
, NULL
);
515 if (!list_empty(&mm
->mmlist
)) {
516 spin_lock(&mmlist_lock
);
517 list_del(&mm
->mmlist
);
518 spin_unlock(&mmlist_lock
);
522 module_put(mm
->binfmt
->module
);
526 EXPORT_SYMBOL_GPL(mmput
);
529 * get_task_mm - acquire a reference to the task's mm
531 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
532 * this kernel workthread has transiently adopted a user mm with use_mm,
533 * to do its AIO) is not set and if so returns a reference to it, after
534 * bumping up the use count. User must release the mm via mmput()
535 * after use. Typically used by /proc and ptrace.
537 struct mm_struct
*get_task_mm(struct task_struct
*task
)
539 struct mm_struct
*mm
;
544 if (task
->flags
& PF_KTHREAD
)
547 atomic_inc(&mm
->mm_users
);
552 EXPORT_SYMBOL_GPL(get_task_mm
);
554 /* Please note the differences between mmput and mm_release.
555 * mmput is called whenever we stop holding onto a mm_struct,
556 * error success whatever.
558 * mm_release is called after a mm_struct has been removed
559 * from the current process.
561 * This difference is important for error handling, when we
562 * only half set up a mm_struct for a new process and need to restore
563 * the old one. Because we mmput the new mm_struct before
564 * restoring the old one. . .
565 * Eric Biederman 10 January 1998
567 void mm_release(struct task_struct
*tsk
, struct mm_struct
*mm
)
569 struct completion
*vfork_done
= tsk
->vfork_done
;
571 /* Get rid of any futexes when releasing the mm */
573 if (unlikely(tsk
->robust_list
)) {
574 exit_robust_list(tsk
);
575 tsk
->robust_list
= NULL
;
578 if (unlikely(tsk
->compat_robust_list
)) {
579 compat_exit_robust_list(tsk
);
580 tsk
->compat_robust_list
= NULL
;
583 if (unlikely(!list_empty(&tsk
->pi_state_list
)))
584 exit_pi_state_list(tsk
);
587 /* Get rid of any cached register state */
588 deactivate_mm(tsk
, mm
);
590 /* notify parent sleeping on vfork() */
592 tsk
->vfork_done
= NULL
;
593 complete(vfork_done
);
597 * If we're exiting normally, clear a user-space tid field if
598 * requested. We leave this alone when dying by signal, to leave
599 * the value intact in a core dump, and to save the unnecessary
600 * trouble otherwise. Userland only wants this done for a sys_exit.
602 if (tsk
->clear_child_tid
) {
603 if (!(tsk
->flags
& PF_SIGNALED
) &&
604 atomic_read(&mm
->mm_users
) > 1) {
606 * We don't check the error code - if userspace has
607 * not set up a proper pointer then tough luck.
609 put_user(0, tsk
->clear_child_tid
);
610 sys_futex(tsk
->clear_child_tid
, FUTEX_WAKE
,
613 tsk
->clear_child_tid
= NULL
;
618 * Allocate a new mm structure and copy contents from the
619 * mm structure of the passed in task structure.
621 struct mm_struct
*dup_mm(struct task_struct
*tsk
)
623 struct mm_struct
*mm
, *oldmm
= current
->mm
;
633 memcpy(mm
, oldmm
, sizeof(*mm
));
635 /* Initializing for Swap token stuff */
636 mm
->token_priority
= 0;
637 mm
->last_interval
= 0;
639 if (!mm_init(mm
, tsk
))
642 if (init_new_context(tsk
, mm
))
645 dup_mm_exe_file(oldmm
, mm
);
647 err
= dup_mmap(mm
, oldmm
);
651 mm
->hiwater_rss
= get_mm_rss(mm
);
652 mm
->hiwater_vm
= mm
->total_vm
;
654 if (mm
->binfmt
&& !try_module_get(mm
->binfmt
->module
))
660 /* don't put binfmt in mmput, we haven't got module yet */
669 * If init_new_context() failed, we cannot use mmput() to free the mm
670 * because it calls destroy_context()
677 static int copy_mm(unsigned long clone_flags
, struct task_struct
* tsk
)
679 struct mm_struct
* mm
, *oldmm
;
682 tsk
->min_flt
= tsk
->maj_flt
= 0;
683 tsk
->nvcsw
= tsk
->nivcsw
= 0;
684 #ifdef CONFIG_DETECT_HUNG_TASK
685 tsk
->last_switch_count
= tsk
->nvcsw
+ tsk
->nivcsw
;
689 tsk
->active_mm
= NULL
;
692 * Are we cloning a kernel thread?
694 * We need to steal a active VM for that..
