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
, **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
);
318 for (mpnt
= oldmm
->mmap
; mpnt
; mpnt
= mpnt
->vm_next
) {
321 if (mpnt
->vm_flags
& VM_DONTCOPY
) {
322 long pages
= vma_pages(mpnt
);
323 mm
->total_vm
-= pages
;
324 vm_stat_account(mm
, mpnt
->vm_flags
, mpnt
->vm_file
,
329 if (mpnt
->vm_flags
& VM_ACCOUNT
) {
330 unsigned int len
= (mpnt
->vm_end
- mpnt
->vm_start
) >> PAGE_SHIFT
;
331 if (security_vm_enough_memory(len
))
335 tmp
= kmem_cache_alloc(vm_area_cachep
, GFP_KERNEL
);
339 pol
= mpol_dup(vma_policy(mpnt
));
340 retval
= PTR_ERR(pol
);
342 goto fail_nomem_policy
;
343 vma_set_policy(tmp
, pol
);
344 tmp
->vm_flags
&= ~VM_LOCKED
;
350 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
351 struct address_space
*mapping
= file
->f_mapping
;
354 if (tmp
->vm_flags
& VM_DENYWRITE
)
355 atomic_dec(&inode
->i_writecount
);
356 spin_lock(&mapping
->i_mmap_lock
);
357 if (tmp
->vm_flags
& VM_SHARED
)
358 mapping
->i_mmap_writable
++;
359 tmp
->vm_truncate_count
= mpnt
->vm_truncate_count
;
360 flush_dcache_mmap_lock(mapping
);
361 /* insert tmp into the share list, just after mpnt */
362 vma_prio_tree_add(tmp
, mpnt
);
363 flush_dcache_mmap_unlock(mapping
);
364 spin_unlock(&mapping
->i_mmap_lock
);
368 * Clear hugetlb-related page reserves for children. This only
369 * affects MAP_PRIVATE mappings. Faults generated by the child
370 * are not guaranteed to succeed, even if read-only
372 if (is_vm_hugetlb_page(tmp
))
373 reset_vma_resv_huge_pages(tmp
);
376 * Link in the new vma and copy the page table entries.
379 pprev
= &tmp
->vm_next
;
381 __vma_link_rb(mm
, tmp
, rb_link
, rb_parent
);
382 rb_link
= &tmp
->vm_rb
.rb_right
;
383 rb_parent
= &tmp
->vm_rb
;
386 retval
= copy_page_range(mm
, oldmm
, mpnt
);
388 if (tmp
->vm_ops
&& tmp
->vm_ops
->open
)
389 tmp
->vm_ops
->open(tmp
);
394 /* a new mm has just been created */
395 arch_dup_mmap(oldmm
, mm
);
398 up_write(&mm
->mmap_sem
);
400 up_write(&oldmm
->mmap_sem
);
403 kmem_cache_free(vm_area_cachep
, tmp
);
406 vm_unacct_memory(charge
);
410 static inline int mm_alloc_pgd(struct mm_struct
* mm
)
412 mm
->pgd
= pgd_alloc(mm
);
413 if (unlikely(!mm
->pgd
))
418 static inline void mm_free_pgd(struct mm_struct
* mm
)
420 pgd_free(mm
, mm
->pgd
);
423 #define dup_mmap(mm, oldmm) (0)
424 #define mm_alloc_pgd(mm) (0)
425 #define mm_free_pgd(mm)
426 #endif /* CONFIG_MMU */
428 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(mmlist_lock
);
430 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
431 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
433 static unsigned long default_dump_filter
= MMF_DUMP_FILTER_DEFAULT
;
435 static int __init
coredump_filter_setup(char *s
)
437 default_dump_filter
=
438 (simple_strtoul(s
, NULL
, 0) << MMF_DUMP_FILTER_SHIFT
) &
439 MMF_DUMP_FILTER_MASK
;
443 __setup("coredump_filter=", coredump_filter_setup
);
445 #include <linux/init_task.h>
447 static void mm_init_aio(struct mm_struct
*mm
)
450 spin_lock_init(&mm
->ioctx_lock
);
451 INIT_HLIST_HEAD(&mm
->ioctx_list
);
455 static struct mm_struct
* mm_init(struct mm_struct
* mm
, struct task_struct
*p
)
457 atomic_set(&mm
->mm_users
, 1);
458 atomic_set(&mm
->mm_count
, 1);
459 init_rwsem(&mm
->mmap_sem
);
460 INIT_LIST_HEAD(&mm
->mmlist
);
461 mm
->flags
= (current
->mm
) ?
462 (current
->mm
->flags
& MMF_INIT_MASK
) : default_dump_filter
;
463 mm
->core_state
= NULL
;
465 set_mm_counter(mm
, file_rss
, 0);
466 set_mm_counter(mm
, anon_rss
, 0);
467 spin_lock_init(&mm
->page_table_lock
);
468 mm
->free_area_cache
= TASK_UNMAPPED_BASE
;
469 mm
->cached_hole_size
= ~0UL;
471 mm_init_owner(mm
, p
);
473 if (likely(!mm_alloc_pgd(mm
))) {
475 mmu_notifier_mm_init(mm
);
484 * Allocate and initialize an mm_struct.
486 struct mm_struct
* mm_alloc(void)
488 struct mm_struct
* mm
;
492 memset(mm
, 0, sizeof(*mm
));
493 mm
= mm_init(mm
, current
);
499 * Called when the last reference to the mm
500 * is dropped: either by a lazy thread or by
501 * mmput. Free the page directory and the mm.
