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>
67 #include <linux/user-return-notifier.h>
69 #include <asm/pgtable.h>
70 #include <asm/pgalloc.h>
71 #include <asm/uaccess.h>
72 #include <asm/mmu_context.h>
73 #include <asm/cacheflush.h>
74 #include <asm/tlbflush.h>
76 #include <trace/events/sched.h>
79 * Protected counters by write_lock_irq(&tasklist_lock)
81 unsigned long total_forks
; /* Handle normal Linux uptimes. */
82 int nr_threads
; /* The idle threads do not count.. */
84 int max_threads
; /* tunable limit on nr_threads */
86 DEFINE_PER_CPU(unsigned long, process_counts
) = 0;
88 __cacheline_aligned
DEFINE_RWLOCK(tasklist_lock
); /* outer */
90 #ifdef CONFIG_PROVE_RCU
91 int lockdep_tasklist_lock_is_held(void)
93 return lockdep_is_held(&tasklist_lock
);
95 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held
);
96 #endif /* #ifdef CONFIG_PROVE_RCU */
98 int nr_processes(void)
103 for_each_possible_cpu(cpu
)
104 total
+= per_cpu(process_counts
, cpu
);
109 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
110 # define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL)
111 # define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk))
112 static struct kmem_cache
*task_struct_cachep
;
115 #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
116 static inline struct thread_info
*alloc_thread_info(struct task_struct
*tsk
)
118 #ifdef CONFIG_DEBUG_STACK_USAGE
119 gfp_t mask
= GFP_KERNEL
| __GFP_ZERO
;
121 gfp_t mask
= GFP_KERNEL
;
123 return (struct thread_info
*)__get_free_pages(mask
, THREAD_SIZE_ORDER
);
126 static inline void free_thread_info(struct thread_info
*ti
)
128 free_pages((unsigned long)ti
, THREAD_SIZE_ORDER
);
132 /* SLAB cache for signal_struct structures (tsk->signal) */
133 static struct kmem_cache
*signal_cachep
;
135 /* SLAB cache for sighand_struct structures (tsk->sighand) */
136 struct kmem_cache
*sighand_cachep
;
138 /* SLAB cache for files_struct structures (tsk->files) */
139 struct kmem_cache
*files_cachep
;
141 /* SLAB cache for fs_struct structures (tsk->fs) */
142 struct kmem_cache
*fs_cachep
;
144 /* SLAB cache for vm_area_struct structures */
145 struct kmem_cache
*vm_area_cachep
;
147 /* SLAB cache for mm_struct structures (tsk->mm) */
148 static struct kmem_cache
*mm_cachep
;
150 static void account_kernel_stack(struct thread_info
*ti
, int account
)
152 struct zone
*zone
= page_zone(virt_to_page(ti
));
154 mod_zone_page_state(zone
, NR_KERNEL_STACK
, account
);
157 void free_task(struct task_struct
*tsk
)
159 prop_local_destroy_single(&tsk
->dirties
);
160 account_kernel_stack(tsk
->stack
, -1);
161 free_thread_info(tsk
->stack
);
162 rt_mutex_debug_task_free(tsk
);
163 ftrace_graph_exit_task(tsk
);
164 free_task_struct(tsk
);
166 EXPORT_SYMBOL(free_task
);
168 void __put_task_struct(struct task_struct
*tsk
)
170 WARN_ON(!tsk
->exit_state
);
171 WARN_ON(atomic_read(&tsk
->usage
));
172 WARN_ON(tsk
== current
);
175 delayacct_tsk_free(tsk
);
177 if (!profile_handoff_task(tsk
))
182 * macro override instead of weak attribute alias, to workaround
183 * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
185 #ifndef arch_task_cache_init
186 #define arch_task_cache_init()
189 void __init
fork_init(unsigned long mempages
)
191 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
192 #ifndef ARCH_MIN_TASKALIGN
193 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
195 /* create a slab on which task_structs can be allocated */
197 kmem_cache_create("task_struct", sizeof(struct task_struct
),
198 ARCH_MIN_TASKALIGN
, SLAB_PANIC
| SLAB_NOTRACK
, NULL
);
201 /* do the arch specific task caches init */
202 arch_task_cache_init();
205 * The default maximum number of threads is set to a safe
206 * value: the thread structures can take up at most half
209 max_threads
= mempages
/ (8 * THREAD_SIZE
/ PAGE_SIZE
);
212 * we need to allow at least 20 threads to boot a system
217 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_cur
= max_threads
/2;
218 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_max
= max_threads
/2;
219 init_task
.signal
->rlim
[RLIMIT_SIGPENDING
] =
220 init_task
.signal
->rlim
[RLIMIT_NPROC
];
223 int __attribute__((weak
)) arch_dup_task_struct(struct task_struct
*dst
,
224 struct task_struct
*src
)
230 static struct task_struct
*dup_task_struct(struct task_struct
*orig
)
232 struct task_struct
*tsk
;
233 struct thread_info
*ti
;
234 unsigned long *stackend
;
238 prepare_to_copy(orig
);
240 tsk
= alloc_task_struct();
244 ti
= alloc_thread_info(tsk
);
246 free_task_struct(tsk
);
250 err
= arch_dup_task_struct(tsk
, orig
);
256 err
= prop_local_init_single(&tsk
->dirties
);
260 setup_thread_stack(tsk
, orig
);
261 clear_user_return_notifier(tsk
);
262 stackend
= end_of_stack(tsk
);
263 *stackend
= STACK_END_MAGIC
; /* for overflow detection */
265 #ifdef CONFIG_CC_STACKPROTECTOR