700 if (clone_flags
& CLONE_VM
) {
701 atomic_inc(&oldmm
->mm_users
);
712 /* Initializing for Swap token stuff */
713 mm
->token_priority
= 0;
714 mm
->last_interval
= 0;
724 static int copy_fs(unsigned long clone_flags
, struct task_struct
*tsk
)
726 struct fs_struct
*fs
= current
->fs
;
727 if (clone_flags
& CLONE_FS
) {
728 /* tsk->fs is already what we want */
729 write_lock(&fs
->lock
);
731 write_unlock(&fs
->lock
);
735 write_unlock(&fs
->lock
);
738 tsk
->fs
= copy_fs_struct(fs
);
744 static int copy_files(unsigned long clone_flags
, struct task_struct
* tsk
)
746 struct files_struct
*oldf
, *newf
;
750 * A background process may not have any files ...
752 oldf
= current
->files
;
756 if (clone_flags
& CLONE_FILES
) {
757 atomic_inc(&oldf
->count
);
761 newf
= dup_fd(oldf
, &error
);
771 static int copy_io(unsigned long clone_flags
, struct task_struct
*tsk
)
774 struct io_context
*ioc
= current
->io_context
;
779 * Share io context with parent, if CLONE_IO is set
781 if (clone_flags
& CLONE_IO
) {
782 tsk
->io_context
= ioc_task_link(ioc
);
783 if (unlikely(!tsk
->io_context
))
785 } else if (ioprio_valid(ioc
->ioprio
)) {
786 tsk
->io_context
= alloc_io_context(GFP_KERNEL
, -1);
787 if (unlikely(!tsk
->io_context
))
790 tsk
->io_context
->ioprio
= ioc
->ioprio
;
796 static int copy_sighand(unsigned long clone_flags
, struct task_struct
*tsk
)
798 struct sighand_struct
*sig
;
800 if (clone_flags
& CLONE_SIGHAND
) {
801 atomic_inc(¤t
->sighand
->count
);
804 sig
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
805 rcu_assign_pointer(tsk
->sighand
, sig
);
808 atomic_set(&sig
->count
, 1);
809 memcpy(sig
->action
, current
->sighand
->action
, sizeof(sig
->action
));
813 void __cleanup_sighand(struct sighand_struct
*sighand
)
815 if (atomic_dec_and_test(&sighand
->count
))
816 kmem_cache_free(sighand_cachep
, sighand
);
821 * Initialize POSIX timer handling for a thread group.
823 static void posix_cpu_timers_init_group(struct signal_struct
*sig
)
825 /* Thread group counters. */
826 thread_group_cputime_init(sig
);
828 /* Expiration times and increments. */
829 sig
->it
[CPUCLOCK_PROF
].expires
= cputime_zero
;
830 sig
->it
[CPUCLOCK_PROF
].incr
= cputime_zero
;
831 sig
->it
[CPUCLOCK_VIRT
].expires
= cputime_zero
;
832 sig
->it
[CPUCLOCK_VIRT
].incr
= cputime_zero
;
834 /* Cached expiration times. */
835 sig
->cputime_expires
.prof_exp
= cputime_zero
;
836 sig
->cputime_expires
.virt_exp
= cputime_zero
;
837 sig
->cputime_expires
.sched_exp
= 0;
839 if (sig
->rlim
[RLIMIT_CPU
].rlim_cur
!= RLIM_INFINITY
) {
840 sig
->cputime_expires
.prof_exp
=
841 secs_to_cputime(sig
->rlim
[RLIMIT_CPU
].rlim_cur
);
842 sig
->cputimer
.running
= 1;
845 /* The timer lists. */
846 INIT_LIST_HEAD(&sig
->cpu_timers
[0]);
847 INIT_LIST_HEAD(&sig
->cpu_timers
[1]);
848 INIT_LIST_HEAD(&sig
->cpu_timers
[2]);
851 static int copy_signal(unsigned long clone_flags
, struct task_struct
*tsk
)
853 struct signal_struct
*sig
;
855 if (clone_flags
& CLONE_THREAD
)
858 sig
= kmem_cache_alloc(signal_cachep
, GFP_KERNEL
);
863 atomic_set(&sig
->count
, 1);
864 atomic_set(&sig
->live
, 1);
865 init_waitqueue_head(&sig
->wait_chldexit
);
867 if (clone_flags
& CLONE_NEWPID
)
868 sig
->flags
|= SIGNAL_UNKILLABLE
;
869 sig
->group_exit_code
= 0;
870 sig
->group_exit_task
= NULL
;
871 sig
->group_stop_count
= 0;
872 sig
->curr_target
= tsk
;
873 init_sigpending(&sig
->shared_pending
);
874 INIT_LIST_HEAD(&sig
->posix_timers
);
876 hrtimer_init(&sig
->real_timer
, CLOCK_MONOTONIC
, HRTIMER_MODE_REL
);
877 sig
->it_real_incr
.tv64
= 0;
878 sig
->real_timer
.function
= it_real_fn
;
880 sig
->leader