503 void __mmdrop(struct mm_struct
*mm
)
505 BUG_ON(mm
== &init_mm
);
508 mmu_notifier_mm_destroy(mm
);
511 EXPORT_SYMBOL_GPL(__mmdrop
);
514 * Decrement the use count and release all resources for an mm.
516 void mmput(struct mm_struct
*mm
)
520 if (atomic_dec_and_test(&mm
->mm_users
)) {
524 set_mm_exe_file(mm
, NULL
);
525 if (!list_empty(&mm
->mmlist
)) {
526 spin_lock(&mmlist_lock
);
527 list_del(&mm
->mmlist
);
528 spin_unlock(&mmlist_lock
);
532 module_put(mm
->binfmt
->module
);
536 EXPORT_SYMBOL_GPL(mmput
);
539 * get_task_mm - acquire a reference to the task's mm
541 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
542 * this kernel workthread has transiently adopted a user mm with use_mm,
543 * to do its AIO) is not set and if so returns a reference to it, after
544 * bumping up the use count. User must release the mm via mmput()
545 * after use. Typically used by /proc and ptrace.
547 struct mm_struct
*get_task_mm(struct task_struct
*task
)
549 struct mm_struct
*mm
;
554 if (task
->flags
& PF_KTHREAD
)
557 atomic_inc(&mm
->mm_users
);
562 EXPORT_SYMBOL_GPL(get_task_mm
);
564 /* Please note the differences between mmput and mm_release.
565 * mmput is called whenever we stop holding onto a mm_struct,
566 * error success whatever.
568 * mm_release is called after a mm_struct has been removed
569 * from the current process.
571 * This difference is important for error handling, when we
572 * only half set up a mm_struct for a new process and need to restore
573 * the old one. Because we mmput the new mm_struct before
574 * restoring the old one. . .
575 * Eric Biederman 10 January 1998
577 void mm_release(struct task_struct
*tsk
, struct mm_struct
*mm
)
579 struct completion
*vfork_done
= tsk
->vfork_done
;
581 /* Get rid of any futexes when releasing the mm */
583 if (unlikely(tsk
->robust_list
)) {
584 exit_robust_list(tsk
);
585 tsk
->robust_list
= NULL
;
588 if (unlikely(tsk
->compat_robust_list
)) {
589 compat_exit_robust_list(tsk
);
590 tsk
->compat_robust_list
= NULL
;
593 if (unlikely(!list_empty(&tsk
->pi_state_list
)))
594 exit_pi_state_list(tsk
);
597 /* Get rid of any cached register state */
598 deactivate_mm(tsk
, mm
);
600 /* notify parent sleeping on vfork() */
602 tsk
->vfork_done
= NULL
;
603 complete(vfork_done
);
607 * If we're exiting normally, clear a user-space tid field if
608 * requested. We leave this alone when dying by signal, to leave
609 * the value intact in a core dump, and to save the unnecessary
610 * trouble otherwise. Userland only wants this done for a sys_exit.
612 if (tsk
->clear_child_tid
) {
613 if (!(tsk
->flags
& PF_SIGNALED
) &&
614 atomic_read(&mm
->mm_users
) > 1) {
616 * We don't check the error code - if userspace has
617 * not set up a proper pointer then tough luck.
619 put_user(0, tsk
->clear_child_tid
);
620 sys_futex(tsk
->clear_child_tid
, FUTEX_WAKE
,
623 tsk
->clear_child_tid
= NULL
;
628 * Allocate a new mm structure and copy contents from the
629 * mm structure of the passed in task structure.
631 struct mm_struct
*dup_mm(struct task_struct
*tsk
)
633 struct mm_struct
*mm
, *oldmm
= current
->mm
;
643 memcpy(mm
, oldmm
, sizeof(*mm
));
645 /* Initializing for Swap token stuff */
646 mm
->token_priority
= 0;
647 mm
->last_interval
= 0;
649 if (!mm_init(mm
, tsk
))
652 if (init_new_context(tsk
, mm
))
655 dup_mm_exe_file(oldmm
, mm
);
657 err
= dup_mmap(mm
, oldmm
);
661 mm
->hiwater_rss
= get_mm_rss(mm
);
662 mm
->hiwater_vm
= mm
->total_vm
;
664 if (mm
->binfmt
&& !try_module_get(mm
->binfmt
->module
))
670 /* don't put binfmt in mmput, we haven't got module yet */
679 * If init_new_context() failed, we cannot use mmput() to free the mm
680 * because it calls destroy_context()
687 static int copy_mm(unsigned long clone_flags
, struct task_struct
* tsk
)
689 struct mm_struct
* mm
, *oldmm
;
692 tsk
->min_flt
= tsk
->maj_flt
= 0;
693 tsk
->nvcsw
= tsk
->nivcsw
= 0;
694 #ifdef CONFIG_DETECT_HUNG_TASK
695 tsk
->last_switch_count
= tsk
->nvcsw
+ tsk
->nivcsw
;
699 tsk
->active_mm
= NULL
;
702 * Are we cloning a kernel thread?
704 * We need to steal a active VM for that..
