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/futex.h>
41 #include <linux/compat.h>
42 #include <linux/kthread.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/blkdev.h>
62 #include <linux/fs_struct.h>
63 #include <linux/magic.h>
64 #include <linux/perf_event.h>
65 #include <linux/posix-timers.h>
66 #include <linux/user-return-notifier.h>
67 #include <linux/oom.h>
68 #include <linux/khugepaged.h>
69 #include <linux/signalfd.h>
71 #include <asm/pgtable.h>
72 #include <asm/pgalloc.h>
73 #include <asm/uaccess.h>
74 #include <asm/mmu_context.h>
75 #include <asm/cacheflush.h>
76 #include <asm/tlbflush.h>
78 #include <trace/events/sched.h>
80 #define CREATE_TRACE_POINTS
81 #include <trace/events/task.h>
84 * Protected counters by write_lock_irq(&tasklist_lock)
86 unsigned long total_forks
; /* Handle normal Linux uptimes. */
87 int nr_threads
; /* The idle threads do not count.. */
89 int max_threads
; /* tunable limit on nr_threads */
91 DEFINE_PER_CPU(unsigned long, process_counts
) = 0;
93 __cacheline_aligned
DEFINE_RWLOCK(tasklist_lock
); /* outer */
95 #ifdef CONFIG_PROVE_RCU
96 int lockdep_tasklist_lock_is_held(void)
98 return lockdep_is_held(&tasklist_lock
);
100 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held
);
101 #endif /* #ifdef CONFIG_PROVE_RCU */
103 int nr_processes(void)
108 for_each_possible_cpu(cpu
)
109 total
+= per_cpu(process_counts
, cpu
);
114 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
115 # define alloc_task_struct_node(node) \
116 kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node)
117 # define free_task_struct(tsk) \
118 kmem_cache_free(task_struct_cachep, (tsk))
119 static struct kmem_cache
*task_struct_cachep
;
122 #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
123 static struct thread_info
*alloc_thread_info_node(struct task_struct
*tsk
,
126 #ifdef CONFIG_DEBUG_STACK_USAGE
127 gfp_t mask
= GFP_KERNEL
| __GFP_ZERO
;
129 gfp_t mask
= GFP_KERNEL
;
131 struct page
*page
= alloc_pages_node(node
, mask
, THREAD_SIZE_ORDER
);
133 return page
? page_address(page
) : NULL
;
136 static inline void free_thread_info(struct thread_info
*ti
)
138 free_pages((unsigned long)ti
, THREAD_SIZE_ORDER
);
142 /* SLAB cache for signal_struct structures (tsk->signal) */
143 static struct kmem_cache
*signal_cachep
;
145 /* SLAB cache for sighand_struct structures (tsk->sighand) */
146 struct kmem_cache
*sighand_cachep
;
148 /* SLAB cache for files_struct structures (tsk->files) */
149 struct kmem_cache
*files_cachep
;
151 /* SLAB cache for fs_struct structures (tsk->fs) */
152 struct kmem_cache
*fs_cachep
;
154 /* SLAB cache for vm_area_struct structures */
155 struct kmem_cache
*vm_area_cachep
;
157 /* SLAB cache for mm_struct structures (tsk->mm) */
158 static struct kmem_cache
*mm_cachep
;
160 static void account_kernel_stack(struct thread_info
*ti
, int account
)
162 struct zone
*zone
= page_zone(virt_to_page(ti
));
164 mod_zone_page_state(zone
, NR_KERNEL_STACK
, account
);
167 void free_task(struct task_struct
*tsk
)
169 account_kernel_stack(tsk
->stack
, -1);
170 free_thread_info(tsk
->stack
);
171 rt_mutex_debug_task_free(tsk
);
172 ftrace_graph_exit_task(tsk
);
173 free_task_struct(tsk
);
175 EXPORT_SYMBOL(free_task
);
177 static inline void free_signal_struct(struct signal_struct
*sig
)
179 taskstats_tgid_free(sig
);
180 sched_autogroup_exit(sig
);
181 kmem_cache_free(signal_cachep
, sig
);
184 static inline void put_signal_struct(struct signal_struct
*sig
)
186 if (atomic_dec_and_test(&sig
->sigcnt
))
187 free_signal_struct(sig
);
190 void __put_task_struct(struct task_struct
*tsk
)
192 WARN_ON(!tsk
->exit_state
);
193 WARN_ON(atomic_read(&tsk
->usage
));
194 WARN_ON(tsk
== current
);
197 delayacct_tsk_free(tsk
);
198 put_signal_struct(tsk
->signal
);
200 if (!profile_handoff_task(tsk
))
203 EXPORT_SYMBOL_GPL(__put_task_struct
);
206 * macro override instead of weak attribute alias, to workaround
207 * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
209 #ifndef arch_task_cache_init
210 #define arch_task_cache_init()
213 void __init
fork_init(unsigned long mempages
)
215 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
216 #ifndef ARCH_MIN_TASKALIGN
217 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
219 /* create a slab on which task_structs can be allocated */
221 kmem_cache_create("task_struct", sizeof(struct task_struct
),
222 ARCH_MIN_TASKALIGN
, SLAB_PANIC
| SLAB_NOTRACK
, NULL
);
225 /* do the arch specific task caches init */
226 arch_task_cache_init();
229 * The default maximum number of threads is set to a safe
230 * value: the thread structures can take up at most half
233 max_threads
= mempages
/ (8 * THREAD_SIZE
/ PAGE_SIZE
);
236 * we need to allow at least 20 threads to boot a system
238 if (max_threads
< 20)
241 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_cur
= max_threads
/2;
242 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_max
= max_threads
/2;
243 init_task
.signal
->rlim
[RLIMIT_SIGPENDING
] =
244 init_task
.signal
->rlim
[RLIMIT_NPROC
];
247 int __attribute__((weak