= 0; /* session leadership doesn't inherit */
881 sig
->tty_old_pgrp
= NULL
;
884 sig
->utime
= sig
->stime
= sig
->cutime
= sig
->cstime
= cputime_zero
;
885 sig
->gtime
= cputime_zero
;
886 sig
->cgtime
= cputime_zero
;
887 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
888 sig
->prev_utime
= sig
->prev_stime
= cputime_zero
;
890 sig
->nvcsw
= sig
->nivcsw
= sig
->cnvcsw
= sig
->cnivcsw
= 0;
891 sig
->min_flt
= sig
->maj_flt
= sig
->cmin_flt
= sig
->cmaj_flt
= 0;
892 sig
->inblock
= sig
->oublock
= sig
->cinblock
= sig
->coublock
= 0;
893 sig
->maxrss
= sig
->cmaxrss
= 0;
894 task_io_accounting_init(&sig
->ioac
);
895 sig
->sum_sched_runtime
= 0;
896 taskstats_tgid_init(sig
);
898 task_lock(current
->group_leader
);
899 memcpy(sig
->rlim
, current
->signal
->rlim
, sizeof sig
->rlim
);
900 task_unlock(current
->group_leader
);
902 posix_cpu_timers_init_group(sig
);
904 acct_init_pacct(&sig
->pacct
);
908 sig
->oom_adj
= current
->signal
->oom_adj
;
913 void __cleanup_signal(struct signal_struct
*sig
)
915 thread_group_cputime_free(sig
);
916 tty_kref_put(sig
->tty
);
917 kmem_cache_free(signal_cachep
, sig
);
920 static void copy_flags(unsigned long clone_flags
, struct task_struct
*p
)
922 unsigned long new_flags
= p
->flags
;
924 new_flags
&= ~PF_SUPERPRIV
;
925 new_flags
|= PF_FORKNOEXEC
;
926 new_flags
|= PF_STARTING
;
927 p
->flags
= new_flags
;
928 clear_freeze_flag(p
);
931 SYSCALL_DEFINE1(set_tid_address
, int __user
*, tidptr
)
933 current
->clear_child_tid
= tidptr
;
935 return task_pid_vnr(current
);
938 static void rt_mutex_init_task(struct task_struct
*p
)
940 spin_lock_init(&p
->pi_lock
);
941 #ifdef CONFIG_RT_MUTEXES
942 plist_head_init(&p
->pi_waiters
, &p
->pi_lock
);
943 p
->pi_blocked_on
= NULL
;
947 #ifdef CONFIG_MM_OWNER
948 void mm_init_owner(struct mm_struct
*mm
, struct task_struct
*p
)
952 #endif /* CONFIG_MM_OWNER */
955 * Initialize POSIX timer handling for a single task.
957 static void posix_cpu_timers_init(struct task_struct
*tsk
)
959 tsk
->cputime_expires
.prof_exp
= cputime_zero
;
960 tsk
->cputime_expires
.virt_exp
= cputime_zero
;
961 tsk
->cputime_expires
.sched_exp
= 0;
962 INIT_LIST_HEAD(&tsk
->cpu_timers
[0]);
963 INIT_LIST_HEAD(&tsk
->cpu_timers
[1]);
964 INIT_LIST_HEAD(&tsk
->cpu_timers
[2]);
968 * This creates a new process as a copy of the old one,
969 * but does not actually start it yet.
971 * It copies the registers, and all the appropriate
972 * parts of the process environment (as per the clone
973 * flags). The actual kick-off is left to the caller.
975 static struct task_struct
*copy_process(unsigned long clone_flags
,
976 unsigned long stack_start
,
977 struct pt_regs
*regs
,
978 unsigned long stack_size
,
979 int __user
*child_tidptr
,
984 struct task_struct
*p
;
985 int cgroup_callbacks_done
= 0;
987 if ((clone_flags
& (CLONE_NEWNS
|CLONE_FS
)) == (CLONE_NEWNS
|CLONE_FS
))
988 return ERR_PTR(-EINVAL
);
991 * Thread groups must share signals as well, and detached threads
992 * can only be started up within the thread group.
994 if ((clone_flags
& CLONE_THREAD
) && !(clone_flags
& CLONE_SIGHAND
))
995 return ERR_PTR(-EINVAL
);
998 * Shared signal handlers imply shared VM. By way of the above,
999 * thread groups also imply shared VM. Blocking this case allows
1000 * for various simplifications in other code.
1002 if ((clone_flags
& CLONE_SIGHAND
) && !(clone_flags
& CLONE_VM
))
1003 return ERR_PTR(-EINVAL
);
1006 * Siblings of global init remain as zombies on exit since they are
1007 * not reaped by their parent (swapper). To solve this and to avoid
1008 * multi-rooted process trees, prevent global and container-inits
1009 * from creating siblings.