266 tsk
->stack_canary
= get_random_int();
269 /* One for us, one for whoever does the "release_task()" (usually parent) */
270 atomic_set(&tsk
->usage
,2);
271 atomic_set(&tsk
->fs_excl
, 0);
272 #ifdef CONFIG_BLK_DEV_IO_TRACE
275 tsk
->splice_pipe
= NULL
;
277 account_kernel_stack(ti
, 1);
282 free_thread_info(ti
);
283 free_task_struct(tsk
);
288 static int dup_mmap(struct mm_struct
*mm
, struct mm_struct
*oldmm
)
290 struct vm_area_struct
*mpnt
, *tmp
, *prev
, **pprev
;
291 struct rb_node
**rb_link
, *rb_parent
;
293 unsigned long charge
;
294 struct mempolicy
*pol
;
296 down_write(&oldmm
->mmap_sem
);
297 flush_cache_dup_mm(oldmm
);
299 * Not linked in yet - no deadlock potential:
301 down_write_nested(&mm
->mmap_sem
, SINGLE_DEPTH_NESTING
);
305 mm
->mmap_cache
= NULL
;
306 mm
->free_area_cache
= oldmm
->mmap_base
;
307 mm
->cached_hole_size
= ~0UL;
309 cpumask_clear(mm_cpumask(mm
));
311 rb_link
= &mm
->mm_rb
.rb_node
;
314 retval
= ksm_fork(mm
, oldmm
);
319 for (mpnt
= oldmm
->mmap
; mpnt
; mpnt
= mpnt
->vm_next
) {
322 if (mpnt
->vm_flags
& VM_DONTCOPY
) {
323 long pages
= vma_pages(mpnt
);
324 mm
->total_vm
-= pages
;
325 vm_stat_account(mm
, mpnt
->vm_flags
, mpnt
->vm_file
,
330 if (mpnt
->vm_flags
& VM_ACCOUNT
) {
331 unsigned int len
= (mpnt
->vm_end
- mpnt
->vm_start
) >> PAGE_SHIFT
;
332 if (security_vm_enough_memory(len
))
336 tmp
= kmem_cache_alloc(vm_area_cachep
, GFP_KERNEL
);
340 INIT_LIST_HEAD(&tmp
->anon_vma_chain
);
341 pol
= mpol_dup(vma_policy(mpnt
));
342 retval
= PTR_ERR(pol
);
344 goto fail_nomem_policy
;
345 vma_set_policy(tmp
, pol
);
346 if (anon_vma_fork(tmp
, mpnt
))
347 goto fail_nomem_anon_vma_fork
;
348 tmp
->vm_flags
&= ~VM_LOCKED
;
350 tmp
->vm_next
= tmp
->vm_prev
= NULL
;
353 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
354 struct address_space
*mapping
= file
->f_mapping
;
357 if (tmp
->vm_flags
& VM_DENYWRITE
)
358 atomic_dec(&inode
->i_writecount
);
359 spin_lock(&mapping
->i_mmap_lock
);
360 if (tmp
->vm_flags
& VM_SHARED
)
361 mapping
->i_mmap_writable
++;
362 tmp
->vm_truncate_count
= mpnt
->vm_truncate_count
;
363 flush_dcache_mmap_lock(mapping
);
364 /* insert tmp into the share list, just after mpnt */
365 vma_prio_tree_add(tmp
, mpnt
);
366 flush_dcache_mmap_unlock(mapping
);
367 spin_unlock(&mapping
->i_mmap_lock
);
371 * Clear hugetlb-related page reserves for children. This only
372 * affects MAP_PRIVATE mappings. Faults generated by the child
373 * are not guaranteed to succeed, even if read-only
375 if (is_vm_hugetlb_page(tmp
))
376 reset_vma_resv_huge_pages(tmp
);
379 * Link in the new vma and copy the page table entries.
382 pprev
= &tmp
->vm_next
;
386 __vma_link_rb(mm
, tmp
, rb_link
, rb_parent
);
387 rb_link
= &tmp
->vm_rb
.rb_right
;
388 rb_parent
= &tmp
->vm_rb
;
391 retval
= copy_page_range(mm
, oldmm
, mpnt
);
393 if (tmp
->vm_ops
&& tmp
->vm_ops
->open
)
394 tmp
->vm_ops
->open(tmp
);
399 /* a new mm has just been created */
400 arch_dup_mmap(oldmm
, mm
);
403 up_write(&mm
->mmap_sem
);
405 up_write(&oldmm
->mmap_sem
);
407 fail_nomem_anon_vma_fork
:
410 kmem_cache_free(vm_area_cachep
, tmp
);
413 vm_unacct_memory(charge
);
417 static inline int mm_alloc_pgd(struct mm_struct
* mm
)
419 mm
->pgd
= pgd_alloc(mm
);
420 if (unlikely(!mm
->pgd
))
425 static inline void mm_free_pgd(struct mm_struct
* mm
)
427 pgd_free(mm
, mm
->pgd
);
430 #define dup_mmap(mm, oldmm) (0)
431 #define mm_alloc_pgd(mm) (0)
432 #define mm_free_pgd(mm)
433 #endif /* CONFIG_MMU */
435 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(mmlist_lock
);
437 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
438 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
440 static unsigned long default_dump_filter
= MMF_DUMP_FILTER_DEFAULT
;
442 static int __init
coredump_filter_setup(char *s
)
444 default_dump_filter
=
445 (simple_strtoul(s
, NULL
, 0) << MMF_DUMP_FILTER_SHIFT
) &
446 MMF_DUMP_FILTER_MASK
;
450 __setup("coredump_filter=", coredump_filter_setup
);
452 #include <linux/init_task.h>
454 static void mm_init_aio(struct mm_struct
*mm
)
457 spin_lock_init(&mm
->ioctx_lock
);
458 INIT_HLIST_HEAD(&mm
->ioctx_list
);
462 static struct mm_struct
* mm_init(struct mm_struct
* mm
, struct task_struct
*p
)
464 atomic_set(&mm
->mm_users
, 1);
465 atomic_set(&mm
->mm_count
, 1);
466 init_rwsem(&mm
->mmap_sem
);
467 INIT_LIST_HEAD(&mm
->mmlist
);
468 mm
->flags
= (current
->mm
) ?
469 (current
->mm
->flags
& MMF_INIT_MASK
) : default_dump_filter
;
470 mm
->core_state
= NULL
;
472 memset(&mm
->rss_stat
, 0, sizeof(mm
->rss_stat
));
473 spin_lock_init(&mm
->page_table_lock
);
474 mm
->free_area_cache
= TASK_UNMAPPED_BASE
;
475 mm
->cached_hole_size
= ~0UL;
477 mm_init_owner(mm
, p
);
479 if (likely(!mm_alloc_pgd(mm
))) {
481 mmu_notifier_mm_init(mm
);
490 * Allocate and initialize an mm_struct.