)) arch_dup_task_struct(struct task_struct
*dst
,
248 struct task_struct
*src
)
254 static struct task_struct
*dup_task_struct(struct task_struct
*orig
)
256 struct task_struct
*tsk
;
257 struct thread_info
*ti
;
258 unsigned long *stackend
;
259 int node
= tsk_fork_get_node(orig
);
262 prepare_to_copy(orig
);
264 tsk
= alloc_task_struct_node(node
);
268 ti
= alloc_thread_info_node(tsk
, node
);
270 free_task_struct(tsk
);
274 err
= arch_dup_task_struct(tsk
, orig
);
280 setup_thread_stack(tsk
, orig
);
281 clear_user_return_notifier(tsk
);
282 clear_tsk_need_resched(tsk
);
283 stackend
= end_of_stack(tsk
);
284 *stackend
= STACK_END_MAGIC
; /* for overflow detection */
286 #ifdef CONFIG_CC_STACKPROTECTOR
287 tsk
->stack_canary
= get_random_int();
291 * One for us, one for whoever does the "release_task()" (usually
294 atomic_set(&tsk
->usage
, 2);
295 #ifdef CONFIG_BLK_DEV_IO_TRACE
298 tsk
->splice_pipe
= NULL
;
300 account_kernel_stack(ti
, 1);
305 free_thread_info(ti
);
306 free_task_struct(tsk
);
311 static int dup_mmap(struct mm_struct
*mm
, struct mm_struct
*oldmm
)
313 struct vm_area_struct
*mpnt
, *tmp
, *prev
, **pprev
;
314 struct rb_node
**rb_link
, *rb_parent
;
316 unsigned long charge
;
317 struct mempolicy
*pol
;
319 down_write(&oldmm
->mmap_sem
);
320 flush_cache_dup_mm(oldmm
);
322 * Not linked in yet - no deadlock potential:
324 down_write_nested(&mm
->mmap_sem
, SINGLE_DEPTH_NESTING
);
328 mm
->mmap_cache
= NULL
;
329 mm
->free_area_cache
= oldmm
->mmap_base
;
330 mm
->cached_hole_size
= ~0UL;
332 cpumask_clear(mm_cpumask(mm
));
334 rb_link
= &mm
->mm_rb
.rb_node
;
337 retval
= ksm_fork(mm
, oldmm
);
340 retval
= khugepaged_fork(mm
, oldmm
);
345 for (mpnt
= oldmm
->mmap
; mpnt
; mpnt
= mpnt
->vm_next
) {
348 if (mpnt
->vm_flags
& VM_DONTCOPY
) {
349 long pages
= vma_pages(mpnt
);
350 mm
->total_vm
-= pages
;
351 vm_stat_account(mm
, mpnt
->vm_flags
, mpnt
->vm_file
,
356 if (mpnt
->vm_flags
& VM_ACCOUNT
) {
357 unsigned int len
= (mpnt
->vm_end
- mpnt
->vm_start
) >> PAGE_SHIFT
;
358 if (security_vm_enough_memory(len
))
362 tmp
= kmem_cache_alloc(vm_area_cachep
, GFP_KERNEL
);
366 INIT_LIST_HEAD(&tmp
->anon_vma_chain
);
367 pol
= mpol_dup(vma_policy(mpnt
));
368 retval
= PTR_ERR(pol
);
370 goto fail_nomem_policy
;
371 vma_set_policy(tmp
, pol
);
373 if (anon_vma_fork(tmp
, mpnt
))
374 goto fail_nomem_anon_vma_fork
;
375 tmp
->vm_flags
&= ~VM_LOCKED
;
376 tmp
->vm_next
= tmp
->vm_prev
= NULL
;
379 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
380 struct address_space
*mapping
= file
->f_mapping
;
383 if (tmp
->vm_flags
& VM_DENYWRITE
)
384 atomic_dec(&inode
->i_writecount
);
385 mutex_lock(&mapping
->i_mmap_mutex
);
386 if (tmp
->vm_flags
& VM_SHARED
)
387 mapping
->i_mmap_writable
++;
388 flush_dcache_mmap_lock(mapping
);
389 /* insert tmp into the share list, just after mpnt */
390 vma_prio_tree_add(tmp
, mpnt
);
391 flush_dcache_mmap_unlock(mapping
);
392 mutex_unlock(&mapping
->i_mmap_mutex
);
396 * Clear hugetlb-related page reserves for children. This only
397 * affects MAP_PRIVATE mappings. Faults generated by the child
398 * are not guaranteed to succeed, even if read-only
400 if (is_vm_hugetlb_page(tmp
))
401 reset_vma_resv_huge_pages(tmp
);
404 * Link in the new vma and copy the page table entries.
407 pprev
= &tmp
->vm_next
;
411 __vma_link_rb(mm
, tmp
, rb_link
, rb_parent
);
412 rb_link
= &tmp
->vm_rb
.rb_right
;
413 rb_parent
= &tmp
->vm_rb
;
416 retval
= copy_page_range(mm
, oldmm
, mpnt
);
418 if (tmp
->vm_ops
&& tmp
->vm_ops
->open
)
419 tmp
->vm_ops
->open(tmp
);
424 /* a new mm has just been created */
425 arch_dup_mmap(oldmm
, mm
);
428 up_write(&mm
->mmap_sem
);
430 up_write(&oldmm
->mmap_sem
);
432 fail_nomem_anon_vma_fork
:
435 kmem_cache_free(vm_area_cachep
, tmp
);
438 vm_unacct_memory(charge
);
442 static inline int mm_alloc_pgd(struct mm_struct
*mm
)
444 mm
->pgd
= pgd_alloc(mm
);
445 if (unlikely(!mm
->pgd
))
450 static inline void mm_free_pgd(struct mm_struct
*mm
)
452 pgd_free(mm
, mm
->pgd
);
455 #define dup_mmap(mm, oldmm) (0)
456 #define mm_alloc_pgd(mm) (0)
457 #define mm_free_pgd(mm)
458 #endif /* CONFIG_MMU */
460 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(mmlist_lock
);
462 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
463 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
465 static unsigned long default_dump_filter
= MMF_DUMP_FILTER_DEFAULT
;
467 static int __init
coredump_filter_setup(char *s
)
469 default_dump_filter
=
470 (simple_strtoul(s
, NULL
, 0) << MMF_DUMP_FILTER_SHIFT
) &
471 MMF_DUMP_FILTER_MASK
;
475 __setup("coredump_filter=", coredump_filter_setup
);
477 #include <linux/init_task.h>
479 static void mm_init_aio(struct mm_struct
*mm
)
482 spin_lock_init(&mm
->ioctx_lock
);
483 INIT_HLIST_HEAD(&mm
->ioctx_list
);
487 static struct mm_struct
*mm_init(struct mm_struct
*mm
, struct task_struct
*p
)
489 atomic_set(&mm
->mm_users
, 1);
490 atomic_set(&mm
->mm_count
, 1);
491 init_rwsem(&mm
->mmap_sem
);
492 INIT_LIST_HEAD(&mm
->mmlist
);
493 mm
->flags
= (current
->mm
) ?