1011 if ((clone_flags
& CLONE_PARENT
) &&
1012 current
->signal
->flags
& SIGNAL_UNKILLABLE
)
1013 return ERR_PTR(-EINVAL
);
1015 retval
= security_task_create(clone_flags
);
1020 p
= dup_task_struct(current
);
1024 ftrace_graph_init_task(p
);
1026 rt_mutex_init_task(p
);
1028 #ifdef CONFIG_PROVE_LOCKING
1029 DEBUG_LOCKS_WARN_ON(!p
->hardirqs_enabled
);
1030 DEBUG_LOCKS_WARN_ON(!p
->softirqs_enabled
);
1033 if (atomic_read(&p
->real_cred
->user
->processes
) >=
1034 p
->signal
->rlim
[RLIMIT_NPROC
].rlim_cur
) {
1035 if (!capable(CAP_SYS_ADMIN
) && !capable(CAP_SYS_RESOURCE
) &&
1036 p
->real_cred
->user
!= INIT_USER
)
1040 retval
= copy_creds(p
, clone_flags
);
1045 * If multiple threads are within copy_process(), then this check
1046 * triggers too late. This doesn't hurt, the check is only there
1047 * to stop root fork bombs.
1050 if (nr_threads
>= max_threads
)
1051 goto bad_fork_cleanup_count
;
1053 if (!try_module_get(task_thread_info(p
)->exec_domain
->module
))
1054 goto bad_fork_cleanup_count
;
1057 delayacct_tsk_init(p
); /* Must remain after dup_task_struct() */
1058 copy_flags(clone_flags
, p
);
1059 INIT_LIST_HEAD(&p
->children
);
1060 INIT_LIST_HEAD(&p
->sibling
);
1061 rcu_copy_process(p
);
1062 p
->vfork_done
= NULL
;
1063 spin_lock_init(&p
->alloc_lock
);
1065 init_sigpending(&p
->pending
);
1067 p
->utime
= cputime_zero
;
1068 p
->stime
= cputime_zero
;
1069 p
->gtime
= cputime_zero
;
1070 p
->utimescaled
= cputime_zero
;
1071 p
->stimescaled
= cputime_zero
;
1072 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
1073 p
->prev_utime
= cputime_zero
;
1074 p
->prev_stime
= cputime_zero
;
1077 p
->default_timer_slack_ns
= current
->timer_slack_ns
;
1079 task_io_accounting_init(&p
->ioac
);
1080 acct_clear_integrals(p
);
1082 posix_cpu_timers_init(p
);
1084 p
->lock_depth
= -1; /* -1 = no lock */
1085 do_posix_clock_monotonic_gettime(&p
->start_time
);
1086 p
->real_start_time
= p
->start_time
;
1087 monotonic_to_bootbased(&p
->real_start_time
);
1088 p
->io_context
= NULL
;
1089 p
->audit_context
= NULL
;
1092 p
->mempolicy
= mpol_dup(p
->mempolicy
);
1093 if (IS_ERR(p
->mempolicy
)) {
1094 retval
= PTR_ERR(p
->mempolicy
);
1095 p
->mempolicy
= NULL
;
1096 goto bad_fork_cleanup_cgroup
;
1098 mpol_fix_fork_child_flag(p
);
1100 #ifdef CONFIG_TRACE_IRQFLAGS
1102 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1103 p
->hardirqs_enabled
= 1;
1105 p
->hardirqs_enabled
= 0;
1107 p
->hardirq_enable_ip
= 0;
1108 p
->hardirq_enable_event
= 0;
1109 p
->hardirq_disable_ip
= _THIS_IP_
;
1110 p
->hardirq_disable_event
= 0;
1111 p
->softirqs_enabled
= 1;
1112 p
->softirq_enable_ip
= _THIS_IP_
;
1113 p
->softirq_enable_event
= 0;
1114 p
->softirq_disable_ip
= 0;
1115 p
->softirq_disable_event
= 0;
1116 p
->hardirq_context
= 0;
1117 p
->softirq_context
= 0;
1119 #ifdef CONFIG_LOCKDEP
1120 p
->lockdep_depth
= 0; /* no locks held yet */
1121 p
->curr_chain_key
= 0;
1122 p
->lockdep_recursion
= 0;
1125 #ifdef CONFIG_DEBUG_MUTEXES
1126 p
->blocked_on
= NULL
; /* not blocked yet */
1131 p
->stack_start
= stack_start
;
1133 /* Perform scheduler related setup. Assign this task to a CPU. */
1134 sched_fork(p
, clone_flags
);
1136 retval
= perf_event_init_task(p
);
1138 goto bad_fork_cleanup_policy
;
1140 if ((retval
= audit_alloc(p
)))
1141 goto bad_fork_cleanup_policy
;
1142 /* copy all the process information */
1143 if ((retval
= copy_semundo(clone_flags
, p
)))
1144 goto bad_fork_cleanup_audit
;
1145 if ((retval
= copy_files(clone_flags
, p
)))
1146 goto bad_fork_cleanup_semundo
;
1147 if ((retval
= copy_fs(clone_flags
, p
)))
1148 goto bad_fork_cleanup_files
;
1149 if ((retval
= copy_sighand(clone_flags
, p
)))
1150 goto bad_fork_cleanup_fs
;
1151 if ((retval
= copy_signal(clone_flags
, p
)))
1152 goto bad_fork_cleanup_sighand
;
1153 if ((retval
= copy_mm(clone_flags
, p
)))
1154 goto bad_fork_cleanup_signal
;
1155 if ((retval
= copy_namespaces(clone_flags
, p
)))
1156 goto bad_fork_cleanup_mm
;
1157 if ((retval
= copy_io(clone_flags
, p
)))
1158 goto bad_fork_cleanup_namespaces
;
1159 retval
= copy_thread(clone_flags
, stack_start
, stack_size
, p
, regs
);
1161 goto bad_fork_cleanup_io
;
1163 if (pid
!= &init_struct_pid
) {
1165 pid
= alloc_pid(p
->nsproxy
->pid_ns
);
1167 goto bad_fork_cleanup_io
;
1169 if (clone_flags
& CLONE_NEWPID
) {
1170 retval
= pid_ns_prepare_proc(p
->nsproxy
->pid_ns
);
1172 goto bad_fork_free_pid
;
1176 p
->pid
= pid_nr(pid
);
1178 if (clone_flags
& CLONE_THREAD
)
1179 p
->tgid
= current
->tgid
;
1181 if (current
->nsproxy
!= p
->nsproxy
) {
1182 retval
= ns_cgroup_clone(p
, pid
);
1184 goto bad_fork_free_pid
;
1187 p
->set_child_tid
= (clone_flags
& CLONE_CHILD_SETTID
) ? child_tidptr
: NULL
;
1189 * Clear TID on mm_release()?