492 struct mm_struct
* mm_alloc(void)
494 struct mm_struct
* mm
;
498 memset(mm
, 0, sizeof(*mm
));
499 mm
= mm_init(mm
, current
);
505 * Called when the last reference to the mm
506 * is dropped: either by a lazy thread or by
507 * mmput. Free the page directory and the mm.
509 void __mmdrop(struct mm_struct
*mm
)
511 BUG_ON(mm
== &init_mm
);
514 mmu_notifier_mm_destroy(mm
);
517 EXPORT_SYMBOL_GPL(__mmdrop
);
520 * Decrement the use count and release all resources for an mm.
522 void mmput(struct mm_struct
*mm
)
526 if (atomic_dec_and_test(&mm
->mm_users
)) {
530 set_mm_exe_file(mm
, NULL
);
531 if (!list_empty(&mm
->mmlist
)) {
532 spin_lock(&mmlist_lock
);
533 list_del(&mm
->mmlist
);
534 spin_unlock(&mmlist_lock
);
538 module_put(mm
->binfmt
->module
);
542 EXPORT_SYMBOL_GPL(mmput
);
545 * get_task_mm - acquire a reference to the task's mm
547 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
548 * this kernel workthread has transiently adopted a user mm with use_mm,
549 * to do its AIO) is not set and if so returns a reference to it, after
550 * bumping up the use count. User must release the mm via mmput()
551 * after use. Typically used by /proc and ptrace.
553 struct mm_struct
*get_task_mm(struct task_struct
*task
)
555 struct mm_struct
*mm
;
560 if (task
->flags
& PF_KTHREAD
)
563 atomic_inc(&mm
->mm_users
);
568 EXPORT_SYMBOL_GPL(get_task_mm
);
570 /* Please note the differences between mmput and mm_release.
571 * mmput is called whenever we stop holding onto a mm_struct,
572 * error success whatever.
574 * mm_release is called after a mm_struct has been removed
575 * from the current process.
577 * This difference is important for error handling, when we
578 * only half set up a mm_struct for a new process and need to restore
579 * the old one. Because we mmput the new mm_struct before
580 * restoring the old one. . .
581 * Eric Biederman 10 January 1998
583 void mm_release(struct task_struct
*tsk
, struct mm_struct
*mm
)
585 struct completion
*vfork_done
= tsk
->vfork_done
;
587 /* Get rid of any futexes when releasing the mm */
589 if (unlikely(tsk
->robust_list
)) {
590 exit_robust_list(tsk
);
591 tsk
->robust_list
= NULL
;
594 if (unlikely(tsk
->compat_robust_list
)) {
595 compat_exit_robust_list(tsk
);
596 tsk
->compat_robust_list
= NULL
;
599 if (unlikely(!list_empty(&tsk
->pi_state_list
)))
600 exit_pi_state_list(tsk
);
603 /* Get rid of any cached register state */
604 deactivate_mm(tsk
, mm
);
606 /* notify parent sleeping on vfork() */
608 tsk
->vfork_done
= NULL
;
609 complete(vfork_done
);
613 * If we're exiting normally, clear a user-space tid field if
614 * requested. We leave this alone when dying by signal, to leave
615 * the value intact in a core dump, and to save the unnecessary
616 * trouble otherwise. Userland only wants this done for a sys_exit.
618 if (tsk
->clear_child_tid
) {
619 if (!(tsk
->flags
& PF_SIGNALED
) &&
620 atomic_read(&mm
->mm_users
) > 1) {
622 * We don't check the error code - if userspace has
623 * not set up a proper pointer then tough luck.
625 put_user(0, tsk
->clear_child_tid
);
626 sys_futex(tsk
->clear_child_tid
, FUTEX_WAKE
,
629 tsk
->clear_child_tid
= NULL
;
634 * Allocate a new mm structure and copy contents from the
635 * mm structure of the passed in task structure.
637 struct mm_struct
*dup_mm(struct task_struct
*tsk
)
639 struct mm_struct
*mm
, *oldmm
= current
->mm
;
649 memcpy(mm
, oldmm
, sizeof(*mm
));
651 /* Initializing for Swap token stuff */
652 mm
->token_priority
= 0;
653 mm
->last_interval
= 0;
655 if (!mm_init(mm
, tsk
))
658 if (init_new_context(tsk
, mm
))
661 dup_mm_exe_file(oldmm
, mm
);
663 err
= dup_mmap(mm
, oldmm
);
667 mm
->hiwater_rss
= get_mm_rss(mm
);
668 mm
->hiwater_vm
= mm
->total_vm
;
670 if (mm
->binfmt
&& !try_module_get(mm
->binfmt
->module
))
676 /* don't put binfmt in mmput, we haven't got module yet */
685 * If init_new_context() failed, we cannot use mmput() to free the mm
686 * because it calls destroy_context()
693 static int copy_mm(unsigned long clone_flags
, struct task_struct
* tsk
)
695 struct mm_struct
* mm
, *oldmm
;
698 tsk
->min_flt
= tsk
->maj_flt
= 0;
699 tsk
->nvcsw
= tsk
->nivcsw
= 0;
700 #ifdef CONFIG_DETECT_HUNG_TASK
701 tsk
->last_switch_count
= tsk
->nvcsw
+ tsk
->nivcsw
;
705 tsk
->active_mm
= NULL
;
708 * Are we cloning a kernel thread?
710 * We need to steal a active VM for that..
716 if (clone_flags
& CLONE_VM
) {
717 atomic_inc(&oldmm
->mm_users
);
728 /* Initializing for Swap token stuff */
729 mm
->token_priority
= 0;
730 mm
->last_interval
= 0;
740 static int copy_fs(unsigned long clone_flags
, struct task_struct
*tsk
)
742 struct fs_struct
*fs
= current
->fs
;
743 if (clone_flags
& CLONE_FS
) {
744 /* tsk->fs is already what we want */
745 write_lock(&fs
->lock
);
747 write_unlock(&fs
->lock
);
751 write_unlock(&fs
->lock
);
754 tsk
->fs
= copy_fs_struct(fs
);
760 static int copy_files(unsigned long clone_flags
, struct task_struct
* tsk
)
762 struct files_struct
*oldf
, *newf
;
766 * A background process may not have any files ...