710 if (clone_flags
& CLONE_VM
) {
711 atomic_inc(&oldmm
->mm_users
);
722 /* Initializing for Swap token stuff */
723 mm
->token_priority
= 0;
724 mm
->last_interval
= 0;
734 static int copy_fs(unsigned long clone_flags
, struct task_struct
*tsk
)
736 struct fs_struct
*fs
= current
->fs
;
737 if (clone_flags
& CLONE_FS
) {
738 /* tsk->fs is already what we want */
739 write_lock(&fs
->lock
);
741 write_unlock(&fs
->lock
);
745 write_unlock(&fs
->lock
);
748 tsk
->fs
= copy_fs_struct(fs
);
754 static int copy_files(unsigned long clone_flags
, struct task_struct
* tsk
)
756 struct files_struct
*oldf
, *newf
;
760 * A background process may not have any files ...
762 oldf
= current
->files
;
766 if (clone_flags
& CLONE_FILES
) {
767 atomic_inc(&oldf
->count
);
771 newf
= dup_fd(oldf
, &error
);
781 static int copy_io(unsigned long clone_flags
, struct task_struct
*tsk
)
784 struct io_context
*ioc
= current
->io_context
;
789 * Share io context with parent, if CLONE_IO is set
791 if (clone_flags
& CLONE_IO
) {
792 tsk
->io_context
= ioc_task_link(ioc
);
793 if (unlikely(!tsk
->io_context
))
795 } else if (ioprio_valid(ioc
->ioprio
)) {
796 tsk
->io_context
= alloc_io_context(GFP_KERNEL
, -1);
797 if (unlikely(!tsk
->io_context
))
800 tsk
->io_context
->ioprio
= ioc
->ioprio
;
806 static int copy_sighand(unsigned long clone_flags
, struct task_struct
*tsk
)
808 struct sighand_struct
*sig
;
810 if (clone_flags
& CLONE_SIGHAND
) {
811 atomic_inc(¤t
->sighand
->count
);
814 sig
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
815 rcu_assign_pointer(tsk
->sighand
, sig
);
818 atomic_set(&sig
->count
, 1);
819 memcpy(sig
->action
, current
->sighand
->action
, sizeof(sig
->action
));
823 void __cleanup_sighand(struct sighand_struct
*sighand
)
825 if (atomic_dec_and_test(&sighand
->count
))
826 kmem_cache_free(sighand_cachep
, sighand
);
831 * Initialize POSIX timer handling for a thread group.
833 static void posix_cpu_timers_init_group(struct signal_struct
*sig
)
835 /* Thread group counters. */
836 thread_group_cputime_init(sig
);
838 /* Expiration times and increments. */
839 sig
->it
[CPUCLOCK_PROF
].expires
= cputime_zero
;
840 sig
->it
[CPUCLOCK_PROF
].incr
= cputime_zero
;
841 sig
->it
[CPUCLOCK_VIRT
].expires
= cputime_zero
;
842 sig
->it
[CPUCLOCK_VIRT
].incr
= cputime_zero
;
844 /* Cached expiration times. */
845 sig
->cputime_expires
.prof_exp
= cputime_zero
;
846 sig
->cputime_expires
.virt_exp
= cputime_zero
;
847 sig
->cputime_expires
.sched_exp
= 0;
849 if (sig
->rlim
[RLIMIT_CPU
].rlim_cur
!= RLIM_INFINITY
) {
850 sig
->cputime_expires
.prof_exp
=
851 secs_to_cputime(sig
->rlim
[RLIMIT_CPU
].rlim_cur
);
852 sig
->cputimer
.running
= 1;
855 /* The timer lists. */
856 INIT_LIST_HEAD(&sig
->cpu_timers
[0]);
857 INIT_LIST_HEAD(&sig
->cpu_timers
[1]);
858 INIT_LIST_HEAD(&sig
->cpu_timers
[2]);
861 static int copy_signal(unsigned long clone_flags
, struct task_struct
*tsk
)
863 struct signal_struct
*sig
;
865 if (clone_flags
& CLONE_THREAD
)
868 sig
= kmem_cache_alloc(signal_cachep
, GFP_KERNEL
);
873 atomic_set(&sig
->count
, 1);
874 atomic_set(&sig
->live
, 1);
875 init_waitqueue_head(&sig
->wait_chldexit
);
877 if (clone_flags
& CLONE_NEWPID
)
878 sig
->flags
|= SIGNAL_UNKILLABLE
;
879 sig
->group_exit_code
= 0;
880 sig
->group_exit_task
= NULL
;
881 sig
->group_stop_count
= 0;
882 sig
->curr_target
= tsk
;
883 init_sigpending(&sig
->shared_pending
);
884 INIT_LIST_HEAD(&sig
->posix_timers
);
886 hrtimer_init(&sig
->real_timer
, CLOCK_MONOTONIC
, HRTIMER_MODE_REL
);
887 sig
->it_real_incr
.tv64
= 0;
888 sig
->real_timer
.function
= it_real_fn
;
890 sig
->leader
= 0; /* session leadership doesn't inherit */
891 sig
->tty_old_pgrp
= NULL
;
894 sig
->utime
= sig
->stime
= sig
->cutime
= sig
->cstime
= cputime_zero
;
895 sig
->gtime
= cputime_zero
;
896 sig
->cgtime
= cputime_zero
;
897 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
898 sig
->prev_utime
= sig
->prev_stime
= cputime_zero
;
900 sig
->nvcsw
= sig
->nivcsw
= sig
->cnvcsw
= sig
->cnivcsw
= 0;
901 sig
->min_flt
= sig
->maj_flt
= sig