1191 p
->clear_child_tid
= (clone_flags
& CLONE_CHILD_CLEARTID
) ? child_tidptr
: NULL
;
1193 p
->robust_list
= NULL
;
1194 #ifdef CONFIG_COMPAT
1195 p
->compat_robust_list
= NULL
;
1197 INIT_LIST_HEAD(&p
->pi_state_list
);
1198 p
->pi_state_cache
= NULL
;
1201 * sigaltstack should be cleared when sharing the same VM
1203 if ((clone_flags
& (CLONE_VM
|CLONE_VFORK
)) == CLONE_VM
)
1204 p
->sas_ss_sp
= p
->sas_ss_size
= 0;
1207 * Syscall tracing should be turned off in the child regardless
1210 clear_tsk_thread_flag(p
, TIF_SYSCALL_TRACE
);
1211 #ifdef TIF_SYSCALL_EMU
1212 clear_tsk_thread_flag(p
, TIF_SYSCALL_EMU
);
1214 clear_all_latency_tracing(p
);
1216 /* ok, now we should be set up.. */
1217 p
->exit_signal
= (clone_flags
& CLONE_THREAD
) ? -1 : (clone_flags
& CSIGNAL
);
1218 p
->pdeath_signal
= 0;
1222 * Ok, make it visible to the rest of the system.
1223 * We dont wake it up yet.
1225 p
->group_leader
= p
;
1226 INIT_LIST_HEAD(&p
->thread_group
);
1228 /* Now that the task is set up, run cgroup callbacks if
1229 * necessary. We need to run them before the task is visible
1230 * on the tasklist. */
1231 cgroup_fork_callbacks(p
);
1232 cgroup_callbacks_done
= 1;
1234 /* Need tasklist lock for parent etc handling! */
1235 write_lock_irq(&tasklist_lock
);
1238 * The task hasn't been attached yet, so its cpus_allowed mask will
1239 * not be changed, nor will its assigned CPU.
1241 * The cpus_allowed mask of the parent may have changed after it was
1242 * copied first time - so re-copy it here, then check the child's CPU
1243 * to ensure it is on a valid CPU (and if not, just force it back to
1244 * parent's CPU). This avoids alot of nasty races.
1246 p
->cpus_allowed
= current
->cpus_allowed
;
1247 p
->rt
.nr_cpus_allowed
= current
->rt
.nr_cpus_allowed
;
1248 if (unlikely(!cpu_isset(task_cpu(p
), p
->cpus_allowed
) ||
1249 !cpu_online(task_cpu(p
))))
1250 set_task_cpu(p
, smp_processor_id());
1252 /* CLONE_PARENT re-uses the old parent */
1253 if (clone_flags
& (CLONE_PARENT
|CLONE_THREAD
)) {
1254 p
->real_parent
= current
->real_parent
;
1255 p
->parent_exec_id
= current
->parent_exec_id
;
1257 p
->real_parent
= current
;
1258 p
->parent_exec_id
= current
->self_exec_id
;
1261 spin_lock(¤t
->sighand
->siglock
);
1264 * Process group and session signals need to be delivered to just the
1265 * parent before the fork or both the parent and the child after the
1266 * fork. Restart if a signal comes in before we add the new process to
1267 * it's process group.
1268 * A fatal signal pending means that current will exit, so the new
1269 * thread can't slip out of an OOM kill (or normal SIGKILL).