494 (current
->mm
->flags
& MMF_INIT_MASK
) : default_dump_filter
;
495 mm
->core_state
= NULL
;
497 memset(&mm
->rss_stat
, 0, sizeof(mm
->rss_stat
));
498 spin_lock_init(&mm
->page_table_lock
);
499 mm
->free_area_cache
= TASK_UNMAPPED_BASE
;
500 mm
->cached_hole_size
= ~0UL;
502 mm_init_owner(mm
, p
);
504 if (likely(!mm_alloc_pgd(mm
))) {
506 mmu_notifier_mm_init(mm
);
514 static void check_mm(struct mm_struct
*mm
)
518 for (i
= 0; i
< NR_MM_COUNTERS
; i
++) {
519 long x
= atomic_long_read(&mm
->rss_stat
.count
[i
]);
522 printk(KERN_ALERT
"BUG: Bad rss-counter state "
523 "mm:%p idx:%d val:%ld\n", mm
, i
, x
);
526 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
527 VM_BUG_ON(mm
->pmd_huge_pte
);
532 * Allocate and initialize an mm_struct.
534 struct mm_struct
*mm_alloc(void)
536 struct mm_struct
*mm
;
542 memset(mm
, 0, sizeof(*mm
));
544 return mm_init(mm
, current
);
548 * Called when the last reference to the mm
549 * is dropped: either by a lazy thread or by
550 * mmput. Free the page directory and the mm.
552 void __mmdrop(struct mm_struct
*mm
)
554 BUG_ON(mm
== &init_mm
);
557 mmu_notifier_mm_destroy(mm
);
561 EXPORT_SYMBOL_GPL(__mmdrop
);
564 * Decrement the use count and release all resources for an mm.
566 void mmput(struct mm_struct
*mm
)
570 if (atomic_dec_and_test(&mm
->mm_users
)) {
573 khugepaged_exit(mm
); /* must run before exit_mmap */
575 set_mm_exe_file(mm
, NULL
);
576 if (!list_empty(&mm
->mmlist
)) {
577 spin_lock(&mmlist_lock
);
578 list_del(&mm
->mmlist
);
579 spin_unlock(&mmlist_lock
);
583 module_put(mm
->binfmt
->module
);
587 EXPORT_SYMBOL_GPL(mmput
);
590 * We added or removed a vma mapping the executable. The vmas are only mapped
591 * during exec and are not mapped with the mmap system call.
592 * Callers must hold down_write() on the mm's mmap_sem for these
594 void added_exe_file_vma(struct mm_struct
*mm
)
596 mm
->num_exe_file_vmas
++;
599 void removed_exe_file_vma(struct mm_struct
*mm
)
601 mm
->num_exe_file_vmas
--;
602 if ((mm
->num_exe_file_vmas
== 0) && mm
->exe_file
) {
609 void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
)
612 get_file(new_exe_file
);
615 mm
->exe_file
= new_exe_file
;
616 mm
->num_exe_file_vmas
= 0;
619 struct file
*get_mm_exe_file(struct mm_struct
*mm
)
621 struct file
*exe_file
;
623 /* We need mmap_sem to protect against races with removal of
624 * VM_EXECUTABLE vmas */
625 down_read(&mm
->mmap_sem
);
626 exe_file
= mm
->exe_file
;
629 up_read(&mm
->mmap_sem
);
633 static void dup_mm_exe_file(struct mm_struct
*oldmm
, struct mm_struct
*newmm
)
635 /* It's safe to write the exe_file pointer without exe_file_lock because
636 * this is called during fork when the task is not yet in /proc */
637 newmm
->exe_file
= get_mm_exe_file(oldmm
);
641 * get_task_mm - acquire a reference to the task's mm
643 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
644 * this kernel workthread has transiently adopted a user mm with use_mm,
645 * to do its AIO) is not set and if so returns a reference to it, after
646 * bumping up the use count. User must release the mm via mmput()
647 * after use. Typically used by /proc and ptrace.
649 struct mm_struct
*get_task_mm(struct task_struct
*task
)
651 struct mm_struct
*mm
;
656 if (task
->flags
& PF_KTHREAD
)
659 atomic_inc(&mm
->mm_users
);
664 EXPORT_SYMBOL_GPL(get_task_mm
);
666 struct mm_struct
*mm_access(struct task_struct
*task
, unsigned int mode
)
668 struct mm_struct
*mm
;
671 err
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
675 mm
= get_task_mm(task
);
676 if (mm
&& mm
!= current
->mm
&&
677 !ptrace_may_access(task
, mode
)) {
679 mm
= ERR_PTR(-EACCES
);
681 mutex_unlock(&task
->signal
->cred_guard_mutex
);
686 static void complete_vfork_done(struct task_struct
*tsk
)
688 struct completion
*vfork
;
691 vfork
= tsk
->vfork_done
;
693 tsk
->vfork_done
= NULL
;
699 static int wait_for_vfork_done(struct task_struct
*child
,
700 struct completion
*vfork
)
704 freezer_do_not_count();
705 killed
= wait_for_completion_killable(vfork
);
710 child
->vfork_done
= NULL
;
714 put_task_struct(child
);
718 /* Please note the differences between mmput and mm_release.
719 * mmput is called whenever we stop holding onto a mm_struct,
720 * error success whatever.
722 * mm_release is called after a mm_struct has been removed
723 * from the current process.
725 * This difference is important for error handling, when we
726 * only half set up a mm_struct for a new process and need to restore
727 * the old one. Because we mmput the new mm_struct before
728 * restoring the old one. . .
729 * Eric Biederman 10 January 1998
731 void mm_release(struct task_struct
*tsk
, struct mm_struct
*mm
)
733 /* Get rid of any futexes when releasing the mm */
735 if (unlikely(tsk
->robust_list
)) {
736 exit_robust_list(tsk
);
737 tsk
->robust_list
= NULL
;
740 if (unlikely(tsk
->compat_robust_list
)) {
741 compat_exit_robust_list(tsk
);
742 tsk
->compat_robust_list
= NULL
;
745 if (unlikely(!list_empty(&tsk
->pi_state_list
)))
746 exit_pi_state_list(tsk
);
749 /* Get rid of any cached register state */
750 deactivate_mm(tsk
, mm
);
753 complete_vfork_done(tsk
);
756 * If we're exiting normally, clear a user-space tid field if
757 * requested. We leave this alone when dying by signal, to leave
758 * the value intact in a core dump, and to save the unnecessary
759 * trouble, say, a killed vfork parent shouldn't touch this mm.
760 * Userland only wants this done for a sys_exit.
762 if (tsk
->clear_child_tid
) {
763 if (!(tsk
->flags
& PF_SIGNALED
) &&
764 atomic_read(&mm
->mm_users
) > 1) {
766 * We don't check the error code - if userspace has
767 * not set up a proper pointer then tough luck.