->cmin_flt
= sig
->cmaj_flt
= 0;
902 sig
->inblock
= sig
->oublock
= sig
->cinblock
= sig
->coublock
= 0;
903 sig
->maxrss
= sig
->cmaxrss
= 0;
904 task_io_accounting_init(&sig
->ioac
);
905 sig
->sum_sched_runtime
= 0;
906 taskstats_tgid_init(sig
);
908 task_lock(current
->group_leader
);
909 memcpy(sig
->rlim
, current
->signal
->rlim
, sizeof sig
->rlim
);
910 task_unlock(current
->group_leader
);
912 posix_cpu_timers_init_group(sig
);
914 acct_init_pacct(&sig
->pacct
);
918 sig
->oom_adj
= current
->signal
->oom_adj
;
923 void __cleanup_signal(struct signal_struct
*sig
)
925 thread_group_cputime_free(sig
);
926 tty_kref_put(sig
->tty
);
927 kmem_cache_free(signal_cachep
, sig
);
930 static void copy_flags(unsigned long clone_flags
, struct task_struct
*p
)
932 unsigned long new_flags
= p
->flags
;
934 new_flags
&= ~PF_SUPERPRIV
;
935 new_flags
|= PF_FORKNOEXEC
;
936 new_flags
|= PF_STARTING
;
937 p
->flags
= new_flags
;
938 clear_freeze_flag(p
);
941 SYSCALL_DEFINE1(set_tid_address
, int __user
*, tidptr
)
943 current
->clear_child_tid
= tidptr
;
945 return task_pid_vnr(current
);
948 static void rt_mutex_init_task(struct task_struct
*p
)
950 raw_spin_lock_init(&p
->pi_lock
);
951 #ifdef CONFIG_RT_MUTEXES
952 plist_head_init_raw(&p
->pi_waiters
, &p
->pi_lock
);
953 p
->pi_blocked_on
= NULL
;
957 #ifdef CONFIG_MM_OWNER
958 void mm_init_owner(struct mm_struct
*mm
, struct task_struct
*p
)
962 #endif /* CONFIG_MM_OWNER */
965 * Initialize POSIX timer handling for a single task.
967 static void posix_cpu_timers_init(struct task_struct
*tsk
)
969 tsk
->cputime_expires
.prof_exp
= cputime_zero
;
970 tsk
->cputime_expires
.virt_exp
= cputime_zero
;
971 tsk
->cputime_expires
.sched_exp
= 0;
972 INIT_LIST_HEAD(&tsk
->cpu_timers
[0]);
973 INIT_LIST_HEAD(&tsk
->cpu_timers
[1]);
974 INIT_LIST_HEAD(&tsk
->cpu_timers
[2]);
978 * This creates a new process as a copy of the old one,
979 * but does not actually start it yet.
981 * It copies the registers, and all the appropriate
982 * parts of the process environment (as per the clone
983 * flags). The actual kick-off is left to the caller.
985 static struct task_struct
*copy_process(unsigned long clone_flags
,
986 unsigned long stack_start
,
987 struct pt_regs
*regs
,
988 unsigned long stack_size
,
989 int __user
*child_tidptr
,
994 struct task_struct
*p
;
995 int cgroup_callbacks_done
= 0;
997 if ((clone_flags
& (CLONE_NEWNS
|CLONE_FS
)) == (CLONE_NEWNS
|CLONE_FS
))
998 return ERR_PTR(-EINVAL
);
1001 * Thread groups must share signals as well, and detached threads
1002 * can only be started up within the thread group.
1004 if ((clone_flags
& CLONE_THREAD
) && !(clone_flags
& CLONE_SIGHAND
))
1005 return ERR_PTR(-EINVAL
);
1008 * Shared signal handlers imply shared VM. By way of the above,
1009 * thread groups also imply shared VM. Blocking this case allows
1010 * for various simplifications in other code.
1012 if ((clone_flags
& CLONE_SIGHAND
) && !(clone_flags
& CLONE_VM
))
1013 return ERR_PTR(-EINVAL
);
1016 * Siblings of global init remain as zombies on exit since they are
1017 * not reaped by their parent (swapper). To solve this and to avoid
1018 * multi-rooted process trees, prevent global and container-inits
1019 * from creating siblings.
1021 if ((clone_flags
& CLONE_PARENT
) &&
1022 current
->signal
->flags
& SIGNAL_UNKILLABLE
)
1023 return ERR_PTR(-EINVAL
);
1025 retval
= security_task_create(clone_flags
);
1030 p
= dup_task_struct(current
);
1034 ftrace_graph_init_task(p
);
1036 rt_mutex_init_task(p
);
1038 #ifdef CONFIG_PROVE_LOCKING
1039 DEBUG_LOCKS_WARN_ON(!p
->hardirqs_enabled
);
1040 DEBUG_LOCKS_WARN_ON(!p
->softirqs_enabled
);
1043 if (atomic_read(&p
->real_cred
->user
->processes
) >=
1044 p
->signal
->rlim
[RLIMIT_NPROC
].rlim_cur
) {
1045 if (!capable(CAP_SYS_ADMIN
) && !capable(CAP_SYS_RESOURCE
) &&
1046 p
->real_cred
->user
!= INIT_USER
)
1050 retval
= copy_creds(p
, clone_flags
);
1055 * If multiple threads are within copy_process(), then this check
1056 * triggers too late. This doesn't hurt, the check is only there
1057 * to stop root fork bombs.
