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/seccomp.h>
38 #include <linux/swap.h>
39 #include <linux/syscalls.h>
40 #include <linux/jiffies.h>
41 #include <linux/futex.h>
42 #include <linux/compat.h>
43 #include <linux/kthread.h>
44 #include <linux/task_io_accounting_ops.h>
45 #include <linux/rcupdate.h>
46 #include <linux/ptrace.h>
47 #include <linux/mount.h>
48 #include <linux/audit.h>
49 #include <linux/memcontrol.h>
50 #include <linux/ftrace.h>
51 #include <linux/proc_fs.h>
52 #include <linux/profile.h>
53 #include <linux/rmap.h>
54 #include <linux/ksm.h>
55 #include <linux/acct.h>
56 #include <linux/tsacct_kern.h>
57 #include <linux/cn_proc.h>
58 #include <linux/freezer.h>
59 #include <linux/delayacct.h>
60 #include <linux/taskstats_kern.h>
61 #include <linux/random.h>
62 #include <linux/tty.h>
63 #include <linux/blkdev.h>
64 #include <linux/fs_struct.h>
65 #include <linux/magic.h>
66 #include <linux/perf_event.h>
67 #include <linux/posix-timers.h>
68 #include <linux/user-return-notifier.h>
69 #include <linux/oom.h>
70 #include <linux/khugepaged.h>
71 #include <linux/signalfd.h>
72 #include <linux/uprobes.h>
74 #include <asm/pgtable.h>
75 #include <asm/pgalloc.h>
76 #include <asm/uaccess.h>
77 #include <asm/mmu_context.h>
78 #include <asm/cacheflush.h>
79 #include <asm/tlbflush.h>
81 #include <trace/events/sched.h>
83 #define CREATE_TRACE_POINTS
84 #include <trace/events/task.h>
87 * Protected counters by write_lock_irq(&tasklist_lock)
89 unsigned long total_forks
; /* Handle normal Linux uptimes. */
90 int nr_threads
; /* The idle threads do not count.. */
92 int max_threads
; /* tunable limit on nr_threads */
94 DEFINE_PER_CPU(unsigned long, process_counts
) = 0;
96 __cacheline_aligned
DEFINE_RWLOCK(tasklist_lock
); /* outer */
98 #ifdef CONFIG_PROVE_RCU
99 int lockdep_tasklist_lock_is_held(void)
101 return lockdep_is_held(&tasklist_lock
);
103 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held
);
104 #endif /* #ifdef CONFIG_PROVE_RCU */
106 int nr_processes(void)
111 for_each_possible_cpu(cpu
)
112 total
+= per_cpu(process_counts
, cpu
);
117 void __weak
arch_release_task_struct(struct task_struct
*tsk
)
121 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
122 static struct kmem_cache
*task_struct_cachep
;
124 static inline struct task_struct
*alloc_task_struct_node(int node
)
126 return kmem_cache_alloc_node(task_struct_cachep
, GFP_KERNEL
, node
);
129 static inline void free_task_struct(struct task_struct
*tsk
)
131 kmem_cache_free(task_struct_cachep
, tsk
);
135 void __weak
arch_release_thread_info(struct thread_info
*ti
)
139 #ifndef CONFIG_ARCH_THREAD_INFO_ALLOCATOR
142 * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a
143 * kmemcache based allocator.
145 # if THREAD_SIZE >= PAGE_SIZE
146 static struct thread_info
*alloc_thread_info_node(struct task_struct
*tsk
,
149 struct page
*page
= alloc_pages_node(node
, THREADINFO_GFP
,
152 return page
? page_address(page
) : NULL
;
155 static inline void free_thread_info(struct thread_info
*ti
)
157 free_pages((unsigned long)ti
, THREAD_SIZE_ORDER
);
160 static struct kmem_cache
*thread_info_cache
;
162 static struct thread_info
*alloc_thread_info_node(struct task_struct
*tsk
,
165 return kmem_cache_alloc_node(thread_info_cache
, THREADINFO_GFP
, node
);
168 static void free_thread_info(struct thread_info
*ti
)
170 kmem_cache_free(thread_info_cache
, ti
);
173 void thread_info_cache_init(void)
175 thread_info_cache
= kmem_cache_create("thread_info", THREAD_SIZE
,
176 THREAD_SIZE
, 0, NULL
);
177 BUG_ON(thread_info_cache
== NULL
);
182 /* SLAB cache for signal_struct structures (tsk->signal) */
183 static struct kmem_cache
*signal_cachep
;
185 /* SLAB cache for sighand_struct structures (tsk->sighand) */
186 struct kmem_cache
*sighand_cachep
;
188 /* SLAB cache for files_struct structures (tsk->files) */
189 struct kmem_cache
*files_cachep
;
191 /* SLAB cache for fs_struct structures (tsk->fs) */
192 struct kmem_cache
*fs_cachep
;
194 /* SLAB cache for vm_area_struct structures */
195 struct kmem_cache
*vm_area_cachep
;
197 /* SLAB cache for mm_struct structures (tsk->mm) */
198 static struct kmem_cache
*mm_cachep
;
200 static void account_kernel_stack(struct thread_info
*ti
, int account
)
202 struct zone
*zone
= page_zone(virt_to_page(ti
));
204 mod_zone_page_state(zone
, NR_KERNEL_STACK
, account
);
207 void free_task(struct task_struct
*tsk
)
209 account_kernel_stack(tsk
->stack
, -1);
210 arch_release_thread_info(tsk
->stack
);
211 free_thread_info(tsk
->stack
);
212 rt_mutex_debug_task_free(tsk
);
213 ftrace_graph_exit_task(tsk
);
214 put_seccomp_filter(tsk
);
215 arch_release_task_struct(tsk
);
216 free_task_struct(tsk
);
218 EXPORT_SYMBOL(free_task
);
220 static inline void free_signal_struct(struct signal_struct
*sig
)
222 taskstats_tgid_free(sig
);
223 sched_autogroup_exit(sig
);
224 kmem_cache_free(signal_cachep
, sig
);
227 static inline void put_signal_struct(struct signal_struct
*sig
)
229 if (atomic_dec_and_test(&sig
->sigcnt
))
230 free_signal_struct(sig
);
233 void __put_task_struct(struct task_struct
*tsk
)
235 WARN_ON(!tsk
->exit_state
);
236 WARN_ON(atomic_read(&tsk
->usage
));
237 WARN_ON(tsk
== current
);
239 security_task_free(tsk
);
241 delayacct_tsk_free(tsk
);
242 put_signal_struct(tsk
->signal
);
244 if (!profile_handoff_task(tsk
))
247 EXPORT_SYMBOL_GPL(__put_task_struct
);
249 void __init __weak
arch_task_cache_init(void) { }
251 void __init
fork_init(unsigned long mempages
)
253 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
254 #ifndef ARCH_MIN_TASKALIGN
255 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
257 /* create a slab on which task_structs can be allocated */