769 put_user(0, tsk
->clear_child_tid
);
770 sys_futex(tsk
->clear_child_tid
, FUTEX_WAKE
,
773 tsk
->clear_child_tid
= NULL
;
778 * Allocate a new mm structure and copy contents from the
779 * mm structure of the passed in task structure.
781 struct mm_struct
*dup_mm(struct task_struct
*tsk
)
783 struct mm_struct
*mm
, *oldmm
= current
->mm
;
793 memcpy(mm
, oldmm
, sizeof(*mm
));
796 /* Initializing for Swap token stuff */
797 mm
->token_priority
= 0;
798 mm
->last_interval
= 0;
800 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
801 mm
->pmd_huge_pte
= NULL
;
804 if (!mm_init(mm
, tsk
))
807 if (init_new_context(tsk
, mm
))
810 dup_mm_exe_file(oldmm
, mm
);
812 err
= dup_mmap(mm
, oldmm
);
816 mm
->hiwater_rss
= get_mm_rss(mm
);
817 mm
->hiwater_vm
= mm
->total_vm
;
819 if (mm
->binfmt
&& !try_module_get(mm
->binfmt
->module
))
825 /* don't put binfmt in mmput, we haven't got module yet */
834 * If init_new_context() failed, we cannot use mmput() to free the mm
835 * because it calls destroy_context()
842 static int copy_mm(unsigned long clone_flags
, struct task_struct
*tsk
)
844 struct mm_struct
*mm
, *oldmm
;
847 tsk
->min_flt
= tsk
->maj_flt
= 0;
848 tsk
->nvcsw
= tsk
->nivcsw
= 0;
849 #ifdef CONFIG_DETECT_HUNG_TASK
850 tsk
->last_switch_count
= tsk
->nvcsw
+ tsk
->nivcsw
;
854 tsk
->active_mm
= NULL
;
857 * Are we cloning a kernel thread?
859 * We need to steal a active VM for that..
865 if (clone_flags
& CLONE_VM
) {
866 atomic_inc(&oldmm
->mm_users
);
877 /* Initializing for Swap token stuff */
878 mm
->token_priority
= 0;
879 mm
->last_interval
= 0;
889 static int copy_fs(unsigned long clone_flags
, struct task_struct
*tsk
)
891 struct fs_struct
*fs
= current
->fs
;
892 if (clone_flags
& CLONE_FS
) {
893 /* tsk->fs is already what we want */
894 spin_lock(&fs
->lock
);
896 spin_unlock(&fs
->lock
);
900 spin_unlock(&fs
->lock
);
903 tsk
->fs
= copy_fs_struct(fs
);
909 static int copy_files(unsigned long clone_flags
, struct task_struct
*tsk
)
911 struct files_struct
*oldf
, *newf
;
915 * A background process may not have any files ...
917 oldf
= current
->files
;
921 if (clone_flags
& CLONE_FILES
) {
922 atomic_inc(&oldf
->count
);
926 newf
= dup_fd(oldf
, &error
);
936 static int copy_io(unsigned long clone_flags
, struct task_struct
*tsk
)
939 struct io_context
*ioc
= current
->io_context
;
940 struct io_context
*new_ioc
;
945 * Share io context with parent, if CLONE_IO is set
947 if (clone_flags
& CLONE_IO
) {
948 tsk
->io_context
= ioc_task_link(ioc
);
949 if (unlikely(!tsk
->io_context
))
951 } else if (ioprio_valid(ioc
->ioprio
)) {
952 new_ioc
= get_task_io_context(tsk
, GFP_KERNEL
, NUMA_NO_NODE
);
953 if (unlikely(!new_ioc
))
956 new_ioc
->ioprio
= ioc
->ioprio
;
957 put_io_context(new_ioc
);
963 static int copy_sighand(unsigned long clone_flags
, struct task_struct
*tsk
)
965 struct sighand_struct
*sig
;
967 if (clone_flags
& CLONE_SIGHAND
) {
968 atomic_inc(¤t
->sighand
->count
);
971 sig
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
972 rcu_assign_pointer(tsk
->sighand
, sig
);
975 atomic_set(&sig
->count
, 1);
976 memcpy(sig
->action
, current
->sighand
->action
, sizeof(sig
->action
));
980 void __cleanup_sighand(struct sighand_struct
*sighand
)
982 if (atomic_dec_and_test(&sighand
->count
)) {
983 signalfd_cleanup(sighand
);
984 kmem_cache_free(sighand_cachep
, sighand
);
990 * Initialize POSIX timer handling for a thread group.
992 static void posix_cpu_timers_init_group(struct signal_struct
*sig
)
994 unsigned long cpu_limit
;
996 /* Thread group counters. */
997 thread_group_cputime_init(sig
);
999 cpu_limit
= ACCESS_ONCE(sig
->rlim
[RLIMIT_CPU
].rlim_cur
);
1000 if (cpu_limit
!= RLIM_INFINITY
) {
1001 sig
->cputime_expires
.prof_exp
= secs_to_cputime(cpu_limit
);
1002 sig
->cputimer
.running
= 1;
1005 /* The timer lists. */
1006 INIT_LIST_HEAD(&sig
->cpu_timers
[0]);
1007 INIT_LIST_HEAD(&sig
->cpu_timers
[1]);
1008 INIT_LIST_HEAD(&sig
->cpu_timers
[2]);
1011 static int copy_signal(unsigned long clone_flags
, struct task_struct
*tsk
)
1013 struct signal_struct
*sig
;
1015 if (clone_flags
& CLONE_THREAD
)
1018 sig
= kmem_cache_zalloc(signal_cachep
, GFP_KERNEL
);
1023 sig
->nr_threads
= 1;
1024 atomic_set(&sig
->live
, 1);
1025 atomic_set(&sig
->sigcnt
, 1);
1026 init_waitqueue_head(&sig
->wait_chldexit
);
1027 if (clone_flags
& CLONE_NEWPID
)
1028 sig
->flags
|= SIGNAL_UNKILLABLE
;
1029 sig
->curr_target
= tsk
;
1030 init_sigpending(&sig
->shared_pending
);
1031 INIT_LIST_HEAD(&sig
->posix_timers
);
1033 hrtimer_init(&sig
->real_timer
, CLOCK_MONOTONIC
, HRTIMER_MODE_REL
);
1034 sig
->real_timer
.function
= it_real_fn
;
1036 task_lock(current
->group_leader
);
1037 memcpy(sig
->rlim
, current
->signal
->rlim
, sizeof sig
->rlim
);
1038 task_unlock(current
->group_leader
);
1040 posix_cpu_timers_init_group(sig
);
1042 tty_audit_fork(sig
);
1043 sched_autogroup_fork(sig
);
1045 #ifdef CONFIG_CGROUPS
1046 init_rwsem(&sig
->group_rwsem
);
1049 sig
->oom_adj
= current
->signal
->oom_adj
;
1050 sig
->oom_score_adj
= current
->signal
->oom_score_adj
;
1051 sig
->oom_score_adj_min
= current
->signal
->oom_score_adj_min
;
1053 mutex_init(&sig
->cred_guard_mutex
);
1058 static void copy_flags(unsigned long clone_flags
, struct task_struct
*p
)
1060 unsigned long new_flags
= p
->flags
;
1062 new_flags
&= ~(PF_SUPERPRIV
| PF_WQ_WORKER
);
1063 new_flags
|= PF_FORKNOEXEC
;
1064 p
->flags
= new_flags
;
1067 SYSCALL_DEFINE1(set_tid_address
, int __user
*, tidptr
)
1069 current
->clear_child_tid
= tidptr
;
1071 return task_pid_vnr(current
);
1074 static void rt_mutex_init_task(struct task_struct
*p
)
1076 raw_spin_lock_init(&p
->pi_lock
);
1077 #ifdef CONFIG_RT_MUTEXES
1078 plist_head_init(&p
->pi_waiters
);
1079 p
->pi_blocked_on
= NULL
;
1083 #ifdef CONFIG_MM_OWNER
1084 void mm_init_owner(struct mm_struct
*mm
, struct task_struct
*p
)
1088 #endif /* CONFIG_MM_OWNER */
1091 * Initialize POSIX timer handling for a single task.