1271 recalc_sigpending();
1272 if (signal_pending(current
)) {
1273 spin_unlock(¤t
->sighand
->siglock
);
1274 write_unlock_irq(&tasklist_lock
);
1275 retval
= -ERESTARTNOINTR
;
1276 goto bad_fork_free_pid
;
1279 if (clone_flags
& CLONE_THREAD
) {
1280 atomic_inc(¤t
->signal
->count
);
1281 atomic_inc(¤t
->signal
->live
);
1282 p
->group_leader
= current
->group_leader
;
1283 list_add_tail_rcu(&p
->thread_group
, &p
->group_leader
->thread_group
);
1286 if (likely(p
->pid
)) {
1287 list_add_tail(&p
->sibling
, &p
->real_parent
->children
);
1288 tracehook_finish_clone(p
, clone_flags
, trace
);
1290 if (thread_group_leader(p
)) {
1291 if (clone_flags
& CLONE_NEWPID
)
1292 p
->nsproxy
->pid_ns
->child_reaper
= p
;
1294 p
->signal
->leader_pid
= pid
;
1295 tty_kref_put(p
->signal
->tty
);
1296 p
->signal
->tty
= tty_kref_get(current
->signal
->tty
);
1297 attach_pid(p
, PIDTYPE_PGID
, task_pgrp(current
));
1298 attach_pid(p
, PIDTYPE_SID
, task_session(current
));
1299 list_add_tail_rcu(&p
->tasks
, &init_task
.tasks
);
1300 __get_cpu_var(process_counts
)++;
1302 attach_pid(p
, PIDTYPE_PID
, pid
);
1307 spin_unlock(¤t
->sighand
->siglock
);
1308 write_unlock_irq(&tasklist_lock
);
1309 proc_fork_connector(p
);
1310 cgroup_post_fork(p
);
1315 if (pid
!= &init_struct_pid
)
1317 bad_fork_cleanup_io
:
1318 put_io_context(p
->io_context
);
1319 bad_fork_cleanup_namespaces
:
1320 exit_task_namespaces(p
);
1321 bad_fork_cleanup_mm
:
1324 bad_fork_cleanup_signal
:
1325 if (!(clone_flags
& CLONE_THREAD
))
1326 __cleanup_signal(p
->signal
);
1327 bad_fork_cleanup_sighand
:
1328 __cleanup_sighand(p
->sighand
);
1329 bad_fork_cleanup_fs
:
1330 exit_fs(p
); /* blocking */
1331 bad_fork_cleanup_files
:
1332 exit_files(p
); /* blocking */
1333 bad_fork_cleanup_semundo
:
1335 bad_fork_cleanup_audit
:
1337 bad_fork_cleanup_policy
:
1338 perf_event_free_task(p
);
1340 mpol_put(p
->mempolicy
);
1341 bad_fork_cleanup_cgroup
:
1343 cgroup_exit(p
, cgroup_callbacks_done
);
1344 delayacct_tsk_free(p
);
1345 module_put(task_thread_info(p
)->exec_domain
->module
);
1346 bad_fork_cleanup_count
:
1347 atomic_dec(&p
->cred
->user
->processes
);
1352 return ERR_PTR(retval
);
1355 noinline
struct pt_regs
* __cpuinit
__attribute__((weak
)) idle_regs(struct pt_regs
*regs
)
1357 memset(regs
, 0, sizeof(struct pt_regs
));
1361 struct task_struct
* __cpuinit
fork_idle(int cpu
)
1363 struct task_struct
*task
;
1364 struct pt_regs regs
;
1366 task
= copy_process(CLONE_VM
, 0, idle_regs(®s
), 0, NULL
,
1367 &init_struct_pid
, 0);
1369 init_idle(task
, cpu
);
1375 * Ok, this is the main fork-routine.
1377 * It copies the process, and if successful kick-starts
1378 * it and waits for it to finish using the VM if required.
1380 long do_fork(unsigned long clone_flags
,
1381 unsigned long stack_start
,
1382 struct pt_regs
*regs
,
1383 unsigned long stack_size
,
1384 int __user
*parent_tidptr
,
1385 int __user
*child_tidptr
)
1387 struct task_struct
*p
;
1392 * Do some preliminary argument and permissions checking before we
1393 * actually start allocating stuff
1395 if (clone_flags
& CLONE_NEWUSER
) {
1396 if (clone_flags
& CLONE_THREAD
)
1398 /* hopefully this check will go away when userns support is
1401 if (!capable(CAP_SYS_ADMIN
) || !capable(CAP_SETUID
) ||
1402 !capable(CAP_SETGID
))
1407 * We hope to recycle these flags after 2.6.26
1409 if (unlikely(clone_flags
& CLONE_STOPPED
)) {
1410 static int __read_mostly count
= 100;
1412 if (count
> 0 && printk_ratelimit()) {
1413 char comm
[TASK_COMM_LEN
];
1416 printk(KERN_INFO
"fork(): process `%s' used deprecated "
1417 "clone flags 0x%lx\n",
1418 get_task_comm(comm
, current
),
1419 clone_flags
& CLONE_STOPPED
);
1424 * When called from kernel_thread, don't do user tracing stuff.