259 kmem_cache_create("task_struct", sizeof(struct task_struct
),
260 ARCH_MIN_TASKALIGN
, SLAB_PANIC
| SLAB_NOTRACK
, NULL
);
263 /* do the arch specific task caches init */
264 arch_task_cache_init();
267 * The default maximum number of threads is set to a safe
268 * value: the thread structures can take up at most half
271 max_threads
= mempages
/ (8 * THREAD_SIZE
/ PAGE_SIZE
);
274 * we need to allow at least 20 threads to boot a system
276 if (max_threads
< 20)
279 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_cur
= max_threads
/2;
280 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_max
= max_threads
/2;
281 init_task
.signal
->rlim
[RLIMIT_SIGPENDING
] =
282 init_task
.signal
->rlim
[RLIMIT_NPROC
];
285 int __attribute__((weak
)) arch_dup_task_struct(struct task_struct
*dst
,
286 struct task_struct
*src
)
292 static struct task_struct
*dup_task_struct(struct task_struct
*orig
)
294 struct task_struct
*tsk
;
295 struct thread_info
*ti
;
296 unsigned long *stackend
;
297 int node
= tsk_fork_get_node(orig
);
300 tsk
= alloc_task_struct_node(node
);
304 ti
= alloc_thread_info_node(tsk
, node
);
308 err
= arch_dup_task_struct(tsk
, orig
);
314 setup_thread_stack(tsk
, orig
);
315 clear_user_return_notifier(tsk
);
316 clear_tsk_need_resched(tsk
);
317 stackend
= end_of_stack(tsk
);
318 *stackend
= STACK_END_MAGIC
; /* for overflow detection */
320 #ifdef CONFIG_CC_STACKPROTECTOR
321 tsk
->stack_canary
= get_random_int();
325 * One for us, one for whoever does the "release_task()" (usually
328 atomic_set(&tsk
->usage
, 2);
329 #ifdef CONFIG_BLK_DEV_IO_TRACE
332 tsk
->splice_pipe
= NULL
;
333 tsk
->task_frag
.page
= NULL
;
335 account_kernel_stack(ti
, 1);
340 free_thread_info(ti
);
342 free_task_struct(tsk
);
347 static int dup_mmap(struct mm_struct
*mm
, struct mm_struct
*oldmm
)
349 struct vm_area_struct
*mpnt
, *tmp
, *prev
, **pprev
;
350 struct rb_node
**rb_link
, *rb_parent
;
352 unsigned long charge
;
353 struct mempolicy
*pol
;
355 down_write(&oldmm
->mmap_sem
);
356 flush_cache_dup_mm(oldmm
);
357 uprobe_dup_mmap(oldmm
, mm
);
359 * Not linked in yet - no deadlock potential:
361 down_write_nested(&mm
->mmap_sem
, SINGLE_DEPTH_NESTING
);
365 mm
->mmap_cache
= NULL
;
366 mm
->free_area_cache
= oldmm
->mmap_base
;
367 mm
->cached_hole_size
= ~0UL;
369 cpumask_clear(mm_cpumask(mm
));
371 rb_link
= &mm
->mm_rb
.rb_node
;
374 retval
= ksm_fork(mm
, oldmm
);
377 retval
= khugepaged_fork(mm
, oldmm
);
382 for (mpnt
= oldmm
->mmap
; mpnt
; mpnt
= mpnt
->vm_next
) {
385 if (mpnt
->vm_flags
& VM_DONTCOPY
) {
386 vm_stat_account(mm
, mpnt
->vm_flags
, mpnt
->vm_file
,
391 if (mpnt
->vm_flags
& VM_ACCOUNT
) {
392 unsigned long len
= vma_pages(mpnt
);
394 if (security_vm_enough_memory_mm(oldmm
, len
)) /* sic */
398 tmp
= kmem_cache_alloc(vm_area_cachep
, GFP_KERNEL
);
402 INIT_LIST_HEAD(&tmp
->anon_vma_chain
);
403 pol
= mpol_dup(vma_policy(mpnt
));
404 retval
= PTR_ERR(pol
);
406 goto fail_nomem_policy
;
407 vma_set_policy(tmp
, pol
);
409 if (anon_vma_fork(tmp
, mpnt
))
410 goto fail_nomem_anon_vma_fork
;
411 tmp
->vm_flags
&= ~VM_LOCKED
;
412 tmp
->vm_next
= tmp
->vm_prev
= NULL
;
415 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
416 struct address_space
*mapping
= file
->f_mapping
;
419 if (tmp
->vm_flags
& VM_DENYWRITE
)
420 atomic_dec(&inode
->i_writecount
);
421 mutex_lock(&mapping
->i_mmap_mutex
);
422 if (tmp
->vm_flags
& VM_SHARED
)
423 mapping
->i_mmap_writable
++;
424 flush_dcache_mmap_lock(mapping
);
425 /* insert tmp into the share list, just after mpnt */
426 vma_prio_tree_add(tmp
, mpnt
);
427 flush_dcache_mmap_unlock(mapping
);
428 mutex_unlock(&mapping
->i_mmap_mutex
);
432 * Clear hugetlb-related page reserves for children. This only
433 * affects MAP_PRIVATE mappings. Faults generated by the child
434 * are not guaranteed to succeed, even if read-only
436 if (is_vm_hugetlb_page(tmp
))
437 reset_vma_resv_huge_pages(tmp
);
440 * Link in the new vma and copy the page table entries.
443 pprev
= &tmp
->vm_next
;
447 __vma_link_rb(mm
, tmp
, rb_link
, rb_parent
);
448 rb_link
= &tmp
->vm_rb
.rb_right
;
449 rb_parent
= &tmp
->vm_rb
;
452 retval
= copy_page_range(mm
, oldmm
, mpnt
);
454 if (tmp
->vm_ops
&& tmp
->vm_ops
->open
)
455 tmp
->vm_ops
->open(tmp
);
460 /* a new mm has just been created */
461 arch_dup_mmap(oldmm
, mm
);
464 up_write(&mm
->mmap_sem
);
466 up_write(&oldmm
->mmap_sem
);
468 fail_nomem_anon_vma_fork
:
471 kmem_cache_free(vm_area_cachep
, tmp
);
474 vm_unacct_memory(charge
);
478 static inline int mm_alloc_pgd(struct mm_struct
*mm
)
480 mm
->pgd
= pgd_alloc(mm
);
481 if (unlikely(!mm
->pgd
))
486 static inline void mm_free_pgd(struct mm_struct
*mm
)
488 pgd_free(mm
, mm
->pgd
);
491 #define dup_mmap(mm, oldmm) (0)
492 #define mm_alloc_pgd(mm) (0)
493 #define mm_free_pgd(mm)
494 #endif /* CONFIG_MMU */
496 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(mmlist_lock
);
498 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
499 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
501 static unsigned long default_dump_filter
= MMF_DUMP_FILTER_DEFAULT
;
503 static int __init
coredump_filter_setup(char *s
)
505 default_dump_filter
=
506 (simple_strtoul(s
, NULL
, 0) << MMF_DUMP_FILTER_SHIFT
) &
507 MMF_DUMP_FILTER_MASK
;
511 __setup("coredump_filter=", coredump_filter_setup
);
513 #include <linux/init_task.h>
515 static void mm_init_aio(struct mm_struct
*mm
)
518 spin_lock_init(&mm
->ioctx_lock
);
519 INIT_HLIST_HEAD(&mm
->ioctx_list
);
523 static struct mm_struct
*mm_init(struct mm_struct
*mm
, struct task_struct
*p
)
525 atomic_set(&mm
->mm_users
, 1);
526 atomic_set(&mm
->mm_count
, 1);
527 init_rwsem(&mm
->mmap_sem
);
528 INIT_LIST_HEAD(&mm
->mmlist
);
529 mm
->flags
= (current
->mm
) ?