1093 static void posix_cpu_timers_init(struct task_struct
*tsk
)
1095 tsk
->cputime_expires
.prof_exp
= 0;
1096 tsk
->cputime_expires
.virt_exp
= 0;
1097 tsk
->cputime_expires
.sched_exp
= 0;
1098 INIT_LIST_HEAD(&tsk
->cpu_timers
[0]);
1099 INIT_LIST_HEAD(&tsk
->cpu_timers
[1]);
1100 INIT_LIST_HEAD(&tsk
->cpu_timers
[2]);
1104 * This creates a new process as a copy of the old one,
1105 * but does not actually start it yet.
1107 * It copies the registers, and all the appropriate
1108 * parts of the process environment (as per the clone
1109 * flags). The actual kick-off is left to the caller.
1111 static struct task_struct
*copy_process(unsigned long clone_flags
,
1112 unsigned long stack_start
,
1113 struct pt_regs
*regs
,
1114 unsigned long stack_size
,
1115 int __user
*child_tidptr
,
1120 struct task_struct
*p
;
1121 int cgroup_callbacks_done
= 0;
1123 if ((clone_flags
& (CLONE_NEWNS
|CLONE_FS
)) == (CLONE_NEWNS
|CLONE_FS
))
1124 return ERR_PTR(-EINVAL
);
1127 * Thread groups must share signals as well, and detached threads
1128 * can only be started up within the thread group.
1130 if ((clone_flags
& CLONE_THREAD
) && !(clone_flags
& CLONE_SIGHAND
))
1131 return ERR_PTR(-EINVAL
);
1134 * Shared signal handlers imply shared VM. By way of the above,
1135 * thread groups also imply shared VM. Blocking this case allows
1136 * for various simplifications in other code.
1138 if ((clone_flags
& CLONE_SIGHAND
) && !(clone_flags
& CLONE_VM
))
1139 return ERR_PTR(-EINVAL
);
1142 * Siblings of global init remain as zombies on exit since they are
1143 * not reaped by their parent (swapper). To solve this and to avoid
1144 * multi-rooted process trees, prevent global and container-inits
1145 * from creating siblings.
1147 if ((clone_flags
& CLONE_PARENT
) &&
1148 current
->signal
->flags
& SIGNAL_UNKILLABLE
)
1149 return ERR_PTR(-EINVAL
);
1151 retval
= security_task_create(clone_flags
);
1156 p
= dup_task_struct(current
);
1160 ftrace_graph_init_task(p
);
1162 rt_mutex_init_task(p
);
1164 #ifdef CONFIG_PROVE_LOCKING
1165 DEBUG_LOCKS_WARN_ON(!p
->hardirqs_enabled
);
1166 DEBUG_LOCKS_WARN_ON(!p
->softirqs_enabled
);
1169 if (atomic_read(&p
->real_cred
->user
->processes
) >=
1170 task_rlimit(p
, RLIMIT_NPROC
)) {
1171 if (!capable(CAP_SYS_ADMIN
) && !capable(CAP_SYS_RESOURCE
) &&
1172 p
->real_cred
->user
!= INIT_USER
)
1175 current
->flags
&= ~PF_NPROC_EXCEEDED
;
1177 retval
= copy_creds(p
, clone_flags
);
1182 * If multiple threads are within copy_process(), then this check
1183 * triggers too late. This doesn't hurt, the check is only there
1184 * to stop root fork bombs.