1060 if (nr_threads
>= max_threads
)
1061 goto bad_fork_cleanup_count
;
1063 if (!try_module_get(task_thread_info(p
)->exec_domain
->module
))
1064 goto bad_fork_cleanup_count
;
1067 delayacct_tsk_init(p
); /* Must remain after dup_task_struct() */
1068 copy_flags(clone_flags
, p
);
1069 INIT_LIST_HEAD(&p
->children
);
1070 INIT_LIST_HEAD(&p
->sibling
);
1071 rcu_copy_process(p
);
1072 p
->vfork_done
= NULL
;
1073 spin_lock_init(&p
->alloc_lock
);
1075 init_sigpending(&p
->pending
);
1077 p
->utime
= cputime_zero
;
1078 p
->stime
= cputime_zero
;
1079 p
->gtime
= cputime_zero
;
1080 p
->utimescaled
= cputime_zero
;
1081 p
->stimescaled
= cputime_zero
;
1082 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
1083 p
->prev_utime
= cputime_zero
;
1084 p
->prev_stime
= cputime_zero
;
1087 p
->default_timer_slack_ns
= current
->timer_slack_ns
;
1089 task_io_accounting_init(&p
->ioac
);
1090 acct_clear_integrals(p
);
1092 posix_cpu_timers_init(p
);
1094 p
->lock_depth
= -1; /* -1 = no lock */
1095 do_posix_clock_monotonic_gettime(&p
->start_time
);
1096 p
->real_start_time
= p
->start_time
;
1097 monotonic_to_bootbased(&p
->real_start_time
);
1098 p
->io_context
= NULL
;
1099 p
->audit_context
= NULL
;
1102 p
->mempolicy
= mpol_dup(p
->mempolicy
);
1103 if (IS_ERR(p
->mempolicy
)) {
1104 retval
= PTR_ERR(p
->mempolicy
);
1105 p
->mempolicy
= NULL
;
1106 goto bad_fork_cleanup_cgroup
;
1108 mpol_fix_fork_child_flag(p
);
1110 #ifdef CONFIG_TRACE_IRQFLAGS
1112 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1113 p
->hardirqs_enabled
= 1;
1115 p
->hardirqs_enabled
= 0;
1117 p
->hardirq_enable_ip
= 0;
1118 p
->hardirq_enable_event
= 0;
1119 p
->hardirq_disable_ip
= _THIS_IP_
;
1120 p
->hardirq_disable_event
= 0;
1121 p
->softirqs_enabled
= 1;
1122 p
->softirq_enable_ip
= _THIS_IP_
;
1123 p
->softirq_enable_event
= 0;
1124 p
->softirq_disable_ip
= 0;
1125 p
->softirq_disable_event
= 0;
1126 p
->hardirq_context
= 0;
1127 p
->softirq_context
= 0;
1129 #ifdef CONFIG_LOCKDEP
1130 p
->lockdep_depth
= 0; /* no locks held yet */
1131 p
->curr_chain_key
= 0;
1132 p
->lockdep_recursion
= 0;
1135 #ifdef CONFIG_DEBUG_MUTEXES
1136 p
->blocked_on
= NULL
; /* not blocked yet */
1138 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
1139 p
->memcg_batch
.do_batch
= 0;
1140 p
->memcg_batch
.memcg
= NULL
;
1145 p
->stack_start
= stack_start
;
1147 /* Perform scheduler related setup. Assign this task to a CPU. */
1148 sched_fork(p
, clone_flags
);
1150 retval
= perf_event_init_task(p
);
1152 goto bad_fork_cleanup_policy
;
1154 if ((retval
= audit_alloc(p
)))
1155 goto bad_fork_cleanup_policy
;
1156 /* copy all the process information */
1157 if ((retval
= copy_semundo(clone_flags
, p
)))
1158 goto bad_fork_cleanup_audit
;
1159 if ((retval
= copy_files(clone_flags
, p
)))
1160 goto bad_fork_cleanup_semundo
;
1161 if ((retval
= copy_fs(clone_flags
, p
)))
1162 goto bad_fork_cleanup_files
;
1163 if ((retval
= copy_sighand(clone_flags
, p
)))
1164 goto bad_fork_cleanup_fs
;
1165 if ((retval
= copy_signal(clone_flags
, p
)))
1166 goto bad_fork_cleanup_sighand
;
1167 if ((retval
= copy_mm(clone_flags
, p
)))
1168 goto bad_fork_cleanup_signal
;
1169 if ((retval
= copy_namespaces(clone_flags
, p
)))
1170 goto bad_fork_cleanup_mm
;
1171 if ((retval
= copy_io(clone_flags
, p
)))
1172 goto bad_fork_cleanup_namespaces
;
1173 retval
= copy_thread(clone_flags
, stack_start
, stack_size
, p
, regs
);
1175 goto bad_fork_cleanup_io
;
1177 if (pid
!= &init_struct_pid
) {
1179 pid
= alloc_pid(p
->nsproxy
->pid_ns
);
1181 goto bad_fork_cleanup_io
;
1183 if (clone_flags
& CLONE_NEWPID
) {
1184 retval
= pid_ns_prepare_proc(p
->nsproxy
->pid_ns
);
1186 goto bad_fork_free_pid
;
1190 p
->pid
= pid_nr(pid
);
1192 if (clone_flags
& CLONE_THREAD
)
1193 p
->tgid
= current
->tgid
;
1195 if (current
->nsproxy
!= p
->nsproxy
) {
1196 retval
= ns_cgroup_clone(p
, pid
);
1198 goto bad_fork_free_pid
;
1201 p
->set_child_tid
= (clone_flags
& CLONE_CHILD_SETTID
) ? child_tidptr
: NULL
;
1203 * Clear TID on mm_release()?