1426 if (likely(user_mode(regs
)))
1427 trace
= tracehook_prepare_clone(clone_flags
);
1429 p
= copy_process(clone_flags
, stack_start
, regs
, stack_size
,
1430 child_tidptr
, NULL
, trace
);
1432 * Do this prior waking up the new thread - the thread pointer
1433 * might get invalid after that point, if the thread exits quickly.
1436 struct completion vfork
;
1438 trace_sched_process_fork(current
, p
);
1440 nr
= task_pid_vnr(p
);
1442 if (clone_flags
& CLONE_PARENT_SETTID
)
1443 put_user(nr
, parent_tidptr
);
1445 if (clone_flags
& CLONE_VFORK
) {
1446 p
->vfork_done
= &vfork
;
1447 init_completion(&vfork
);
1450 audit_finish_fork(p
);
1451 tracehook_report_clone(regs
, clone_flags
, nr
, p
);
1454 * We set PF_STARTING at creation in case tracing wants to
1455 * use this to distinguish a fully live task from one that
1456 * hasn't gotten to tracehook_report_clone() yet. Now we
1457 * clear it and set the child going.
1459 p
->flags
&= ~PF_STARTING
;
1461 if (unlikely(clone_flags
& CLONE_STOPPED
)) {
1463 * We'll start up with an immediate SIGSTOP.
1465 sigaddset(&p
->pending
.signal
, SIGSTOP
);
1466 set_tsk_thread_flag(p
, TIF_SIGPENDING
);
1467 __set_task_state(p
, TASK_STOPPED
);
1469 wake_up_new_task(p
, clone_flags
);
1472 tracehook_report_clone_complete(trace
, regs
,
1473 clone_flags
, nr
, p
);
1475 if (clone_flags
& CLONE_VFORK
) {
1476 freezer_do_not_count();
1477 wait_for_completion(&vfork
);
1479 tracehook_report_vfork_done(p
, nr
);
1487 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1488 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1491 static void sighand_ctor(void *data
)
1493 struct sighand_struct
*sighand
= data
;
1495 spin_lock_init(&sighand
->siglock
);
1496 init_waitqueue_head(&sighand
->signalfd_wqh
);
1499 void __init
proc_caches_init(void)
1501 sighand_cachep
= kmem_cache_create("sighand_cache",
1502 sizeof(struct sighand_struct
), 0,
1503 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_DESTROY_BY_RCU
|
1504 SLAB_NOTRACK
, sighand_ctor
);
1505 signal_cachep
= kmem_cache_create("signal_cache",
1506 sizeof(struct signal_struct
), 0,
1507 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1508 files_cachep
= kmem_cache_create("files_cache",
1509 sizeof(struct files_struct
), 0,
1510 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1511 fs_cachep
= kmem_cache_create("fs_cache",
1512 sizeof(struct fs_struct
), 0,
1513 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1514 mm_cachep
= kmem_cache_create("mm_struct",
1515 sizeof(struct mm_struct
), ARCH_MIN_MMSTRUCT_ALIGN
,
1516 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1517 vm_area_cachep
= KMEM_CACHE(vm_area_struct
, SLAB_PANIC
);
1522 * Check constraints on flags passed to the unshare system call and
1523 * force unsharing of additional process context as appropriate.
1525 static void check_unshare_flags(unsigned long *flags_ptr
)
1528 * If unsharing a thread from a thread group, must also
1531 if (*flags_ptr
& CLONE_THREAD
)
1532 *flags_ptr
|= CLONE_VM
;
1535 * If unsharing vm, must also unshare signal handlers.
1537 if (*flags_ptr
& CLONE_VM
)
1538 *flags_ptr
|= CLONE_SIGHAND
;
1541 * If unsharing signal handlers and the task was created
1542 * using CLONE_THREAD, then must unshare the thread
1544 if ((*flags_ptr
& CLONE_SIGHAND
) &&
1545 (atomic_read(¤t
->signal
->count
) > 1))
1546 *flags_ptr
|= CLONE_THREAD
;
1549 * If unsharing namespace, must also unshare filesystem information.