1187 if (nr_threads
>= max_threads
)
1188 goto bad_fork_cleanup_count
;
1190 if (!try_module_get(task_thread_info(p
)->exec_domain
->module
))
1191 goto bad_fork_cleanup_count
;
1194 delayacct_tsk_init(p
); /* Must remain after dup_task_struct() */
1195 copy_flags(clone_flags
, p
);
1196 INIT_LIST_HEAD(&p
->children
);
1197 INIT_LIST_HEAD(&p
->sibling
);
1198 rcu_copy_process(p
);
1199 p
->vfork_done
= NULL
;
1200 spin_lock_init(&p
->alloc_lock
);
1202 init_sigpending(&p
->pending
);
1204 p
->utime
= p
->stime
= p
->gtime
= 0;
1205 p
->utimescaled
= p
->stimescaled
= 0;
1206 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
1207 p
->prev_utime
= p
->prev_stime
= 0;
1209 #if defined(SPLIT_RSS_COUNTING)
1210 memset(&p
->rss_stat
, 0, sizeof(p
->rss_stat
));
1213 p
->default_timer_slack_ns
= current
->timer_slack_ns
;
1215 task_io_accounting_init(&p
->ioac
);
1216 acct_clear_integrals(p
);
1218 posix_cpu_timers_init(p
);
1220 do_posix_clock_monotonic_gettime(&p
->start_time
);
1221 p
->real_start_time
= p
->start_time
;
1222 monotonic_to_bootbased(&p
->real_start_time
);
1223 p
->io_context
= NULL
;
1224 p
->audit_context
= NULL
;
1225 if (clone_flags
& CLONE_THREAD
)
1226 threadgroup_change_begin(current
);
1229 p
->mempolicy
= mpol_dup(p
->mempolicy
);
1230 if (IS_ERR(p
->mempolicy
)) {
1231 retval
= PTR_ERR(p
->mempolicy
);
1232 p
->mempolicy
= NULL
;
1233 goto bad_fork_cleanup_cgroup
;
1235 mpol_fix_fork_child_flag(p
);
1237 #ifdef CONFIG_CPUSETS
1238 p
->cpuset_mem_spread_rotor
= NUMA_NO_NODE
;
1239 p
->cpuset_slab_spread_rotor
= NUMA_NO_NODE
;
1241 #ifdef CONFIG_TRACE_IRQFLAGS
1243 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1244 p
->hardirqs_enabled
= 1;
1246 p
->hardirqs_enabled
= 0;
1248 p
->hardirq_enable_ip
= 0;
1249 p
->hardirq_enable_event
= 0;
1250 p
->hardirq_disable_ip
= _THIS_IP_
;
1251 p
->hardirq_disable_event
= 0;
1252 p
->softirqs_enabled
= 1;
1253 p
->softirq_enable_ip
= _THIS_IP_
;
1254 p
->softirq_enable_event
= 0;
1255 p
->softirq_disable_ip
= 0;
1256 p
->softirq_disable_event
= 0;
1257 p
->hardirq_context
= 0;
1258 p
->softirq_context
= 0;
1260 #ifdef CONFIG_LOCKDEP
1261 p
->lockdep_depth
= 0; /* no locks held yet */
1262 p
->curr_chain_key
= 0;
1263 p
->lockdep_recursion
= 0;
1266 #ifdef CONFIG_DEBUG_MUTEXES
1267 p
->blocked_on
= NULL
; /* not blocked yet */
1269 #ifdef CONFIG_CGROUP_MEM_RES_CTLR
1270 p
->memcg_batch
.do_batch
= 0;
1271 p
->memcg_batch
.memcg
= NULL
;
1274 /* Perform scheduler related setup. Assign this task to a CPU. */
1277 retval
= perf_event_init_task(p
);
1279 goto bad_fork_cleanup_policy
;
1280 retval
= audit_alloc(p
);
1282 goto bad_fork_cleanup_policy
;
1283 /* copy all the process information */
1284 retval
= copy_semundo(clone_flags
, p
);
1286 goto bad_fork_cleanup_audit
;
1287 retval
= copy_files(clone_flags
, p
);
1289 goto bad_fork_cleanup_semundo
;
1290 retval
= copy_fs(clone_flags
, p
);
1292 goto bad_fork_cleanup_files
;
1293 retval
= copy_sighand(clone_flags
, p
);
1295 goto bad_fork_cleanup_fs
;
1296 retval
= copy_signal(clone_flags
, p
);
1298 goto bad_fork_cleanup_sighand
;
1299 retval
= copy_mm(clone_flags
, p
);
1301 goto bad_fork_cleanup_signal
;
1302 retval
= copy_namespaces(clone_flags
, p
);
1304 goto bad_fork_cleanup_mm
;
1305 retval
= copy_io(clone_flags
, p
);
1307 goto bad_fork_cleanup_namespaces
;
1308 retval
= copy_thread(clone_flags
, stack_start
, stack_size
, p
, regs
);
1310 goto bad_fork_cleanup_io
;
1312 if (pid
!= &init_struct_pid
) {
1314 pid
= alloc_pid(p
->nsproxy
->pid_ns
);
1316 goto bad_fork_cleanup_io
;
1319 p
->pid
= pid_nr(pid
);
1321 if (clone_flags
& CLONE_THREAD
)
1322 p
->tgid
= current
->tgid
;
1324 p
->set_child_tid
= (clone_flags
& CLONE_CHILD_SETTID
) ? child_tidptr
: NULL
;
1326 * Clear TID on mm_release()?
1328 p
->clear_child_tid
= (clone_flags
& CLONE_CHILD_CLEARTID
) ? child_tidptr
: NULL
;
1333 p
->robust_list
= NULL
;
1334 #ifdef CONFIG_COMPAT
1335 p
->compat_robust_list
= NULL
;
1337 INIT_LIST_HEAD(&p
->pi_state_list
);
1338 p
->pi_state_cache
= NULL
;
1341 * sigaltstack should be cleared when sharing the same VM
1343 if ((clone_flags
& (CLONE_VM
|CLONE_VFORK
)) == CLONE_VM
)
1344 p
->sas_ss_sp
= p
->sas_ss_size
= 0;
1347 * Syscall tracing and stepping should be turned off in the
1348 * child regardless of CLONE_PTRACE.
1350 user_disable_single_step(p
);
1351 clear_tsk_thread_flag(p
, TIF_SYSCALL_TRACE
);
1352 #ifdef TIF_SYSCALL_EMU
1353 clear_tsk_thread_flag(p
, TIF_SYSCALL_EMU
);
1355 clear_all_latency_tracing(p
);
1357 /* ok, now we should be set up.. */
1358 if (clone_flags
& CLONE_THREAD
)
1359 p
->exit_signal
= -1;
1360 else if (clone_flags
& CLONE_PARENT
)
1361 p
->exit_signal
= current
->group_leader
->exit_signal
;
1363 p
->exit_signal
= (clone_flags
& CSIGNAL
);
1365 p
->pdeath_signal
= 0;
1369 p
->nr_dirtied_pause
= 128 >> (PAGE_SHIFT
- 10);
1370 p
->dirty_paused_when
= 0;
1373 * Ok, make it visible to the rest of the system.
1374 * We dont wake it up yet.
1376 p
->group_leader
= p
;
1377 INIT_LIST_HEAD(&p
->thread_group
);
1379 /* Now that the task is set up, run cgroup callbacks if
1380 * necessary. We need to run them before the task is visible
1381 * on the tasklist. */
1382 cgroup_fork_callbacks(p
);
1383 cgroup_callbacks_done
= 1;
1385 /* Need tasklist lock for parent etc handling! */
1386 write_lock_irq(&tasklist_lock
);
1388 /* CLONE_PARENT re-uses the old parent */
1389 if (clone_flags
& (CLONE_PARENT
|CLONE_THREAD
)) {
1390 p
->real_parent
= current
->real_parent
;
1391 p
->parent_exec_id
= current
->parent_exec_id
;
1393 p
->real_parent
= current
;
1394 p
->parent_exec_id
= current
->self_exec_id
;
1397 spin_lock(¤t
->sighand
->siglock
);
1400 * Process group and session signals need to be delivered to just the
1401 * parent before the fork or both the parent and the child after the
1402 * fork. Restart if a signal comes in before we add the new process to
1403 * it's process group.