768 oldf
= current
->files
;
772 if (clone_flags
& CLONE_FILES
) {
773 atomic_inc(&oldf
->count
);
777 newf
= dup_fd(oldf
, &error
);
787 static int copy_io(unsigned long clone_flags
, struct task_struct
*tsk
)
790 struct io_context
*ioc
= current
->io_context
;
795 * Share io context with parent, if CLONE_IO is set
797 if (clone_flags
& CLONE_IO
) {
798 tsk
->io_context
= ioc_task_link(ioc
);
799 if (unlikely(!tsk
->io_context
))
801 } else if (ioprio_valid(ioc
->ioprio
)) {
802 tsk
->io_context
= alloc_io_context(GFP_KERNEL
, -1);
803 if (unlikely(!tsk
->io_context
))
806 tsk
->io_context
->ioprio
= ioc
->ioprio
;
812 static int copy_sighand(unsigned long clone_flags
, struct task_struct
*tsk
)
814 struct sighand_struct
*sig
;
816 if (clone_flags
& CLONE_SIGHAND
) {
817 atomic_inc(¤t
->sighand
->count
);
820 sig
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
821 rcu_assign_pointer(tsk
->sighand
, sig
);
824 atomic_set(&sig
->count
, 1);
825 memcpy(sig
->action
, current
->sighand
->action
, sizeof(sig
->action
));
829 void __cleanup_sighand(struct sighand_struct
*sighand
)
831 if (atomic_dec_and_test(&sighand
->count
))
832 kmem_cache_free(sighand_cachep
, sighand
);
837 * Initialize POSIX timer handling for a thread group.
839 static void posix_cpu_timers_init_group(struct signal_struct
*sig
)
841 unsigned long cpu_limit
;
843 /* Thread group counters. */
844 thread_group_cputime_init(sig
);
846 cpu_limit
= ACCESS_ONCE(sig
->rlim
[RLIMIT_CPU
].rlim_cur
);
847 if (cpu_limit
!= RLIM_INFINITY
) {
848 sig
->cputime_expires
.prof_exp
= secs_to_cputime(cpu_limit
);
849 sig
->cputimer
.running
= 1;
852 /* The timer lists. */
853 INIT_LIST_HEAD(&sig
->cpu_timers
[0]);
854 INIT_LIST_HEAD(&sig
->cpu_timers
[1]);
855 INIT_LIST_HEAD(&sig
->cpu_timers
[2]);
858 static int copy_signal(unsigned long clone_flags
, struct task_struct
*tsk
)
860 struct signal_struct
*sig
;
862 if (clone_flags
& CLONE_THREAD
)
865 sig
= kmem_cache_zalloc(signal_cachep
, GFP_KERNEL
);
870 atomic_set(&sig
->count
, 1);
871 atomic_set(&sig
->live
, 1);
872 init_waitqueue_head(&sig
->wait_chldexit
);
873 if (clone_flags
& CLONE_NEWPID
)
874 sig
->flags
|= SIGNAL_UNKILLABLE
;
875 sig
->curr_target
= tsk
;
876 init_sigpending(&sig
->shared_pending
);
877 INIT_LIST_HEAD(&sig
->posix_timers
);
879 hrtimer_init(&sig
->real_timer
, CLOCK_MONOTONIC
, HRTIMER_MODE_REL
);
880 sig
->real_timer
.function
= it_real_fn
;
882 task_lock(current
->group_leader
);
883 memcpy(sig
->rlim
, current
->signal
->rlim
, sizeof sig
->rlim
);
884 task_unlock(current
->group_leader
);
886 posix_cpu_timers_init_group(sig
);
890 sig
->oom_adj
= current
->signal
->oom_adj
;
895 void __cleanup_signal(struct signal_struct
*sig
)
897 thread_group_cputime_free(sig
);
898 tty_kref_put(sig
->tty
);
899 kmem_cache_free(signal_cachep
, sig
);
902 static void copy_flags(unsigned long clone_flags
, struct task_struct
*p
)
904 unsigned long new_flags
= p
->flags
;
906 new_flags
&= ~PF_SUPERPRIV
;
907 new_flags
|= PF_FORKNOEXEC
;
908 new_flags
|= PF_STARTING
;
909 p
->flags
= new_flags
;
910 clear_freeze_flag(p
);
913 SYSCALL_DEFINE1(set_tid_address
, int __user
*, tidptr
)
915 current
->clear_child_tid
= tidptr
;
917 return task_pid_vnr(current
);
920 static void rt_mutex_init_task(struct task_struct
*p
)
922 raw_spin_lock_init(&p
->pi_lock
);
923 #ifdef CONFIG_RT_MUTEXES
924 plist_head_init_raw(&p
->pi_waiters
, &p
->pi_lock
);
925 p
->pi_blocked_on
= NULL
;
929 #ifdef CONFIG_MM_OWNER
930 void mm_init_owner(struct mm_struct
*mm
, struct task_struct
*p
)
934 #endif /* CONFIG_MM_OWNER */
937 * Initialize POSIX timer handling for a single task.
939 static void posix_cpu_timers_init(struct task_struct
*tsk
)
941 tsk
->cputime_expires
.prof_exp
= cputime_zero
;
942 tsk
->cputime_expires
.virt_exp
= cputime_zero
;
943 tsk
->cputime_expires
.sched_exp
= 0;
944 INIT_LIST_HEAD(&tsk
->cpu_timers
[0]);
945 INIT_LIST_HEAD(&tsk
->cpu_timers
[1]);
946 INIT_LIST_HEAD(&tsk
->cpu_timers
[2]);
950 * This creates a new process as a copy of the old one,
951 * but does not actually start it yet.
953 * It copies the registers, and all the appropriate
954 * parts of the process environment (as per the clone
955 * flags). The actual kick-off is left to the caller.
957 static struct task_struct
*copy_process(unsigned long clone_flags
,
958 unsigned long stack_start
,
959 struct pt_regs
*regs
,
960 unsigned long stack_size
,
961 int __user
*child_tidptr
,
966 struct task_struct
*p
;
967 int cgroup_callbacks_done
= 0;
969 if ((clone_flags
& (CLONE_NEWNS
|CLONE_FS
)) == (CLONE_NEWNS
|CLONE_FS
))
970 return ERR_PTR(-EINVAL
);
973 * Thread groups must share signals as well, and detached threads
974 * can only be started up within the thread group.