1551 if (*flags_ptr
& CLONE_NEWNS
)
1552 *flags_ptr
|= CLONE_FS
;
1556 * Unsharing of tasks created with CLONE_THREAD is not supported yet
1558 static int unshare_thread(unsigned long unshare_flags
)
1560 if (unshare_flags
& CLONE_THREAD
)
1567 * Unshare the filesystem structure if it is being shared
1569 static int unshare_fs(unsigned long unshare_flags
, struct fs_struct
**new_fsp
)
1571 struct fs_struct
*fs
= current
->fs
;
1573 if (!(unshare_flags
& CLONE_FS
) || !fs
)
1576 /* don't need lock here; in the worst case we'll do useless copy */
1580 *new_fsp
= copy_fs_struct(fs
);
1588 * Unsharing of sighand is not supported yet
1590 static int unshare_sighand(unsigned long unshare_flags
, struct sighand_struct
**new_sighp
)
1592 struct sighand_struct
*sigh
= current
->sighand
;
1594 if ((unshare_flags
& CLONE_SIGHAND
) && atomic_read(&sigh
->count
) > 1)
1601 * Unshare vm if it is being shared
1603 static int unshare_vm(unsigned long unshare_flags
, struct mm_struct
**new_mmp
)
1605 struct mm_struct
*mm
= current
->mm
;
1607 if ((unshare_flags
& CLONE_VM
) &&
1608 (mm
&& atomic_read(&mm
->mm_users
) > 1)) {
1616 * Unshare file descriptor table if it is being shared
1618 static int unshare_fd(unsigned long unshare_flags
, struct files_struct
**new_fdp
)
1620 struct files_struct
*fd
= current
->files
;
1623 if ((unshare_flags
& CLONE_FILES
) &&
1624 (fd
&& atomic_read(&fd
->count
) > 1)) {
1625 *new_fdp
= dup_fd(fd
, &error
);
1634 * unshare allows a process to 'unshare' part of the process
1635 * context which was originally shared using clone. copy_*
1636 * functions used by do_fork() cannot be used here directly
1637 * because they modify an inactive task_struct that is being
1638 * constructed. Here we are modifying the current, active,
1641 SYSCALL_DEFINE1(unshare
, unsigned long, unshare_flags
)
1644 struct fs_struct
*fs
, *new_fs
= NULL
;
1645 struct sighand_struct
*new_sigh
= NULL
;
1646 struct mm_struct
*mm
, *new_mm
= NULL
, *active_mm
= NULL
;
1647 struct files_struct
*fd
, *new_fd
= NULL
;
1648 struct nsproxy
*new_nsproxy
= NULL
;
1651 check_unshare_flags(&unshare_flags
);
1653 /* Return -EINVAL for all unsupported flags */
1655 if (unshare_flags
& ~(CLONE_THREAD
|CLONE_FS
|CLONE_NEWNS
|CLONE_SIGHAND
|
1656 CLONE_VM
|CLONE_FILES
|CLONE_SYSVSEM
|
1657 CLONE_NEWUTS
|CLONE_NEWIPC
|CLONE_NEWNET
))
1658 goto bad_unshare_out
;
1661 * CLONE_NEWIPC must also detach from the undolist: after switching
1662 * to a new ipc namespace, the semaphore arrays from the old
1663 * namespace are unreachable.
1665 if (unshare_flags
& (CLONE_NEWIPC
|CLONE_SYSVSEM
))
1667 if ((err
= unshare_thread(unshare_flags
)))
1668 goto bad_unshare_out
;
1669 if ((err
= unshare_fs(unshare_flags
, &new_fs
)))
1670 goto bad_unshare_cleanup_thread
;
1671 if ((err
= unshare_sighand(unshare_flags
, &new_sigh
)))
1672 goto bad_unshare_cleanup_fs
;
1673 if ((err
= unshare_vm(unshare_flags
, &new_mm
)))
1674 goto bad_unshare_cleanup_sigh
;
1675 if ((err
= unshare_fd(unshare_flags
, &new_fd
)))
1676 goto bad_unshare_cleanup_vm
;
1677 if ((err
= unshare_nsproxy_namespaces(unshare_flags
, &new_nsproxy
,
1679 goto bad_unshare_cleanup_fd
;
1681 if (new_fs
|| new_mm
|| new_fd
|| do_sysvsem
|| new_nsproxy
) {
1684 * CLONE_SYSVSEM is equivalent to sys_exit().
1690 switch_task_namespaces(current
, new_nsproxy
);
1698 write_lock(&fs
->lock
);
1699 current
->fs
= new_fs
;
1704 write_unlock(&fs
->lock
);
1709 active_mm
= current
->active_mm
;
1710 current
->mm
= new_mm
;
1711 current
->active_mm
= new_mm
;
1712 activate_mm(active_mm
, new_mm
);
1717 fd
= current
->files
;
1718 current
->files
= new_fd
;
1722 task_unlock(current
);
1726 put_nsproxy(new_nsproxy
);
1728 bad_unshare_cleanup_fd
:
1730 put_files_struct(new_fd
);
1732 bad_unshare_cleanup_vm
:
1736 bad_unshare_cleanup_sigh
:
1738 if (atomic_dec_and_test(&new_sigh
->count
))
1739 kmem_cache_free(sighand_cachep
, new_sigh
);
1741 bad_unshare_cleanup_fs
:
1743 free_fs_struct(new_fs
);
1745 bad_unshare_cleanup_thread
:
1751 * Helper to unshare the files of the current task.
1752 * We don't want to expose copy_files internals to
1753 * the exec layer of the kernel.
1756 int unshare_files(struct files_struct
**displaced
)
1758 struct task_struct
*task
= current
;
1759 struct files_struct
*copy
= NULL
;
1762 error
= unshare_fd(CLONE_FILES
, ©
);
1763 if (error
|| !copy
) {
1767 *displaced
= task
->files
;