1404 * A fatal signal pending means that current will exit, so the new
1405 * thread can't slip out of an OOM kill (or normal SIGKILL).
1407 recalc_sigpending();
1408 if (signal_pending(current
)) {
1409 spin_unlock(¤t
->sighand
->siglock
);
1410 write_unlock_irq(&tasklist_lock
);
1411 retval
= -ERESTARTNOINTR
;
1412 goto bad_fork_free_pid
;
1415 if (clone_flags
& CLONE_THREAD
) {
1416 current
->signal
->nr_threads
++;
1417 atomic_inc(¤t
->signal
->live
);
1418 atomic_inc(¤t
->signal
->sigcnt
);
1419 p
->group_leader
= current
->group_leader
;
1420 list_add_tail_rcu(&p
->thread_group
, &p
->group_leader
->thread_group
);
1423 if (likely(p
->pid
)) {
1424 ptrace_init_task(p
, (clone_flags
& CLONE_PTRACE
) || trace
);
1426 if (thread_group_leader(p
)) {
1427 if (is_child_reaper(pid
))
1428 p
->nsproxy
->pid_ns
->child_reaper
= p
;
1430 p
->signal
->leader_pid
= pid
;
1431 p
->signal
->tty
= tty_kref_get(current
->signal
->tty
);
1432 attach_pid(p
, PIDTYPE_PGID
, task_pgrp(current
));
1433 attach_pid(p
, PIDTYPE_SID
, task_session(current
));
1434 list_add_tail(&p
->sibling
, &p
->real_parent
->children
);
1435 list_add_tail_rcu(&p
->tasks
, &init_task
.tasks
);
1436 __this_cpu_inc(process_counts
);
1438 attach_pid(p
, PIDTYPE_PID
, pid
);
1443 spin_unlock(¤t
->sighand
->siglock
);
1444 write_unlock_irq(&tasklist_lock
);
1445 proc_fork_connector(p
);
1446 cgroup_post_fork(p
);
1447 if (clone_flags
& CLONE_THREAD
)
1448 threadgroup_change_end(current
);
1451 trace_task_newtask(p
, clone_flags
);
1456 if (pid
!= &init_struct_pid
)
1458 bad_fork_cleanup_io
:
1461 bad_fork_cleanup_namespaces
:
1462 exit_task_namespaces(p
);
1463 bad_fork_cleanup_mm
:
1466 bad_fork_cleanup_signal
:
1467 if (!(clone_flags
& CLONE_THREAD
))
1468 free_signal_struct(p
->signal
);
1469 bad_fork_cleanup_sighand
:
1470 __cleanup_sighand(p
->sighand
);
1471 bad_fork_cleanup_fs
:
1472 exit_fs(p
); /* blocking */
1473 bad_fork_cleanup_files
:
1474 exit_files(p
); /* blocking */
1475 bad_fork_cleanup_semundo
:
1477 bad_fork_cleanup_audit
:
1479 bad_fork_cleanup_policy
:
1480 perf_event_free_task(p
);
1482 mpol_put(p
->mempolicy
);
1483 bad_fork_cleanup_cgroup
:
1485 if (clone_flags
& CLONE_THREAD
)
1486 threadgroup_change_end(current
);
1487 cgroup_exit(p
, cgroup_callbacks_done
);
1488 delayacct_tsk_free(p
);
1489 module_put(task_thread_info(p
)->exec_domain
->module
);
1490 bad_fork_cleanup_count
:
1491 atomic_dec(&p
->cred
->user
->processes
);
1496 return ERR_PTR(retval
);
1499 noinline
struct pt_regs
* __cpuinit
__attribute__((weak
)) idle_regs(struct pt_regs
*regs
)
1501 memset(regs
, 0, sizeof(struct pt_regs
));
1505 static inline void init_idle_pids(struct pid_link
*links
)
1509 for (type
= PIDTYPE_PID
; type
< PIDTYPE_MAX
; ++type
) {
1510 INIT_HLIST_NODE(&links
[type
].node
); /* not really needed */
1511 links
[type
].pid
= &init_struct_pid
;
1515 struct task_struct
* __cpuinit
fork_idle(int cpu
)
1517 struct task_struct
*task
;
1518 struct pt_regs regs
;
1520 task
= copy_process(CLONE_VM
, 0, idle_regs(®s
), 0, NULL
,
1521 &init_struct_pid
, 0);
1522 if (!IS_ERR(task
)) {
1523 init_idle_pids(task
->pids
);
1524 init_idle(task
, cpu
);
1531 * Ok, this is the main fork-routine.
1533 * It copies the process, and if successful kick-starts
1534 * it and waits for it to finish using the VM if required.
1536 long do_fork(unsigned long clone_flags
,
1537 unsigned long stack_start
,
1538 struct pt_regs
*regs
,
1539 unsigned long stack_size
,
1540 int __user
*parent_tidptr
,
1541 int __user
*child_tidptr
)
1543 struct task_struct
*p
;
1548 * Do some preliminary argument and permissions checking before we
1549 * actually start allocating stuff
1551 if (clone_flags
& CLONE_NEWUSER
) {
1552 if (clone_flags
& CLONE_THREAD
)
1554 /* hopefully this check will go away when userns support is
1557 if (!capable(CAP_SYS_ADMIN
) || !capable(CAP_SETUID
) ||
1558 !capable(CAP_SETGID
))
1563 * Determine whether and which event to report to ptracer. When
1564 * called from kernel_thread or CLONE_UNTRACED is explicitly
1565 * requested, no event is reported; otherwise, report if the event
1566 * for the type of forking is enabled.
1568 if (likely(user_mode(regs
)) && !(clone_flags
& CLONE_UNTRACED
)) {
1569 if (clone_flags
& CLONE_VFORK
)
1570 trace
= PTRACE_EVENT_VFORK
;
1571 else if ((clone_flags
& CSIGNAL
) != SIGCHLD
)
1572 trace
= PTRACE_EVENT_CLONE
;
1574 trace
= PTRACE_EVENT_FORK
;
1576 if (likely(!ptrace_event_enabled(current
, trace
)))
1580 p
= copy_process(clone_flags
, stack_start
, regs
, stack_size
,
1581 child_tidptr
, NULL
, trace
);
1583 * Do this prior waking up the new thread - the thread pointer
1584 * might get invalid after that point, if the thread exits quickly.