530 (current
->mm
->flags
& MMF_INIT_MASK
) : default_dump_filter
;
531 mm
->core_state
= NULL
;
533 memset(&mm
->rss_stat
, 0, sizeof(mm
->rss_stat
));
534 spin_lock_init(&mm
->page_table_lock
);
535 mm
->free_area_cache
= TASK_UNMAPPED_BASE
;
536 mm
->cached_hole_size
= ~0UL;
538 mm_init_owner(mm
, p
);
540 if (likely(!mm_alloc_pgd(mm
))) {
542 mmu_notifier_mm_init(mm
);
550 static void check_mm(struct mm_struct
*mm
)
554 for (i
= 0; i
< NR_MM_COUNTERS
; i
++) {
555 long x
= atomic_long_read(&mm
->rss_stat
.count
[i
]);
558 printk(KERN_ALERT
"BUG: Bad rss-counter state "
559 "mm:%p idx:%d val:%ld\n", mm
, i
, x
);
562 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
563 VM_BUG_ON(mm
->pmd_huge_pte
);
568 * Allocate and initialize an mm_struct.
570 struct mm_struct
*mm_alloc(void)
572 struct mm_struct
*mm
;
578 memset(mm
, 0, sizeof(*mm
));
580 return mm_init(mm
, current
);
584 * Called when the last reference to the mm
585 * is dropped: either by a lazy thread or by
586 * mmput. Free the page directory and the mm.
588 void __mmdrop(struct mm_struct
*mm
)
590 BUG_ON(mm
== &init_mm
);
593 mmu_notifier_mm_destroy(mm
);
597 EXPORT_SYMBOL_GPL(__mmdrop
);
600 * Decrement the use count and release all resources for an mm.
602 void mmput(struct mm_struct
*mm
)
606 if (atomic_dec_and_test(&mm
->mm_users
)) {
607 uprobe_clear_state(mm
);
610 khugepaged_exit(mm
); /* must run before exit_mmap */
612 set_mm_exe_file(mm
, NULL
);
613 if (!list_empty(&mm
->mmlist
)) {
614 spin_lock(&mmlist_lock
);
615 list_del(&mm
->mmlist
);
616 spin_unlock(&mmlist_lock
);
619 module_put(mm
->binfmt
->module
);
623 EXPORT_SYMBOL_GPL(mmput
);
625 void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
)
628 get_file(new_exe_file
);
631 mm
->exe_file
= new_exe_file
;
634 struct file
*get_mm_exe_file(struct mm_struct
*mm
)
636 struct file
*exe_file
;
638 /* We need mmap_sem to protect against races with removal of exe_file */
639 down_read(&mm
->mmap_sem
);
640 exe_file
= mm
->exe_file
;
643 up_read(&mm
->mmap_sem
);
647 static void dup_mm_exe_file(struct mm_struct
*oldmm
, struct mm_struct
*newmm
)
649 /* It's safe to write the exe_file pointer without exe_file_lock because
650 * this is called during fork when the task is not yet in /proc */
651 newmm
->exe_file
= get_mm_exe_file(oldmm
);
655 * get_task_mm - acquire a reference to the task's mm
657 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
658 * this kernel workthread has transiently adopted a user mm with use_mm,
659 * to do its AIO) is not set and if so returns a reference to it, after
660 * bumping up the use count. User must release the mm via mmput()
661 * after use. Typically used by /proc and ptrace.
663 struct mm_struct
*get_task_mm(struct task_struct
*task
)
665 struct mm_struct
*mm
;
670 if (task
->flags
& PF_KTHREAD
)
673 atomic_inc(&mm
->mm_users
);
678 EXPORT_SYMBOL_GPL(get_task_mm
);
680 struct mm_struct
*mm_access(struct task_struct
*task
, unsigned int mode
)
682 struct mm_struct
*mm
;
685 err
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
689 mm
= get_task_mm(task
);
690 if (mm
&& mm
!= current
->mm
&&
691 !ptrace_may_access(task
, mode
)) {
693 mm
= ERR_PTR(-EACCES
);
695 mutex_unlock(&task
->signal
->cred_guard_mutex
);
700 static void complete_vfork_done(struct task_struct
*tsk
)
702 struct completion
*vfork
;
705 vfork
= tsk
->vfork_done
;
707 tsk
->vfork_done
= NULL
;
713 static int wait_for_vfork_done(struct task_struct
*child
,
714 struct completion
*vfork
)
718 freezer_do_not_count();
719 killed
= wait_for_completion_killable(vfork
);
724 child
->vfork_done
= NULL
;
728 put_task_struct(child
);
732 /* Please note the differences between mmput and mm_release.
733 * mmput is called whenever we stop holding onto a mm_struct,
734 * error success whatever.
736 * mm_release is called after a mm_struct has been removed
737 * from the current process.
739 * This difference is important for error handling, when we
740 * only half set up a mm_struct for a new process and need to restore
741 * the old one. Because we mmput the new mm_struct before
742 * restoring the old one. . .
743 * Eric Biederman 10 January 1998
745 void mm_release(struct task_struct
*tsk
, struct mm_struct
*mm
)
747 /* Get rid of any futexes when releasing the mm */
749 if (unlikely(tsk
->robust_list
)) {
750 exit_robust_list(tsk
);
751 tsk
->robust_list
= NULL
;
754 if (unlikely(tsk
->compat_robust_list
)) {
755 compat_exit_robust_list(tsk
);
756 tsk
->compat_robust_list
= NULL
;
759 if (unlikely(!list_empty(&tsk
->pi_state_list
)))
760 exit_pi_state_list(tsk
);
763 uprobe_free_utask(tsk
);
765 /* Get rid of any cached register state */
766 deactivate_mm(tsk
, mm
);
769 * If we're exiting normally, clear a user-space tid field if
770 * requested. We leave this alone when dying by signal, to leave
771 * the value intact in a core dump, and to save the unnecessary
772 * trouble, say, a killed vfork parent shouldn't touch this mm.