1205 p
->clear_child_tid
= (clone_flags
& CLONE_CHILD_CLEARTID
) ? child_tidptr
: NULL
;
1207 p
->robust_list
= NULL
;
1208 #ifdef CONFIG_COMPAT
1209 p
->compat_robust_list
= NULL
;
1211 INIT_LIST_HEAD(&p
->pi_state_list
);
1212 p
->pi_state_cache
= NULL
;
1215 * sigaltstack should be cleared when sharing the same VM
1217 if ((clone_flags
& (CLONE_VM
|CLONE_VFORK
)) == CLONE_VM
)
1218 p
->sas_ss_sp
= p
->sas_ss_size
= 0;
1221 * Syscall tracing and stepping should be turned off in the
1222 * child regardless of CLONE_PTRACE.
1224 user_disable_single_step(p
);
1225 clear_tsk_thread_flag(p
, TIF_SYSCALL_TRACE
);
1226 #ifdef TIF_SYSCALL_EMU
1227 clear_tsk_thread_flag(p
, TIF_SYSCALL_EMU
);
1229 clear_all_latency_tracing(p
);
1231 /* ok, now we should be set up.. */
1232 p
->exit_signal
= (clone_flags
& CLONE_THREAD
) ? -1 : (clone_flags
& CSIGNAL
);
1233 p
->pdeath_signal
= 0;
1237 * Ok, make it visible to the rest of the system.
1238 * We dont wake it up yet.
1240 p
->group_leader
= p
;
1241 INIT_LIST_HEAD(&p
->thread_group
);
1243 /* Now that the task is set up, run cgroup callbacks if
1244 * necessary. We need to run them before the task is visible
1245 * on the tasklist. */
1246 cgroup_fork_callbacks(p
);
1247 cgroup_callbacks_done
= 1;
1249 /* Need tasklist lock for parent etc handling! */
1250 write_lock_irq(&tasklist_lock
);
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 tracehook_finish_clone(p
, clone_flags
, trace
);
1289 if (thread_group_leader(p
)) {
1290 if (clone_flags
& CLONE_NEWPID
)
1291 p
->nsproxy
->pid_ns
->child_reaper
= p
;
1293 p
->signal
->leader_pid
= pid
;
1294 tty_kref_put(p
->signal
->tty
);
1295 p
->signal
->tty
= tty_kref_get(current
->signal
->tty
);
1296 attach_pid(p
, PIDTYPE_PGID
, task_pgrp(current
));
1297 attach_pid(p
, PIDTYPE_SID
, task_session(current
));
1298 list_add_tail(&p
->sibling
, &p
->real_parent
->children
);
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
:
1320 bad_fork_cleanup_namespaces
:
1321 exit_task_namespaces(p
);
1322 bad_fork_cleanup_mm
:
1325 bad_fork_cleanup_signal
:
1326 if (!(clone_flags
& CLONE_THREAD
))
1327 __cleanup_signal(p
->signal
);
1328 bad_fork_cleanup_sighand
:
1329 __cleanup_sighand(p
->sighand
);
1330 bad_fork_cleanup_fs
:
1331 exit_fs(p
); /* blocking */
1332 bad_fork_cleanup_files
:
1333 exit_files(p
); /* blocking */
1334 bad_fork_cleanup_semundo
:
1336 bad_fork_cleanup_audit
:
1338 bad_fork_cleanup_policy
:
1339 perf_event_free_task(p
);
1341 mpol_put(p
->mempolicy
);
1342 bad_fork_cleanup_cgroup
:
1344 cgroup_exit(p
, cgroup_callbacks_done
);
1345 delayacct_tsk_free(p
);
1346 module_put(task_thread_info(p
)->exec_domain
->module
);
1347 bad_fork_cleanup_count
:
1348 atomic_dec(&p
->cred
->user
->processes
);
1353 return ERR_PTR(retval
);
1356 noinline
struct pt_regs
* __cpuinit
__attribute__((weak
)) idle_regs(struct pt_regs
*regs
)
1358 memset(regs
, 0, sizeof(struct pt_regs
));
1362 struct task_struct
* __cpuinit
fork_idle(int cpu
)
1364 struct task_struct
*task
;
1365 struct pt_regs regs
;
1367 task
= copy_process(CLONE_VM
, 0, idle_regs(®s
), 0, NULL
,
1368 &init_struct_pid
, 0);
1370 init_idle(task
, cpu
);
1376 * Ok, this is the main fork-routine.
1378 * It copies the process, and if successful kick-starts
1379 * it and waits for it to finish using the VM if required.