1587 struct completion vfork
;
1589 trace_sched_process_fork(current
, p
);
1591 nr
= task_pid_vnr(p
);
1593 if (clone_flags
& CLONE_PARENT_SETTID
)
1594 put_user(nr
, parent_tidptr
);
1596 if (clone_flags
& CLONE_VFORK
) {
1597 p
->vfork_done
= &vfork
;
1598 init_completion(&vfork
);
1602 wake_up_new_task(p
);
1604 /* forking complete and child started to run, tell ptracer */
1605 if (unlikely(trace
))
1606 ptrace_event(trace
, nr
);
1608 if (clone_flags
& CLONE_VFORK
) {
1609 if (!wait_for_vfork_done(p
, &vfork
))
1610 ptrace_event(PTRACE_EVENT_VFORK_DONE
, nr
);
1618 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1619 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1622 static void sighand_ctor(void *data
)
1624 struct sighand_struct
*sighand
= data
;
1626 spin_lock_init(&sighand
->siglock
);
1627 init_waitqueue_head(&sighand
->signalfd_wqh
);
1630 void __init
proc_caches_init(void)
1632 sighand_cachep
= kmem_cache_create("sighand_cache",
1633 sizeof(struct sighand_struct
), 0,
1634 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_DESTROY_BY_RCU
|
1635 SLAB_NOTRACK
, sighand_ctor
);
1636 signal_cachep
= kmem_cache_create("signal_cache",
1637 sizeof(struct signal_struct
), 0,
1638 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1639 files_cachep
= kmem_cache_create("files_cache",
1640 sizeof(struct files_struct
), 0,
1641 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1642 fs_cachep
= kmem_cache_create("fs_cache",
1643 sizeof(struct fs_struct
), 0,
1644 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1646 * FIXME! The "sizeof(struct mm_struct)" currently includes the
1647 * whole struct cpumask for the OFFSTACK case. We could change
1648 * this to *only* allocate as much of it as required by the
1649 * maximum number of CPU's we can ever have. The cpumask_allocation
1650 * is at the end of the structure, exactly for that reason.
1652 mm_cachep
= kmem_cache_create("mm_struct",
1653 sizeof(struct mm_struct
), ARCH_MIN_MMSTRUCT_ALIGN
,
1654 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1655 vm_area_cachep
= KMEM_CACHE(vm_area_struct
, SLAB_PANIC
);
1657 nsproxy_cache_init();
1661 * Check constraints on flags passed to the unshare system call.
1663 static int check_unshare_flags(unsigned long unshare_flags
)
1665 if (unshare_flags
& ~(CLONE_THREAD
|CLONE_FS
|CLONE_NEWNS
|CLONE_SIGHAND
|
1666 CLONE_VM
|CLONE_FILES
|CLONE_SYSVSEM
|
1667 CLONE_NEWUTS
|CLONE_NEWIPC
|CLONE_NEWNET
))
1670 * Not implemented, but pretend it works if there is nothing to
1671 * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND
1672 * needs to unshare vm.
1674 if (unshare_flags
& (CLONE_THREAD
| CLONE_SIGHAND
| CLONE_VM
)) {
1675 /* FIXME: get_task_mm() increments ->mm_users */
1676 if (atomic_read(¤t
->mm
->mm_users
) > 1)
1684 * Unshare the filesystem structure if it is being shared
1686 static int unshare_fs(unsigned long unshare_flags
, struct fs_struct
**new_fsp
)
1688 struct fs_struct
*fs
= current
->fs
;
1690 if (!(unshare_flags
& CLONE_FS
) || !fs
)
1693 /* don't need lock here; in the worst case we'll do useless copy */
1697 *new_fsp
= copy_fs_struct(fs
);
1705 * Unshare file descriptor table if it is being shared
1707 static int unshare_fd(unsigned long unshare_flags
, struct files_struct
**new_fdp
)
1709 struct files_struct
*fd
= current
->files
;
1712 if ((unshare_flags
& CLONE_FILES
) &&
1713 (fd
&& atomic_read(&fd
->count
) > 1)) {
1714 *new_fdp
= dup_fd(fd
, &error
);
1723 * unshare allows a process to 'unshare' part of the process
1724 * context which was originally shared using clone. copy_*
1725 * functions used by do_fork() cannot be used here directly
1726 * because they modify an inactive task_struct that is being
1727 * constructed. Here we are modifying the current, active,
1730 SYSCALL_DEFINE1(unshare
, unsigned long, unshare_flags
)
1732 struct fs_struct
*fs
, *new_fs
= NULL
;
1733 struct files_struct
*fd
, *new_fd
= NULL
;
1734 struct nsproxy
*new_nsproxy
= NULL
;
1738 err
= check_unshare_flags(unshare_flags
);
1740 goto bad_unshare_out
;
1743 * If unsharing namespace, must also unshare filesystem information.
1745 if (unshare_flags
& CLONE_NEWNS
)
1746 unshare_flags
|= CLONE_FS
;
1748 * CLONE_NEWIPC must also detach from the undolist: after switching
1749 * to a new ipc namespace, the semaphore arrays from the old
1750 * namespace are unreachable.
1752 if (unshare_flags
& (CLONE_NEWIPC
|CLONE_SYSVSEM
))
1754 err
= unshare_fs(unshare_flags
, &new_fs
);
1756 goto bad_unshare_out
;
1757 err
= unshare_fd(unshare_flags
, &new_fd
);
1759 goto bad_unshare_cleanup_fs
;
1760 err
= unshare_nsproxy_namespaces(unshare_flags
, &new_nsproxy
, new_fs
);
1762 goto bad_unshare_cleanup_fd
;
1764 if (new_fs
|| new_fd
|| do_sysvsem
|| new_nsproxy
) {
1767 * CLONE_SYSVSEM is equivalent to sys_exit().
1773 switch_task_namespaces(current
, new_nsproxy
);
1781 spin_lock(&fs
->lock
);
1782 current
->fs
= new_fs
;
1787 spin_unlock(&fs
->lock
);
1791 fd
= current
->files
;
1792 current
->files
= new_fd
;
1796 task_unlock(current
);
1800 put_nsproxy(new_nsproxy
);
1802 bad_unshare_cleanup_fd
:
1804 put_files_struct(new_fd
);
1806 bad_unshare_cleanup_fs
:
1808 free_fs_struct(new_fs
);
1815 * Helper to unshare the files of the current task.
1816 * We don't want to expose copy_files internals to
1817 * the exec layer of the kernel.
1820 int unshare_files(struct files_struct
**displaced
)
1822 struct task_struct
*task
= current
;
1823 struct files_struct
*copy
= NULL
;
1826 error
= unshare_fd(CLONE_FILES
, ©
);
1827 if (error
|| !copy
) {
1831 *displaced
= task
->files
;