773 * Userland only wants this done for a sys_exit.
775 if (tsk
->clear_child_tid
) {
776 if (!(tsk
->flags
& PF_SIGNALED
) &&
777 atomic_read(&mm
->mm_users
) > 1) {
779 * We don't check the error code - if userspace has
780 * not set up a proper pointer then tough luck.
782 put_user(0, tsk
->clear_child_tid
);
783 sys_futex(tsk
->clear_child_tid
, FUTEX_WAKE
,
786 tsk
->clear_child_tid
= NULL
;
790 * All done, finally we can wake up parent and return this mm to him.
791 * Also kthread_stop() uses this completion for synchronization.
794 complete_vfork_done(tsk
);
798 * Allocate a new mm structure and copy contents from the
799 * mm structure of the passed in task structure.
801 struct mm_struct
*dup_mm(struct task_struct
*tsk
)
803 struct mm_struct
*mm
, *oldmm
= current
->mm
;
813 memcpy(mm
, oldmm
, sizeof(*mm
));
816 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
817 mm
->pmd_huge_pte
= NULL
;
819 if (!mm_init(mm
, tsk
))
822 if (init_new_context(tsk
, mm
))
825 dup_mm_exe_file(oldmm
, mm
);
827 err
= dup_mmap(mm
, oldmm
);
831 mm
->hiwater_rss
= get_mm_rss(mm
);
832 mm
->hiwater_vm
= mm
->total_vm
;
834 if (mm
->binfmt
&& !try_module_get(mm
->binfmt
->module
))
840 /* don't put binfmt in mmput, we haven't got module yet */
849 * If init_new_context() failed, we cannot use mmput() to free the mm
850 * because it calls destroy_context()
857 static int copy_mm(unsigned long clone_flags
, struct task_struct
*tsk
)
859 struct mm_struct
*mm
, *oldmm
;
862 tsk
->min_flt
= tsk
->maj_flt
= 0;
863 tsk
->nvcsw
= tsk
->nivcsw
= 0;
864 #ifdef CONFIG_DETECT_HUNG_TASK
865 tsk
->last_switch_count
= tsk
->nvcsw
+ tsk
->nivcsw
;
869 tsk
->active_mm
= NULL
;
872 * Are we cloning a kernel thread?
874 * We need to steal a active VM for that..
880 if (clone_flags
& CLONE_VM
) {
881 atomic_inc(&oldmm
->mm_users
);
900 static int copy_fs(unsigned long clone_flags
, struct task_struct
*tsk
)
902 struct fs_struct
*fs
= current
->fs
;
903 if (clone_flags
& CLONE_FS
) {
904 /* tsk->fs is already what we want */
905 spin_lock(&fs
->lock
);
907 spin_unlock(&fs
->lock
);
911 spin_unlock(&fs
->lock
);
914 tsk
->fs
= copy_fs_struct(fs
);
920 static int copy_files(unsigned long clone_flags
, struct task_struct
*tsk
)
922 struct files_struct
*oldf
, *newf
;
926 * A background process may not have any files ...
928 oldf
= current
->files
;
932 if (clone_flags
& CLONE_FILES
) {
933 atomic_inc(&oldf
->count
);
937 newf
= dup_fd(oldf
, &error
);
947 static int copy_io(unsigned long clone_flags
, struct task_struct
*tsk
)
950 struct io_context
*ioc
= current
->io_context
;
951 struct io_context
*new_ioc
;
956 * Share io context with parent, if CLONE_IO is set
958 if (clone_flags
& CLONE_IO
) {
960 tsk
->io_context
= ioc
;
961 } else if (ioprio_valid(ioc
->ioprio
)) {
962 new_ioc
= get_task_io_context(tsk
, GFP_KERNEL
, NUMA_NO_NODE
);
963 if (unlikely(!new_ioc
))
966 new_ioc
->ioprio
= ioc
->ioprio
;
967 put_io_context(new_ioc
);
973 static int copy_sighand(unsigned long clone_flags
, struct task_struct
*tsk
)
975 struct sighand_struct
*sig
;
977 if (clone_flags
& CLONE_SIGHAND
) {
978 atomic_inc(¤t
->sighand
->count
);
981 sig
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
982 rcu_assign_pointer(tsk
->sighand
, sig
);
985 atomic_set(&sig
->count
, 1);
986 memcpy(sig
->action
, current
->sighand
->action
, sizeof(sig
->action
));
990 void __cleanup_sighand(struct sighand_struct
*sighand
)
992 if (atomic_dec_and_test(&sighand
->count
)) {
993 signalfd_cleanup(sighand
);
994 kmem_cache_free(sighand_cachep
, sighand
);
1000 * Initialize POSIX timer handling for a thread group.