976 if ((clone_flags
& CLONE_THREAD
) && !(clone_flags
& CLONE_SIGHAND
))
977 return ERR_PTR(-EINVAL
);
980 * Shared signal handlers imply shared VM. By way of the above,
981 * thread groups also imply shared VM. Blocking this case allows
982 * for various simplifications in other code.
984 if ((clone_flags
& CLONE_SIGHAND
) && !(clone_flags
& CLONE_VM
))
985 return ERR_PTR(-EINVAL
);
988 * Siblings of global init remain as zombies on exit since they are
989 * not reaped by their parent (swapper). To solve this and to avoid
990 * multi-rooted process trees, prevent global and container-inits
991 * from creating siblings.
993 if ((clone_flags
& CLONE_PARENT
) &&
994 current
->signal
->flags
& SIGNAL_UNKILLABLE
)
995 return ERR_PTR(-EINVAL
);
997 retval
= security_task_create(clone_flags
);
1002 p
= dup_task_struct(current
);
1006 ftrace_graph_init_task(p
);
1008 rt_mutex_init_task(p
);
1010 #ifdef CONFIG_PROVE_LOCKING
1011 DEBUG_LOCKS_WARN_ON(!p
->hardirqs_enabled
);
1012 DEBUG_LOCKS_WARN_ON(!p
->softirqs_enabled
);
1015 if (atomic_read(&p
->real_cred
->user
->processes
) >=
1016 task_rlimit(p
, RLIMIT_NPROC
)) {
1017 if (!capable(CAP_SYS_ADMIN
) && !capable(CAP_SYS_RESOURCE
) &&
1018 p
->real_cred
->user
!= INIT_USER
)
1022 retval
= copy_creds(p
, clone_flags
);
1027 * If multiple threads are within copy_process(), then this check
1028 * triggers too late. This doesn't hurt, the check is only there
1029 * to stop root fork bombs.
1032 if (nr_threads
>= max_threads
)
1033 goto bad_fork_cleanup_count
;
1035 if (!try_module_get(task_thread_info(p
)->exec_domain
->module
))
1036 goto bad_fork_cleanup_count
;
1039 delayacct_tsk_init(p
); /* Must remain after dup_task_struct() */
1040 copy_flags(clone_flags
, p
);
1041 INIT_LIST_HEAD(&p
->children
);
1042 INIT_LIST_HEAD(&p
->sibling
);
1043 rcu_copy_process(p
);
1044 p
->vfork_done
= NULL
;
1045 spin_lock_init(&p
->alloc_lock
);
1047 init_sigpending(&p
->pending
);
1049 p
->utime
= cputime_zero
;
1050 p
->stime
= cputime_zero
;
1051 p
->gtime
= cputime_zero
;
1052 p
->utimescaled
= cputime_zero
;
1053 p
->stimescaled
= cputime_zero
;
1054 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
1055 p
->prev_utime
= cputime_zero
;
1056 p
->prev_stime
= cputime_zero
;
1058 #if defined(SPLIT_RSS_COUNTING)
1059 memset(&p
->rss_stat
, 0, sizeof(p
->rss_stat
));
1062 p
->default_timer_slack_ns
= current
->timer_slack_ns
;
1064 task_io_accounting_init(&p
->ioac
);
1065 acct_clear_integrals(p
);
1067 posix_cpu_timers_init(p
);
1069 p
->lock_depth
= -1; /* -1 = no lock */
1070 do_posix_clock_monotonic_gettime(&p
->start_time
);
1071 p
->real_start_time
= p
->start_time
;
1072 monotonic_to_bootbased(&p
->real_start_time
);
1073 p
->io_context
= NULL
;
1074 p
->audit_context
= NULL
;
1077 p
->mempolicy
= mpol_dup(p
->mempolicy
);
1078 if (IS_ERR(p
->mempolicy
)) {
1079 retval
= PTR_ERR(p
->mempolicy
);
1080 p
->mempolicy
= NULL
;
1081 goto bad_fork_cleanup_cgroup
;
1083 mpol_fix_fork_child_flag(p
);
1085 #ifdef CONFIG_TRACE_IRQFLAGS
1087 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1088 p
->hardirqs_enabled
= 1;
1090 p
->hardirqs_enabled
= 0;
1092 p
->hardirq_enable_ip
= 0;
1093 p
->hardirq_enable_event
= 0;
1094 p
->hardirq_disable_ip
= _THIS_IP_
;
1095 p
->hardirq_disable_event
= 0;
1096 p
->softirqs_enabled
= 1;
1097 p
->softirq_enable_ip
= _THIS_IP_
;
1098 p
->softirq_enable_event
= 0;
1099 p
->softirq_disable_ip
= 0;
1100 p
->softirq_disable_event
= 0;
1101 p
->hardirq_context
= 0;
1102 p
->softirq_context
= 0;
1104 #ifdef CONFIG_LOCKDEP
1105 p
->lockdep_depth
= 0; /* no locks held yet */
1106 p
->curr_chain_key
= 0;
1107 p
->lockdep_recursion
= 0;
1110 #ifdef CONFIG_DEBUG_MUTEXES
1111 p
->blocked_on
= NULL
; /* not blocked yet */
1113 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
1114 p
->memcg_batch
.do_batch
= 0;
1115 p
->memcg_batch
.memcg
= NULL
;
1120 /* Perform scheduler related setup. Assign this task to a CPU. */
1121 sched_fork(p
, clone_flags
);
1123 retval
= perf_event_init_task(p
);
1125 goto bad_fork_cleanup_policy
;
1127 if ((retval
= audit_alloc(p
)))
1128 goto bad_fork_cleanup_policy
;
1129 /* copy all the process information */
1130 if ((retval
= copy_semundo(clone_flags
, p
)))
1131 goto bad_fork_cleanup_audit
;
1132 if ((retval
= copy_files(clone_flags
, p
)))
1133 goto bad_fork_cleanup_semundo
;
1134 if ((retval
= copy_fs(clone_flags
, p
)))
1135 goto bad_fork_cleanup_files
;
1136 if ((retval
= copy_sighand(clone_flags
, p
)))
1137 goto bad_fork_cleanup_fs
;
1138 if ((retval
= copy_signal(clone_flags
, p
)))
1139 goto bad_fork_cleanup_sighand
;
1140 if ((retval
= copy_mm(clone_flags
, p
)))
1141 goto bad_fork_cleanup_signal
;
1142 if ((retval
= copy_namespaces(clone_flags
, p
)))
1143 goto bad_fork_cleanup_mm
;
1144 if ((retval
= copy_io(clone_flags
, p
)))
1145 goto bad_fork_cleanup_namespaces
;
1146 retval
= copy_thread(clone_flags
, stack_start
, stack_size
, p
, regs
);
1148 goto bad_fork_cleanup_io
;
1150 if (pid
!= &init_struct_pid
) {
1152 pid
= alloc_pid(p
->nsproxy
->pid_ns
);
1154 goto bad_fork_cleanup_io
;
1156 if (clone_flags
& CLONE_NEWPID
) {
1157 retval
= pid_ns_prepare_proc(p
->nsproxy
->pid_ns
);
1159 goto bad_fork_free_pid
;
1163 p
->pid
= pid_nr(pid
);
1165 if (clone_flags
& CLONE_THREAD
)
1166 p
->tgid
= current
->tgid
;
1168 if (current
->nsproxy
!= p
->nsproxy
) {
1169 retval
= ns_cgroup_clone(p
, pid
);
1171 goto bad_fork_free_pid
;
1174 p
->set_child_tid
= (clone_flags
& CLONE_CHILD_SETTID
) ? child_tidptr
: NULL
;
1176 * Clear TID on mm_release()?