1381 long do_fork(unsigned long clone_flags
,
1382 unsigned long stack_start
,
1383 struct pt_regs
*regs
,
1384 unsigned long stack_size
,
1385 int __user
*parent_tidptr
,
1386 int __user
*child_tidptr
)
1388 struct task_struct
*p
;
1393 * Do some preliminary argument and permissions checking before we
1394 * actually start allocating stuff
1396 if (clone_flags
& CLONE_NEWUSER
) {
1397 if (clone_flags
& CLONE_THREAD
)
1399 /* hopefully this check will go away when userns support is
1402 if (!capable(CAP_SYS_ADMIN
) || !capable(CAP_SETUID
) ||
1403 !capable(CAP_SETGID
))
1408 * We hope to recycle these flags after 2.6.26
1410 if (unlikely(clone_flags
& CLONE_STOPPED
)) {
1411 static int __read_mostly count
= 100;
1413 if (count
> 0 && printk_ratelimit()) {
1414 char comm
[TASK_COMM_LEN
];
1417 printk(KERN_INFO
"fork(): process `%s' used deprecated "
1418 "clone flags 0x%lx\n",
1419 get_task_comm(comm
, current
),
1420 clone_flags
& CLONE_STOPPED
);
1425 * When called from kernel_thread, don't do user tracing stuff.
1427 if (likely(user_mode(regs
)))
1428 trace
= tracehook_prepare_clone(clone_flags
);
1430 p
= copy_process(clone_flags
, stack_start
, regs
, stack_size
,
1431 child_tidptr
, NULL
, trace
);
1433 * Do this prior waking up the new thread - the thread pointer
1434 * might get invalid after that point, if the thread exits quickly.
1437 struct completion vfork
;
1439 trace_sched_process_fork(current
, p
);
1441 nr
= task_pid_vnr(p
);
1443 if (clone_flags
& CLONE_PARENT_SETTID
)
1444 put_user(nr
, parent_tidptr
);
1446 if (clone_flags
& CLONE_VFORK
) {
1447 p
->vfork_done
= &vfork
;
1448 init_completion(&vfork
);
1451 audit_finish_fork(p
);
1452 tracehook_report_clone(regs
, clone_flags
, nr
, p
);
1455 * We set PF_STARTING at creation in case tracing wants to
1456 * use this to distinguish a fully live task from one that
1457 * hasn't gotten to tracehook_report_clone() yet. Now we
1458 * clear it and set the child going.
1460 p
->flags
&= ~PF_STARTING
;
1462 if (unlikely(clone_flags
& CLONE_STOPPED
)) {
1464 * We'll start up with an immediate SIGSTOP.
1466 sigaddset(&p
->pending
.signal
, SIGSTOP
);
1467 set_tsk_thread_flag(p
, TIF_SIGPENDING
);
1468 __set_task_state(p
, TASK_STOPPED
);
1470 wake_up_new_task(p
, clone_flags
);
1473 tracehook_report_clone_complete(trace
, regs
,
1474 clone_flags
, nr
, p
);
1476 if (clone_flags
& CLONE_VFORK
) {
1477 freezer_do_not_count();
1478 wait_for_completion(&vfork
);
1480 tracehook_report_vfork_done(p
, nr
);
1488 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1489 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1492 static void sighand_ctor(void *data
)
1494 struct sighand_struct
*sighand
= data
;
1496 spin_lock_init(&sighand
->siglock
);
1497 init_waitqueue_head(&sighand
->signalfd_wqh
);
1500 void __init
proc_caches_init(void)
1502 sighand_cachep
= kmem_cache_create("sighand_cache",
1503 sizeof(struct sighand_struct
), 0,
1504 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_DESTROY_BY_RCU
|
1505 SLAB_NOTRACK
, sighand_ctor
);
1506 signal_cachep
= kmem_cache_create("signal_cache",
1507 sizeof(struct signal_struct
), 0,
1508 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1509 files_cachep
= kmem_cache_create("files_cache",
1510 sizeof(struct files_struct
), 0,
1511 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1512 fs_cachep
= kmem_cache_create("fs_cache",
1513 sizeof(struct fs_struct
), 0,
1514 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1515 mm_cachep
= kmem_cache_create("mm_struct",
1516 sizeof(struct mm_struct
), ARCH_MIN_MMSTRUCT_ALIGN
,
1517 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1518 vm_area_cachep
= KMEM_CACHE(vm_area_struct
, SLAB_PANIC
);
1523 * Check constraints on flags passed to the unshare system call and
1524 * force unsharing of additional process context as appropriate.
1526 static void check_unshare_flags(unsigned long *flags_ptr
)
1529 * If unsharing a thread from a thread group, must also
1532 if (*flags_ptr
& CLONE_THREAD
)
1533 *flags_ptr
|= CLONE_VM
;
1536 * If unsharing vm, must also unshare signal handlers.
1538 if (*flags_ptr
& CLONE_VM
)
1539 *flags_ptr
|= CLONE_SIGHAND
;
1542 * If unsharing signal handlers and the task was created
1543 * using CLONE_THREAD, then must unshare the thread
1545 if ((*flags_ptr
& CLONE_SIGHAND
) &&
1546 (atomic_read(¤t
->signal
->count
) > 1))
1547 *flags_ptr
|= CLONE_THREAD
;
1550 * If unsharing namespace, must also unshare filesystem information.