1002 static void posix_cpu_timers_init_group(struct signal_struct
*sig
)
1004 unsigned long cpu_limit
;
1006 /* Thread group counters. */
1007 thread_group_cputime_init(sig
);
1009 cpu_limit
= ACCESS_ONCE(sig
->rlim
[RLIMIT_CPU
].rlim_cur
);
1010 if (cpu_limit
!= RLIM_INFINITY
) {
1011 sig
->cputime_expires
.prof_exp
= secs_to_cputime(cpu_limit
);
1012 sig
->cputimer
.running
= 1;
1015 /* The timer lists. */
1016 INIT_LIST_HEAD(&sig
->cpu_timers
[0]);
1017 INIT_LIST_HEAD(&sig
->cpu_timers
[1]);
1018 INIT_LIST_HEAD(&sig
->cpu_timers
[2]);
1021 static int copy_signal(unsigned long clone_flags
, struct task_struct
*tsk
)
1023 struct signal_struct
*sig
;
1025 if (clone_flags
& CLONE_THREAD
)
1028 sig
= kmem_cache_zalloc(signal_cachep
, GFP_KERNEL
);
1033 sig
->nr_threads
= 1;
1034 atomic_set(&sig
->live
, 1);
1035 atomic_set(&sig
->sigcnt
, 1);
1036 init_waitqueue_head(&sig
->wait_chldexit
);
1037 if (clone_flags
& CLONE_NEWPID
)
1038 sig
->flags
|= SIGNAL_UNKILLABLE
;
1039 sig
->curr_target
= tsk
;
1040 init_sigpending(&sig
->shared_pending
);
1041 INIT_LIST_HEAD(&sig
->posix_timers
);
1043 hrtimer_init(&sig
->real_timer
, CLOCK_MONOTONIC
, HRTIMER_MODE_REL
);
1044 sig
->real_timer
.function
= it_real_fn
;
1046 task_lock(current
->group_leader
);
1047 memcpy(sig
->rlim
, current
->signal
->rlim
, sizeof sig
->rlim
);
1048 task_unlock(current
->group_leader
);
1050 posix_cpu_timers_init_group(sig
);
1052 tty_audit_fork(sig
);
1053 sched_autogroup_fork(sig
);
1055 #ifdef CONFIG_CGROUPS
1056 init_rwsem(&sig
->group_rwsem
);
1059 sig
->oom_adj
= current
->signal
->oom_adj
;
1060 sig
->oom_score_adj
= current
->signal
->oom_score_adj
;
1061 sig
->oom_score_adj_min
= current
->signal
->oom_score_adj_min
;
1063 sig
->has_child_subreaper
= current
->signal
->has_child_subreaper
||
1064 current
->signal
->is_child_subreaper
;
1066 mutex_init(&sig
->cred_guard_mutex
);
1071 static void copy_flags(unsigned long clone_flags
, struct task_struct
*p
)
1073 unsigned long new_flags
= p
->flags
;
1075 new_flags
&= ~(PF_SUPERPRIV
| PF_WQ_WORKER
);
1076 new_flags
|= PF_FORKNOEXEC
;
1077 p
->flags
= new_flags
;
1080 SYSCALL_DEFINE1(set_tid_address
, int __user
*, tidptr
)
1082 current
->clear_child_tid
= tidptr
;
1084 return task_pid_vnr(current
);
1087 static void rt_mutex_init_task(struct task_struct
*p
)
1089 raw_spin_lock_init(&p
->pi_lock
);
1090 #ifdef CONFIG_RT_MUTEXES
1091 plist_head_init(&p
->pi_waiters
);
1092 p
->pi_blocked_on
= NULL
;
1096 #ifdef CONFIG_MM_OWNER
1097 void mm_init_owner(struct mm_struct
*mm
, struct task_struct
*p
)
1101 #endif /* CONFIG_MM_OWNER */
1104 * Initialize POSIX timer handling for a single task.
1106 static void posix_cpu_timers_init(struct task_struct
*tsk
)
1108 tsk
->cputime_expires
.prof_exp
= 0;
1109 tsk
->cputime_expires
.virt_exp
= 0;
1110 tsk
->cputime_expires
.sched_exp
= 0;
1111 INIT_LIST_HEAD(&tsk
->cpu_timers
[0]);
1112 INIT_LIST_HEAD(&tsk
->cpu_timers
[1]);
1113 INIT_LIST_HEAD(&tsk
->cpu_timers
[2]);
1117 * This creates a new process as a copy of the old one,
1118 * but does not actually start it yet.
1120 * It copies the registers, and all the appropriate
1121 * parts of the process environment (as per the clone
1122 * flags). The actual kick-off is left to the caller.
1124 static struct task_struct
*copy_process(unsigned long clone_flags
,
1125 unsigned long stack_start
,
1126 struct pt_regs
*regs
,
1127 unsigned long stack_size
,
1128 int __user
*child_tidptr
,
1133 struct task_struct
*p
;
1134 int cgroup_callbacks_done
= 0;
1136 if ((clone_flags
& (CLONE_NEWNS
|CLONE_FS
)) == (CLONE_NEWNS
|CLONE_FS
))
1137 return ERR_PTR(-EINVAL
);
1140 * Thread groups must share signals as well, and detached threads
1141 * can only be started up within the thread group.
1143 if ((clone_flags
& CLONE_THREAD
) && !(clone_flags
& CLONE_SIGHAND
))
1144 return ERR_PTR(-EINVAL
);
1147 * Shared signal handlers imply shared VM. By way of the above,
1148 * thread groups also imply shared VM. Blocking this case allows
1149 * for various simplifications in other code.
1151 if ((clone_flags
& CLONE_SIGHAND
) && !(clone_flags
& CLONE_VM
))
1152 return ERR_PTR(-EINVAL
);
1155 * Siblings of global init remain as zombies on exit since they are
1156 * not reaped by their parent (swapper). To solve this and to avoid
1157 * multi-rooted process trees, prevent global and container-inits
1158 * from creating siblings.
1160 if ((clone_flags
& CLONE_PARENT
) &&
1161 current
->signal
->flags
& SIGNAL_UNKILLABLE
)
1162 return ERR_PTR(-EINVAL
);
1164 retval
= security_task_create(clone_flags
);
1169 p
= dup_task_struct(current
);
1173 ftrace_graph_init_task(p
);
1174 get_seccomp_filter(p
);
1176 rt_mutex_init_task(p
);
1178 #ifdef CONFIG_PROVE_LOCKING
1179 DEBUG_LOCKS_WARN_ON(!p
->hardirqs_enabled
);
1180 DEBUG_LOCKS_WARN_ON(!p
->softirqs_enabled
);
1183 if (atomic_read(&p
->real_cred
->user
->processes
) >=
1184 task_rlimit(p
, RLIMIT_NPROC
)) {
1185 if (!capable(CAP_SYS_ADMIN
) && !capable(CAP_SYS_RESOURCE
) &&
1186 p
->real_cred
->user
!= INIT_USER
)
1189 current
->flags
&= ~PF_NPROC_EXCEEDED
;
1191 retval
= copy_creds(p
, clone_flags
);
1196 * If multiple threads are within copy_process(), then this check
1197 * triggers too late. This doesn't hurt, the check is only there
1198 * to stop root fork bombs.