1178 p
->clear_child_tid
= (clone_flags
& CLONE_CHILD_CLEARTID
) ? child_tidptr
: NULL
;
1180 p
->robust_list
= NULL
;
1181 #ifdef CONFIG_COMPAT
1182 p
->compat_robust_list
= NULL
;
1184 INIT_LIST_HEAD(&p
->pi_state_list
);
1185 p
->pi_state_cache
= NULL
;
1188 * sigaltstack should be cleared when sharing the same VM
1190 if ((clone_flags
& (CLONE_VM
|CLONE_VFORK
)) == CLONE_VM
)
1191 p
->sas_ss_sp
= p
->sas_ss_size
= 0;
1194 * Syscall tracing and stepping should be turned off in the
1195 * child regardless of CLONE_PTRACE.
1197 user_disable_single_step(p
);
1198 clear_tsk_thread_flag(p
, TIF_SYSCALL_TRACE
);
1199 #ifdef TIF_SYSCALL_EMU
1200 clear_tsk_thread_flag(p
, TIF_SYSCALL_EMU
);
1202 clear_all_latency_tracing(p
);
1204 /* ok, now we should be set up.. */
1205 p
->exit_signal
= (clone_flags
& CLONE_THREAD
) ? -1 : (clone_flags
& CSIGNAL
);
1206 p
->pdeath_signal
= 0;
1210 * Ok, make it visible to the rest of the system.
1211 * We dont wake it up yet.
1213 p
->group_leader
= p
;
1214 INIT_LIST_HEAD(&p
->thread_group
);
1216 /* Now that the task is set up, run cgroup callbacks if
1217 * necessary. We need to run them before the task is visible
1218 * on the tasklist. */
1219 cgroup_fork_callbacks(p
);
1220 cgroup_callbacks_done
= 1;
1222 /* Need tasklist lock for parent etc handling! */
1223 write_lock_irq(&tasklist_lock
);
1225 /* CLONE_PARENT re-uses the old parent */
1226 if (clone_flags
& (CLONE_PARENT
|CLONE_THREAD
)) {
1227 p
->real_parent
= current
->real_parent
;
1228 p
->parent_exec_id
= current
->parent_exec_id
;
1230 p
->real_parent
= current
;
1231 p
->parent_exec_id
= current
->self_exec_id
;
1234 spin_lock(¤t
->sighand
->siglock
);
1237 * Process group and session signals need to be delivered to just the
1238 * parent before the fork or both the parent and the child after the
1239 * fork. Restart if a signal comes in before we add the new process to
1240 * it's process group.
1241 * A fatal signal pending means that current will exit, so the new
1242 * thread can't slip out of an OOM kill (or normal SIGKILL).
1244 recalc_sigpending();
1245 if (signal_pending(current
)) {
1246 spin_unlock(¤t
->sighand
->siglock
);
1247 write_unlock_irq(&tasklist_lock
);
1248 retval
= -ERESTARTNOINTR
;
1249 goto bad_fork_free_pid
;
1252 if (clone_flags
& CLONE_THREAD
) {
1253 atomic_inc(¤t
->signal
->count
);
1254 atomic_inc(¤t
->signal
->live
);
1255 p
->group_leader
= current
->group_leader
;
1256 list_add_tail_rcu(&p
->thread_group
, &p
->group_leader
->thread_group
);
1259 if (likely(p
->pid
)) {
1260 tracehook_finish_clone(p
, clone_flags
, trace
);
1262 if (thread_group_leader(p
)) {
1263 if (clone_flags
& CLONE_NEWPID
)
1264 p
->nsproxy
->pid_ns
->child_reaper
= p
;
1266 p
->signal
->leader_pid
= pid
;
1267 tty_kref_put(p
->signal
->tty
);
1268 p
->signal
->tty
= tty_kref_get(current
->signal
->tty
);
1269 attach_pid(p
, PIDTYPE_PGID
, task_pgrp(current
));
1270 attach_pid(p
, PIDTYPE_SID
, task_session(current
));
1271 list_add_tail(&p
->sibling
, &p
->real_parent
->children
);
1272 list_add_tail_rcu(&p
->tasks
, &init_task
.tasks
);
1273 __get_cpu_var(process_counts
)++;
1275 attach_pid(p
, PIDTYPE_PID
, pid
);
1280 spin_unlock(¤t
->sighand
->siglock
);
1281 write_unlock_irq(&tasklist_lock
);
1282 proc_fork_connector(p
);
1283 cgroup_post_fork(p
);
1288 if (pid
!= &init_struct_pid
)
1290 bad_fork_cleanup_io
:
1293 bad_fork_cleanup_namespaces
:
1294 exit_task_namespaces(p
);
1295 bad_fork_cleanup_mm
:
1298 bad_fork_cleanup_signal
:
1299 if (!(clone_flags
& CLONE_THREAD
))
1300 __cleanup_signal(p
->signal
);
1301 bad_fork_cleanup_sighand
:
1302 __cleanup_sighand(p
->sighand
);
1303 bad_fork_cleanup_fs
:
1304 exit_fs(p
); /* blocking */
1305 bad_fork_cleanup_files
:
1306 exit_files(p
); /* blocking */
1307 bad_fork_cleanup_semundo
:
1309 bad_fork_cleanup_audit
:
1311 bad_fork_cleanup_policy
:
1312 perf_event_free_task(p
);
1314 mpol_put(p
->mempolicy
);
1315 bad_fork_cleanup_cgroup
:
1317 cgroup_exit(p
, cgroup_callbacks_done
);
1318 delayacct_tsk_free(p
);
1319 module_put(task_thread_info(p
)->exec_domain
->module
);
1320 bad_fork_cleanup_count
:
1321 atomic_dec(&p
->cred
->user
->processes
);
1326 return ERR_PTR(retval
);
1329 noinline
struct pt_regs
* __cpuinit
__attribute__((weak
)) idle_regs(struct pt_regs
*regs
)
1331 memset(regs
, 0, sizeof(struct pt_regs
));
1335 struct task_struct
* __cpuinit
fork_idle(int cpu
)
1337 struct task_struct
*task
;
1338 struct pt_regs regs
;
1340 task
= copy_process(CLONE_VM
, 0, idle_regs(®s
), 0, NULL
,
1341 &init_struct_pid
, 0);
1343 init_idle(task
, cpu
);
1349 * Ok, this is the main fork-routine.