1552 if (*flags_ptr
& CLONE_NEWNS
)
1553 *flags_ptr
|= CLONE_FS
;
1557 * Unsharing of tasks created with CLONE_THREAD is not supported yet
1559 static int unshare_thread(unsigned long unshare_flags
)
1561 if (unshare_flags
& CLONE_THREAD
)
1568 * Unshare the filesystem structure if it is being shared
1570 static int unshare_fs(unsigned long unshare_flags
, struct fs_struct
**new_fsp
)
1572 struct fs_struct
*fs
= current
->fs
;
1574 if (!(unshare_flags
& CLONE_FS
) || !fs
)
1577 /* don't need lock here; in the worst case we'll do useless copy */
1581 *new_fsp
= copy_fs_struct(fs
);
1589 * Unsharing of sighand is not supported yet
1591 static int unshare_sighand(unsigned long unshare_flags
, struct sighand_struct
**new_sighp
)
1593 struct sighand_struct
*sigh
= current
->sighand
;
1595 if ((unshare_flags
& CLONE_SIGHAND
) && atomic_read(&sigh
->count
) > 1)
1602 * Unshare vm if it is being shared
1604 static int unshare_vm(unsigned long unshare_flags
, struct mm_struct
**new_mmp
)
1606 struct mm_struct
*mm
= current
->mm
;
1608 if ((unshare_flags
& CLONE_VM
) &&
1609 (mm
&& atomic_read(&mm
->mm_users
) > 1)) {
1617 * Unshare file descriptor table if it is being shared
1619 static int unshare_fd(unsigned long unshare_flags
, struct files_struct
**new_fdp
)
1621 struct files_struct
*fd
= current
->files
;
1624 if ((unshare_flags
& CLONE_FILES
) &&
1625 (fd
&& atomic_read(&fd
->count
) > 1)) {
1626 *new_fdp
= dup_fd(fd
, &error
);
1635 * unshare allows a process to 'unshare' part of the process
1636 * context which was originally shared using clone. copy_*
1637 * functions used by do_fork() cannot be used here directly
1638 * because they modify an inactive task_struct that is being
1639 * constructed. Here we are modifying the current, active,
1642 SYSCALL_DEFINE1(unshare
, unsigned long, unshare_flags
)
1645 struct fs_struct
*fs
, *new_fs
= NULL
;
1646 struct sighand_struct
*new_sigh
= NULL
;
1647 struct mm_struct
*mm
, *new_mm
= NULL
, *active_mm
= NULL
;
1648 struct files_struct
*fd
, *new_fd
= NULL
;
1649 struct nsproxy
*new_nsproxy
= NULL
;
1652 check_unshare_flags(&unshare_flags
);
1654 /* Return -EINVAL for all unsupported flags */
1656 if (unshare_flags
& ~(CLONE_THREAD
|CLONE_FS
|CLONE_NEWNS
|CLONE_SIGHAND
|
1657 CLONE_VM
|CLONE_FILES
|CLONE_SYSVSEM
|
1658 CLONE_NEWUTS
|CLONE_NEWIPC
|CLONE_NEWNET
))
1659 goto bad_unshare_out
;
1662 * CLONE_NEWIPC must also detach from the undolist: after switching
1663 * to a new ipc namespace, the semaphore arrays from the old
1664 * namespace are unreachable.
1666 if (unshare_flags
& (CLONE_NEWIPC
|CLONE_SYSVSEM
))
1668 if ((err
= unshare_thread(unshare_flags
)))
1669 goto bad_unshare_out
;
1670 if ((err
= unshare_fs(unshare_flags
, &new_fs
)))
1671 goto bad_unshare_cleanup_thread
;
1672 if ((err
= unshare_sighand(unshare_flags
, &new_sigh
)))
1673 goto bad_unshare_cleanup_fs
;
1674 if ((err
= unshare_vm(unshare_flags
, &new_mm
)))
1675 goto bad_unshare_cleanup_sigh
;
1676 if ((err
= unshare_fd(unshare_flags
, &new_fd
)))
1677 goto bad_unshare_cleanup_vm
;
1678 if ((err
= unshare_nsproxy_namespaces(unshare_flags
, &new_nsproxy
,
1680 goto bad_unshare_cleanup_fd
;
1682 if (new_fs
|| new_mm
|| new_fd
|| do_sysvsem
|| new_nsproxy
) {
1685 * CLONE_SYSVSEM is equivalent to sys_exit().
1691 switch_task_namespaces(current
, new_nsproxy
);
1699 write_lock(&fs
->lock
);
1700 current
->fs
= new_fs
;
1705 write_unlock(&fs
->lock
);
1710 active_mm
= current
->active_mm
;
1711 current
->mm
= new_mm
;
1712 current
->active_mm
= new_mm
;
1713 activate_mm(active_mm
, new_mm
);
1718 fd
= current
->files
;
1719 current
->files
= new_fd
;
1723 task_unlock(current
);
1727 put_nsproxy(new_nsproxy
);
1729 bad_unshare_cleanup_fd
:
1731 put_files_struct(new_fd
);
1733 bad_unshare_cleanup_vm
:
1737 bad_unshare_cleanup_sigh
:
1739 if (atomic_dec_and_test(&new_sigh
->count
))
1740 kmem_cache_free(sighand_cachep
, new_sigh
);
1742 bad_unshare_cleanup_fs
:
1744 free_fs_struct(new_fs
);
1746 bad_unshare_cleanup_thread
:
1752 * Helper to unshare the files of the current task.
1753 * We don't want to expose copy_files internals to
1754 * the exec layer of the kernel.
1757 int unshare_files(struct files_struct
**displaced
)
1759 struct task_struct
*task
= current
;
1760 struct files_struct
*copy
= NULL
;
1763 error
= unshare_fd(CLONE_FILES
, ©
);
1764 if (error
|| !copy
) {
1768 *displaced
= task
->files
;