1201 if (nr_threads
>= max_threads
)
1202 goto bad_fork_cleanup_count
;
1204 if (!try_module_get(task_thread_info(p
)->exec_domain
->module
))
1205 goto bad_fork_cleanup_count
;
1208 delayacct_tsk_init(p
); /* Must remain after dup_task_struct() */
1209 copy_flags(clone_flags
, p
);
1210 INIT_LIST_HEAD(&p
->children
);
1211 INIT_LIST_HEAD(&p
->sibling
);
1212 rcu_copy_process(p
);
1213 p
->vfork_done
= NULL
;
1214 spin_lock_init(&p
->alloc_lock
);
1216 init_sigpending(&p
->pending
);
1218 p
->utime
= p
->stime
= p
->gtime
= 0;
1219 p
->utimescaled
= p
->stimescaled
= 0;
1220 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
1221 p
->prev_utime
= p
->prev_stime
= 0;
1223 #if defined(SPLIT_RSS_COUNTING)
1224 memset(&p
->rss_stat
, 0, sizeof(p
->rss_stat
));
1227 p
->default_timer_slack_ns
= current
->timer_slack_ns
;
1229 task_io_accounting_init(&p
->ioac
);
1230 acct_clear_integrals(p
);
1232 posix_cpu_timers_init(p
);
1234 do_posix_clock_monotonic_gettime(&p
->start_time
);
1235 p
->real_start_time
= p
->start_time
;
1236 monotonic_to_bootbased(&p
->real_start_time
);
1237 p
->io_context
= NULL
;
1238 p
->audit_context
= NULL
;
1239 if (clone_flags
& CLONE_THREAD
)
1240 threadgroup_change_begin(current
);
1243 p
->mempolicy
= mpol_dup(p
->mempolicy
);
1244 if (IS_ERR(p
->mempolicy
)) {
1245 retval
= PTR_ERR(p
->mempolicy
);
1246 p
->mempolicy
= NULL
;
1247 goto bad_fork_cleanup_cgroup
;
1249 mpol_fix_fork_child_flag(p
);
1251 #ifdef CONFIG_CPUSETS
1252 p
->cpuset_mem_spread_rotor
= NUMA_NO_NODE
;
1253 p
->cpuset_slab_spread_rotor
= NUMA_NO_NODE
;
1254 seqcount_init(&p
->mems_allowed_seq
);
1256 #ifdef CONFIG_TRACE_IRQFLAGS
1258 p
->hardirqs_enabled
= 0;
1259 p
->hardirq_enable_ip
= 0;
1260 p
->hardirq_enable_event
= 0;
1261 p
->hardirq_disable_ip
= _THIS_IP_
;
1262 p
->hardirq_disable_event
= 0;
1263 p
->softirqs_enabled
= 1;
1264 p
->softirq_enable_ip
= _THIS_IP_
;
1265 p
->softirq_enable_event
= 0;
1266 p
->softirq_disable_ip
= 0;
1267 p
->softirq_disable_event
= 0;
1268 p
->hardirq_context
= 0;
1269 p
->softirq_context
= 0;
1271 #ifdef CONFIG_LOCKDEP
1272 p
->lockdep_depth
= 0; /* no locks held yet */
1273 p
->curr_chain_key
= 0;
1274 p
->lockdep_recursion
= 0;
1277 #ifdef CONFIG_DEBUG_MUTEXES
1278 p
->blocked_on
= NULL
; /* not blocked yet */
1281 p
->memcg_batch
.do_batch
= 0;
1282 p
->memcg_batch
.memcg
= NULL
;
1285 /* Perform scheduler related setup. Assign this task to a CPU. */
1288 retval
= perf_event_init_task(p
);
1290 goto bad_fork_cleanup_policy
;
1291 retval
= audit_alloc(p
);
1293 goto bad_fork_cleanup_policy
;
1294 /* copy all the process information */
1295 retval
= copy_semundo(clone_flags
, p
);
1297 goto bad_fork_cleanup_audit
;
1298 retval
= copy_files(clone_flags
, p
);
1300 goto bad_fork_cleanup_semundo
;
1301 retval
= copy_fs(clone_flags
, p
);
1303 goto bad_fork_cleanup_files
;
1304 retval
= copy_sighand(clone_flags
, p
);
1306 goto bad_fork_cleanup_fs
;
1307 retval
= copy_signal(clone_flags
, p
);
1309 goto bad_fork_cleanup_sighand
;
1310 retval
= copy_mm(clone_flags
, p
);
1312 goto bad_fork_cleanup_signal
;
1313 retval
= copy_namespaces(clone_flags
, p
);
1315 goto bad_fork_cleanup_mm
;
1316 retval
= copy_io(clone_flags
, p
);
1318 goto bad_fork_cleanup_namespaces
;
1319 retval
= copy_thread(clone_flags
, stack_start
, stack_size
, p
, regs
);
1321 goto bad_fork_cleanup_io
;
1323 if (pid
!= &init_struct_pid
) {
1325 pid
= alloc_pid(p
->nsproxy
->pid_ns
);
1327 goto bad_fork_cleanup_io
;
1330 p
->pid
= pid_nr(pid
);
1332 if (clone_flags
& CLONE_THREAD
)
1333 p
->tgid
= current
->tgid
;
1335 p
->set_child_tid
= (clone_flags
& CLONE_CHILD_SETTID
) ? child_tidptr
: NULL
;
1337 * Clear TID on mm_release()?
1339 p
->clear_child_tid
= (clone_flags
& CLONE_CHILD_CLEARTID
) ? child_tidptr
: NULL
;
1344 p
->robust_list
= NULL
;
1345 #ifdef CONFIG_COMPAT
1346 p
->compat_robust_list
= NULL
;
1348 INIT_LIST_HEAD(&p
->pi_state_list
);
1349 p
->pi_state_cache
= NULL
;
1351 uprobe_copy_process(p
);
1353 * sigaltstack should be cleared when sharing the same VM
1355 if ((clone_flags
& (CLONE_VM
|CLONE_VFORK
)) == CLONE_VM
)
1356 p
->sas_ss_sp
= p
->sas_ss_size
= 0;
1359 * Syscall tracing and stepping should be turned off in the
1360 * child regardless of CLONE_PTRACE.
1362 user_disable_single_step(p
);
1363 clear_tsk_thread_flag(p
, TIF_SYSCALL_TRACE
);
1364 #ifdef TIF_SYSCALL_EMU
1365 clear_tsk_thread_flag(p
, TIF_SYSCALL_EMU
);
1367 clear_all_latency_tracing(p
);
1369 /* ok, now we should be set up.. */
1370 if (clone_flags
& CLONE_THREAD
)
1371 p
->exit_signal
= -1;
1372 else if (clone_flags
& CLONE_PARENT
)
1373 p
->exit_signal
= current
->group_leader
->exit_signal
;
1375 p
->exit_signal
= (clone_flags
& CSIGNAL
);
1377 p
->pdeath_signal
= 0;
1381 p
->nr_dirtied_pause
= 128 >> (PAGE_SHIFT
- 10);
1382 p
->dirty_paused_when
= 0;
1385 * Ok, make it visible to the rest of the system.
1386 * We dont wake it up yet.
1388 p
->group_leader
= p
;
1389 INIT_LIST_HEAD(&p
->thread_group
);
1390 p
->task_works
= NULL
;
1392 /* Now that the task is set up, run cgroup callbacks if
1393 * necessary. We need to run them before the task is visible
1394 * on the tasklist. */
1395 cgroup_fork_callbacks(p
);
1396 cgroup_callbacks_done
= 1;
1398 /* Need tasklist lock for parent etc handling! */
1399 write_lock_irq(&tasklist_lock
);
1401 /* CLONE_PARENT re-uses the old parent */
1402 if (clone_flags
& (CLONE_PARENT
|CLONE_THREAD
)) {
1403 p
->real_parent
= current
->real_parent
;
1404 p
->parent_exec_id
= current
->parent_exec_id
;
1406 p
->real_parent
= current
;
1407 p
->parent_exec_id
= current
->self_exec_id
;
1410 spin_lock(¤t
->sighand
->siglock
);
1413 * Process group and session signals need to be delivered to just the
1414 * parent before the fork or both the parent and the child after the
1415 * fork. Restart if a signal comes in before we add the new process to
1416 * it's process group.