1351 * It copies the process, and if successful kick-starts
1352 * it and waits for it to finish using the VM if required.
1354 long do_fork(unsigned long clone_flags
,
1355 unsigned long stack_start
,
1356 struct pt_regs
*regs
,
1357 unsigned long stack_size
,
1358 int __user
*parent_tidptr
,
1359 int __user
*child_tidptr
)
1361 struct task_struct
*p
;
1366 * Do some preliminary argument and permissions checking before we
1367 * actually start allocating stuff
1369 if (clone_flags
& CLONE_NEWUSER
) {
1370 if (clone_flags
& CLONE_THREAD
)
1372 /* hopefully this check will go away when userns support is
1375 if (!capable(CAP_SYS_ADMIN
) || !capable(CAP_SETUID
) ||
1376 !capable(CAP_SETGID
))
1381 * We hope to recycle these flags after 2.6.26
1383 if (unlikely(clone_flags
& CLONE_STOPPED
)) {
1384 static int __read_mostly count
= 100;
1386 if (count
> 0 && printk_ratelimit()) {
1387 char comm
[TASK_COMM_LEN
];
1390 printk(KERN_INFO
"fork(): process `%s' used deprecated "
1391 "clone flags 0x%lx\n",
1392 get_task_comm(comm
, current
),
1393 clone_flags
& CLONE_STOPPED
);
1398 * When called from kernel_thread, don't do user tracing stuff.
1400 if (likely(user_mode(regs
)))
1401 trace
= tracehook_prepare_clone(clone_flags
);
1403 p
= copy_process(clone_flags
, stack_start
, regs
, stack_size
,
1404 child_tidptr
, NULL
, trace
);
1406 * Do this prior waking up the new thread - the thread pointer
1407 * might get invalid after that point, if the thread exits quickly.
1410 struct completion vfork
;
1412 trace_sched_process_fork(current
, p
);
1414 nr
= task_pid_vnr(p
);
1416 if (clone_flags
& CLONE_PARENT_SETTID
)
1417 put_user(nr
, parent_tidptr
);
1419 if (clone_flags
& CLONE_VFORK
) {
1420 p
->vfork_done
= &vfork
;
1421 init_completion(&vfork
);
1424 audit_finish_fork(p
);
1425 tracehook_report_clone(regs
, clone_flags
, nr
, p
);
1428 * We set PF_STARTING at creation in case tracing wants to
1429 * use this to distinguish a fully live task from one that
1430 * hasn't gotten to tracehook_report_clone() yet. Now we
1431 * clear it and set the child going.
1433 p
->flags
&= ~PF_STARTING
;
1435 if (unlikely(clone_flags
& CLONE_STOPPED
)) {
1437 * We'll start up with an immediate SIGSTOP.
1439 sigaddset(&p
->pending
.signal
, SIGSTOP
);
1440 set_tsk_thread_flag(p
, TIF_SIGPENDING
);
1441 __set_task_state(p
, TASK_STOPPED
);
1443 wake_up_new_task(p
, clone_flags
);
1446 tracehook_report_clone_complete(trace
, regs
,
1447 clone_flags
, nr
, p
);
1449 if (clone_flags
& CLONE_VFORK
) {
1450 freezer_do_not_count();
1451 wait_for_completion(&vfork
);
1453 tracehook_report_vfork_done(p
, nr
);
1461 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1462 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1465 static void sighand_ctor(void *data
)
1467 struct sighand_struct
*sighand
= data
;
1469 spin_lock_init(&sighand
->siglock
);
1470 init_waitqueue_head(&sighand
->signalfd_wqh
);
1473 void __init
proc_caches_init(void)
1475 sighand_cachep
= kmem_cache_create("sighand_cache",
1476 sizeof(struct sighand_struct
), 0,
1477 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_DESTROY_BY_RCU
|
1478 SLAB_NOTRACK
, sighand_ctor
);
1479 signal_cachep
= kmem_cache_create("signal_cache",
1480 sizeof(struct signal_struct
), 0,
1481 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1482 files_cachep
= kmem_cache_create("files_cache",
1483 sizeof(struct files_struct
), 0,
1484 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1485 fs_cachep
= kmem_cache_create("fs_cache",
1486 sizeof(struct fs_struct
), 0,
1487 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1488 mm_cachep
= kmem_cache_create("mm_struct",
1489 sizeof(struct mm_struct
), ARCH_MIN_MMSTRUCT_ALIGN
,
1490 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1491 vm_area_cachep
= KMEM_CACHE(vm_area_struct
, SLAB_PANIC
);
1496 * Check constraints on flags passed to the unshare system call and
1497 * force unsharing of additional process context as appropriate.