1417 * A fatal signal pending means that current will exit, so the new
1418 * thread can't slip out of an OOM kill (or normal SIGKILL).
1420 recalc_sigpending();
1421 if (signal_pending(current
)) {
1422 spin_unlock(¤t
->sighand
->siglock
);
1423 write_unlock_irq(&tasklist_lock
);
1424 retval
= -ERESTARTNOINTR
;
1425 goto bad_fork_free_pid
;
1428 if (clone_flags
& CLONE_THREAD
) {
1429 current
->signal
->nr_threads
++;
1430 atomic_inc(¤t
->signal
->live
);
1431 atomic_inc(¤t
->signal
->sigcnt
);
1432 p
->group_leader
= current
->group_leader
;
1433 list_add_tail_rcu(&p
->thread_group
, &p
->group_leader
->thread_group
);
1436 if (likely(p
->pid
)) {
1437 ptrace_init_task(p
, (clone_flags
& CLONE_PTRACE
) || trace
);
1439 if (thread_group_leader(p
)) {
1440 if (is_child_reaper(pid
))
1441 p
->nsproxy
->pid_ns
->child_reaper
= p
;
1443 p
->signal
->leader_pid
= pid
;
1444 p
->signal
->tty
= tty_kref_get(current
->signal
->tty
);
1445 attach_pid(p
, PIDTYPE_PGID
, task_pgrp(current
));
1446 attach_pid(p
, PIDTYPE_SID
, task_session(current
));
1447 list_add_tail(&p
->sibling
, &p
->real_parent
->children
);
1448 list_add_tail_rcu(&p
->tasks
, &init_task
.tasks
);
1449 __this_cpu_inc(process_counts
);
1451 attach_pid(p
, PIDTYPE_PID
, pid
);
1456 spin_unlock(¤t
->sighand
->siglock
);
1457 write_unlock_irq(&tasklist_lock
);
1458 proc_fork_connector(p
);
1459 cgroup_post_fork(p
);
1460 if (clone_flags
& CLONE_THREAD
)
1461 threadgroup_change_end(current
);
1464 trace_task_newtask(p
, clone_flags
);
1469 if (pid
!= &init_struct_pid
)
1471 bad_fork_cleanup_io
:
1474 bad_fork_cleanup_namespaces
:
1475 if (unlikely(clone_flags
& CLONE_NEWPID
))
1476 pid_ns_release_proc(p
->nsproxy
->pid_ns
);
1477 exit_task_namespaces(p
);
1478 bad_fork_cleanup_mm
:
1481 bad_fork_cleanup_signal
:
1482 if (!(clone_flags
& CLONE_THREAD
))
1483 free_signal_struct(p
->signal
);
1484 bad_fork_cleanup_sighand
:
1485 __cleanup_sighand(p
->sighand
);
1486 bad_fork_cleanup_fs
:
1487 exit_fs(p
); /* blocking */
1488 bad_fork_cleanup_files
:
1489 exit_files(p
); /* blocking */
1490 bad_fork_cleanup_semundo
:
1492 bad_fork_cleanup_audit
:
1494 bad_fork_cleanup_policy
:
1495 perf_event_free_task(p
);
1497 mpol_put(p
->mempolicy
);
1498 bad_fork_cleanup_cgroup
:
1500 if (clone_flags
& CLONE_THREAD
)
1501 threadgroup_change_end(current
);
1502 cgroup_exit(p
, cgroup_callbacks_done
);
1503 delayacct_tsk_free(p
);
1504 module_put(task_thread_info(p
)->exec_domain
->module
);
1505 bad_fork_cleanup_count
:
1506 atomic_dec(&p
->cred
->user
->processes
);
1511 return ERR_PTR(retval
);
1514 noinline
struct pt_regs
* __cpuinit
__attribute__((weak
)) idle_regs(struct pt_regs
*regs
)
1516 memset(regs
, 0, sizeof(struct pt_regs
));
1520 static inline void init_idle_pids(struct pid_link
*links
)
1524 for (type
= PIDTYPE_PID
; type
< PIDTYPE_MAX
; ++type
) {
1525 INIT_HLIST_NODE(&links
[type
].node
); /* not really needed */
1526 links
[type
].pid
= &init_struct_pid
;
1530 struct task_struct
* __cpuinit
fork_idle(int cpu
)
1532 struct task_struct
*task
;
1533 struct pt_regs regs
;
1535 task
= copy_process(CLONE_VM
, 0, idle_regs(®s
), 0, NULL
,
1536 &init_struct_pid
, 0);
1537 if (!IS_ERR(task
)) {
1538 init_idle_pids(task
->pids
);
1539 init_idle(task
, cpu
);
1546 * Ok, this is the main fork-routine.
1548 * It copies the process, and if successful kick-starts
1549 * it and waits for it to finish using the VM if required.
1551 long do_fork(unsigned long clone_flags
,
1552 unsigned long stack_start
,
1553 struct pt_regs
*regs
,
1554 unsigned long stack_size
,
1555 int __user
*parent_tidptr
,
1556 int __user
*child_tidptr
)
1558 struct task_struct
*p
;
1563 * Do some preliminary argument and permissions checking before we
1564 * actually start allocating stuff
1566 if (clone_flags
& CLONE_NEWUSER
) {
1567 if (clone_flags
& CLONE_THREAD
)
1569 /* hopefully this check will go away when userns support is
1572 if (!capable(CAP_SYS_ADMIN
) || !capable(CAP_SETUID
) ||
1573 !capable(CAP_SETGID
))
1578 * Determine whether and which event to report to ptracer. When
1579 * called from kernel_thread or CLONE_UNTRACED is explicitly
1580 * requested, no event is reported; otherwise, report if the event
1581 * for the type of forking is enabled.
1583 if (likely(user_mode(regs
)) && !(clone_flags
& CLONE_UNTRACED
)) {
1584 if (clone_flags
& CLONE_VFORK
)
1585 trace
= PTRACE_EVENT_VFORK
;
1586 else if ((clone_flags
& CSIGNAL
) != SIGCHLD
)
1587 trace
= PTRACE_EVENT_CLONE
;
1589 trace
= PTRACE_EVENT_FORK
;
1591 if (likely(!ptrace_event_enabled(current
, trace
)))
1595 p
= copy_process(clone_flags
, stack_start
, regs
, stack_size
,
1596 child_tidptr
, NULL
, trace
);
1598 * Do this prior waking up the new thread - the thread pointer
1599 * might get invalid after that point, if the thread exits quickly.
1602 struct completion vfork
;
1604 trace_sched_process_fork(current
, p
);
1606 nr
= task_pid_vnr(p
);
1608 if (clone_flags
& CLONE_PARENT_SETTID
)
1609 put_user(nr
, parent_tidptr
);
1611 if (clone_flags
& CLONE_VFORK
) {
1612 p
->vfork_done
= &vfork
;
1613 init_completion(&vfork
);
1617 wake_up_new_task(p
);
1619 /* forking complete and child started to run, tell ptracer */
1620 if (unlikely(trace
))
1621 ptrace_event(trace
, nr
);
1623 if (clone_flags
& CLONE_VFORK
) {
1624 if (!wait_for_vfork_done(p
, &vfork
))
1625 ptrace_event(PTRACE_EVENT_VFORK_DONE
, nr
);
1633 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1634 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1637 static void sighand_ctor(void *data
)
1639 struct sighand_struct
*sighand
= data
;
1641 spin_lock_init(&sighand
->siglock
);
1642 init_waitqueue_head(&sighand
->signalfd_wqh
);
1645 void __init
proc_caches_init(void)
1647 sighand_cachep
= kmem_cache_create("sighand_cache",
1648 sizeof(struct sighand_struct
), 0,
1649 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_DESTROY_BY_RCU
|
1650 SLAB_NOTRACK
, sighand_ctor
);
1651 signal_cachep
= kmem_cache_create("signal_cache",
1652 sizeof(struct signal_struct
), 0,
1653 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1654 files_cachep
= kmem_cache_create("files_cache",
1655 sizeof(struct files_struct
), 0,
1656 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1657 fs_cachep
= kmem_cache_create("fs_cache",
1658 sizeof(struct fs_struct
), 0,
1659 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1661 * FIXME! The "sizeof(struct mm_struct)" currently includes the
1662 * whole struct cpumask for the OFFSTACK case. We could change
1663 * this to *only* allocate as much of it as required by the
1664 * maximum number of CPU's we can ever have. The cpumask_allocation
1665 * is at the end of the structure, exactly for that reason.
1667 mm_cachep
= kmem_cache_create("mm_struct",
1668 sizeof(struct mm_struct
), ARCH_MIN_MMSTRUCT_ALIGN
,
1669 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1670 vm_area_cachep
= KMEM_CACHE(vm_area_struct
, SLAB_PANIC
);
1672 nsproxy_cache_init();
1676 * Check constraints on flags passed to the unshare system call.
1678 static int check_unshare_flags(unsigned long unshare_flags
)
1680 if (unshare_flags
& ~(CLONE_THREAD
|CLONE_FS
|CLONE_NEWNS
|CLONE_SIGHAND
|
1681 CLONE_VM
|CLONE_FILES
|CLONE_SYSVSEM
|
1682 CLONE_NEWUTS
|CLONE_NEWIPC
|CLONE_NEWNET
))
1685 * Not implemented, but pretend it works if there is nothing to
1686 * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND
1687 * needs to unshare vm.
1689 if (unshare_flags
& (CLONE_THREAD
| CLONE_SIGHAND
| CLONE_VM
)) {
1690 /* FIXME: get_task_mm() increments ->mm_users */
1691 if (atomic_read(¤t
->mm
->mm_users
) > 1)
1699 * Unshare the filesystem structure if it is being shared
1701 static int unshare_fs(unsigned long unshare_flags
, struct fs_struct
**new_fsp
)
1703 struct fs_struct
*fs
= current
->fs
;
1705 if (!(unshare_flags
& CLONE_FS
) || !fs
)
1708 /* don't need lock here; in the worst case we'll do useless copy */
1712 *new_fsp
= copy_fs_struct(fs
);
1720 * Unshare file descriptor table if it is being shared
1722 static int unshare_fd(unsigned long unshare_flags
, struct files_struct
**new_fdp
)
1724 struct files_struct
*fd
= current
->files
;
1727 if ((unshare_flags
& CLONE_FILES
) &&
1728 (fd
&& atomic_read(&fd
->count
) > 1)) {
1729 *new_fdp
= dup_fd(fd
, &error
);
1738 * unshare allows a process to 'unshare' part of the process
1739 * context which was originally shared using clone. copy_*
1740 * functions used by do_fork() cannot be used here directly
1741 * because they modify an inactive task_struct that is being
1742 * constructed. Here we are modifying the current, active,
1745 SYSCALL_DEFINE1(unshare
, unsigned long, unshare_flags
)
1747 struct fs_struct
*fs
, *new_fs
= NULL
;
1748 struct files_struct
*fd
, *new_fd
= NULL
;
1749 struct nsproxy
*new_nsproxy
= NULL
;
1753 err
= check_unshare_flags(unshare_flags
);
1755 goto bad_unshare_out
;
1758 * If unsharing namespace, must also unshare filesystem information.
1760 if (unshare_flags
& CLONE_NEWNS
)
1761 unshare_flags
|= CLONE_FS
;
1763 * CLONE_NEWIPC must also detach from the undolist: after switching
1764 * to a new ipc namespace, the semaphore arrays from the old
1765 * namespace are unreachable.
1767 if (unshare_flags
& (CLONE_NEWIPC
|CLONE_SYSVSEM
))
1769 err
= unshare_fs(unshare_flags
, &new_fs
);
1771 goto bad_unshare_out
;
1772 err
= unshare_fd(unshare_flags
, &new_fd
);
1774 goto bad_unshare_cleanup_fs
;
1775 err
= unshare_nsproxy_namespaces(unshare_flags
, &new_nsproxy
, new_fs
);
1777 goto bad_unshare_cleanup_fd
;
1779 if (new_fs
|| new_fd
|| do_sysvsem
|| new_nsproxy
) {
1782 * CLONE_SYSVSEM is equivalent to sys_exit().
1788 switch_task_namespaces(current
, new_nsproxy
);
1796 spin_lock(&fs
->lock
);
1797 current
->fs
= new_fs
;
1802 spin_unlock(&fs
->lock
);
1806 fd
= current
->files
;
1807 current
->files
= new_fd
;
1811 task_unlock(current
);
1815 put_nsproxy(new_nsproxy
);
1817 bad_unshare_cleanup_fd
:
1819 put_files_struct(new_fd
);
1821 bad_unshare_cleanup_fs
:
1823 free_fs_struct(new_fs
);
1830 * Helper to unshare the files of the current task.
1831 * We don't want to expose copy_files internals to
1832 * the exec layer of the kernel.
1835 int unshare_files(struct files_struct
**displaced
)
1837 struct task_struct
*task
= current
;
1838 struct files_struct
*copy
= NULL
;
1841 error
= unshare_fd(CLONE_FILES
, ©
);
1842 if (error
|| !copy
) {
1846 *displaced
= task
->files
;