1499 static void check_unshare_flags(unsigned long *flags_ptr
)
1502 * If unsharing a thread from a thread group, must also
1505 if (*flags_ptr
& CLONE_THREAD
)
1506 *flags_ptr
|= CLONE_VM
;
1509 * If unsharing vm, must also unshare signal handlers.
1511 if (*flags_ptr
& CLONE_VM
)
1512 *flags_ptr
|= CLONE_SIGHAND
;
1515 * If unsharing signal handlers and the task was created
1516 * using CLONE_THREAD, then must unshare the thread
1518 if ((*flags_ptr
& CLONE_SIGHAND
) &&
1519 (atomic_read(¤t
->signal
->count
) > 1))
1520 *flags_ptr
|= CLONE_THREAD
;
1523 * If unsharing namespace, must also unshare filesystem information.
1525 if (*flags_ptr
& CLONE_NEWNS
)
1526 *flags_ptr
|= CLONE_FS
;
1530 * Unsharing of tasks created with CLONE_THREAD is not supported yet
1532 static int unshare_thread(unsigned long unshare_flags
)
1534 if (unshare_flags
& CLONE_THREAD
)
1541 * Unshare the filesystem structure if it is being shared
1543 static int unshare_fs(unsigned long unshare_flags
, struct fs_struct
**new_fsp
)
1545 struct fs_struct
*fs
= current
->fs
;
1547 if (!(unshare_flags
& CLONE_FS
) || !fs
)
1550 /* don't need lock here; in the worst case we'll do useless copy */
1554 *new_fsp
= copy_fs_struct(fs
);
1562 * Unsharing of sighand is not supported yet
1564 static int unshare_sighand(unsigned long unshare_flags
, struct sighand_struct
**new_sighp
)
1566 struct sighand_struct
*sigh
= current
->sighand
;
1568 if ((unshare_flags
& CLONE_SIGHAND
) && atomic_read(&sigh
->count
) > 1)
1575 * Unshare vm if it is being shared
1577 static int unshare_vm(unsigned long unshare_flags
, struct mm_struct
**new_mmp
)
1579 struct mm_struct
*mm
= current
->mm
;
1581 if ((unshare_flags
& CLONE_VM
) &&
1582 (mm
&& atomic_read(&mm
->mm_users
) > 1)) {
1590 * Unshare file descriptor table if it is being shared
1592 static int unshare_fd(unsigned long unshare_flags
, struct files_struct
**new_fdp
)
1594 struct files_struct
*fd
= current
->files
;
1597 if ((unshare_flags
& CLONE_FILES
) &&
1598 (fd
&& atomic_read(&fd
->count
) > 1)) {
1599 *new_fdp
= dup_fd(fd
, &error
);
1608 * unshare allows a process to 'unshare' part of the process
1609 * context which was originally shared using clone. copy_*
1610 * functions used by do_fork() cannot be used here directly
1611 * because they modify an inactive task_struct that is being
1612 * constructed. Here we are modifying the current, active,
1615 SYSCALL_DEFINE1(unshare
, unsigned long, unshare_flags
)
1618 struct fs_struct
*fs
, *new_fs
= NULL
;
1619 struct sighand_struct
*new_sigh
= NULL
;
1620 struct mm_struct
*mm
, *new_mm
= NULL
, *active_mm
= NULL
;
1621 struct files_struct
*fd
, *new_fd
= NULL
;
1622 struct nsproxy
*new_nsproxy
= 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
|CLONE_NEWNET
))
1632 goto bad_unshare_out
;
1635 * CLONE_NEWIPC must also detach from the undolist: after switching
1636 * to a new ipc namespace, the semaphore arrays from the old
1637 * namespace are unreachable.
1639 if (unshare_flags
& (CLONE_NEWIPC
|CLONE_SYSVSEM
))
1641 if ((err
= unshare_thread(unshare_flags
)))
1642 goto bad_unshare_out
;
1643 if ((err
= unshare_fs(unshare_flags
, &new_fs
)))
1644 goto bad_unshare_cleanup_thread
;
1645 if ((err
= unshare_sighand(unshare_flags
, &new_sigh
)))
1646 goto bad_unshare_cleanup_fs
;
1647 if ((err
= unshare_vm(unshare_flags
, &new_mm
)))
1648 goto bad_unshare_cleanup_sigh
;
1649 if ((err
= unshare_fd(unshare_flags
, &new_fd
)))
1650 goto bad_unshare_cleanup_vm
;
1651 if ((err
= unshare_nsproxy_namespaces(unshare_flags
, &new_nsproxy
,
1653 goto bad_unshare_cleanup_fd
;
1655 if (new_fs
|| new_mm
|| new_fd
|| do_sysvsem
|| new_nsproxy
) {
1658 * CLONE_SYSVSEM is equivalent to sys_exit().
1664 switch_task_namespaces(current
, new_nsproxy
);
1672 write_lock(&fs
->lock
);
1673 current
->fs
= new_fs
;
1678 write_unlock(&fs
->lock
);
1683 active_mm
= current
->active_mm
;
1684 current
->mm
= new_mm
;
1685 current
->active_mm
= new_mm
;
1686 activate_mm(active_mm
, new_mm
);
1691 fd
= current
->files
;
1692 current
->files
= new_fd
;
1696 task_unlock(current
);
1700 put_nsproxy(new_nsproxy
);
1702 bad_unshare_cleanup_fd
:
1704 put_files_struct(new_fd
);
1706 bad_unshare_cleanup_vm
:
1710 bad_unshare_cleanup_sigh
:
1712 if (atomic_dec_and_test(&new_sigh
->count
))
1713 kmem_cache_free(sighand_cachep
, new_sigh
);
1715 bad_unshare_cleanup_fs
:
1717 free_fs_struct(new_fs
);
1719 bad_unshare_cleanup_thread
:
1725 * Helper to unshare the files of the current task.
1726 * We don't want to expose copy_files internals to
1727 * the exec layer of the kernel.
1730 int unshare_files(struct files_struct
**displaced
)
1732 struct task_struct
*task
= current
;
1733 struct files_struct
*copy
= NULL
;
1736 error
= unshare_fd(CLONE_FILES
, ©
);
1737 if (error
|| !copy
) {
1741 *displaced
= task
->files
;