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>
32 #include <linux/vmacache.h>
33 #include <linux/nsproxy.h>
34 #include <linux/capability.h>
35 #include <linux/cpu.h>
36 #include <linux/cgroup.h>
37 #include <linux/security.h>
38 #include <linux/hugetlb.h>
39 #include <linux/seccomp.h>
40 #include <linux/swap.h>
41 #include <linux/syscalls.h>
42 #include <linux/jiffies.h>
43 #include <linux/futex.h>
44 #include <linux/compat.h>
45 #include <linux/kthread.h>
46 #include <linux/task_io_accounting_ops.h>
47 #include <linux/rcupdate.h>
48 #include <linux/ptrace.h>
49 #include <linux/mount.h>
50 #include <linux/audit.h>
51 #include <linux/memcontrol.h>
52 #include <linux/ftrace.h>
53 #include <linux/proc_fs.h>
54 #include <linux/profile.h>
55 #include <linux/rmap.h>
56 #include <linux/ksm.h>
57 #include <linux/acct.h>
58 #include <linux/tsacct_kern.h>
59 #include <linux/cn_proc.h>
60 #include <linux/freezer.h>
61 #include <linux/delayacct.h>
62 #include <linux/taskstats_kern.h>
63 #include <linux/random.h>
64 #include <linux/tty.h>
65 #include <linux/blkdev.h>
66 #include <linux/fs_struct.h>
67 #include <linux/magic.h>
68 #include <linux/perf_event.h>
69 #include <linux/posix-timers.h>
70 #include <linux/user-return-notifier.h>
71 #include <linux/oom.h>
72 #include <linux/khugepaged.h>
73 #include <linux/signalfd.h>
74 #include <linux/uprobes.h>
75 #include <linux/aio.h>
76 #include <linux/compiler.h>
78 #include <asm/pgtable.h>
79 #include <asm/pgalloc.h>
80 #include <asm/uaccess.h>
81 #include <asm/mmu_context.h>
82 #include <asm/cacheflush.h>
83 #include <asm/tlbflush.h>
85 #include <trace/events/sched.h>
87 #define CREATE_TRACE_POINTS
88 #include <trace/events/task.h>
91 * Protected counters by write_lock_irq(&tasklist_lock)
93 unsigned long total_forks
; /* Handle normal Linux uptimes. */
94 int nr_threads
; /* The idle threads do not count.. */
96 int max_threads
; /* tunable limit on nr_threads */
98 DEFINE_PER_CPU(unsigned long, process_counts
) = 0;
100 __cacheline_aligned
DEFINE_RWLOCK(tasklist_lock
); /* outer */
102 #ifdef CONFIG_PROVE_RCU
103 int lockdep_tasklist_lock_is_held(void)
105 return lockdep_is_held(&tasklist_lock
);
107 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held
);
108 #endif /* #ifdef CONFIG_PROVE_RCU */
110 int nr_processes(void)
115 for_each_possible_cpu(cpu
)
116 total
+= per_cpu(process_counts
, cpu
);
121 void __weak
arch_release_task_struct(struct task_struct
*tsk
)
125 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
126 static struct kmem_cache
*task_struct_cachep
;
128 static inline struct task_struct
*alloc_task_struct_node(int node
)
130 return kmem_cache_alloc_node(task_struct_cachep
, GFP_KERNEL
, node
);
133 static inline void free_task_struct(struct task_struct
*tsk
)
135 kmem_cache_free(task_struct_cachep
, tsk
);
139 void __weak
arch_release_thread_info(struct thread_info
*ti
)
143 #ifndef CONFIG_ARCH_THREAD_INFO_ALLOCATOR
146 * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a
147 * kmemcache based allocator.
149 # if THREAD_SIZE >= PAGE_SIZE
150 static struct thread_info
*alloc_thread_info_node(struct task_struct
*tsk
,
153 struct page
*page
= alloc_kmem_pages_node(node
, THREADINFO_GFP
,
156 return page
? page_address(page
) : NULL
;
159 static inline void free_thread_info(struct thread_info
*ti
)
161 free_kmem_pages((unsigned long)ti
, THREAD_SIZE_ORDER
);
164 static struct kmem_cache
*thread_info_cache
;
166 static struct thread_info
*alloc_thread_info_node(struct task_struct
*tsk
,
169 return kmem_cache_alloc_node(thread_info_cache
, THREADINFO_GFP
, node
);
172 static void free_thread_info(struct thread_info
*ti
)
174 kmem_cache_free(thread_info_cache
, ti
);
177 void thread_info_cache_init(void)
179 thread_info_cache
= kmem_cache_create("thread_info", THREAD_SIZE
,
180 THREAD_SIZE
, 0, NULL
);
181 BUG_ON(thread_info_cache
== NULL
);
186 /* SLAB cache for signal_struct structures (tsk->signal) */
187 static struct kmem_cache
*signal_cachep
;
189 /* SLAB cache for sighand_struct structures (tsk->sighand) */
190 struct kmem_cache
*sighand_cachep
;
192 /* SLAB cache for files_struct structures (tsk->files) */
193 struct kmem_cache
*files_cachep
;
195 /* SLAB cache for fs_struct structures (tsk->fs) */
196 struct kmem_cache
*fs_cachep
;
198 /* SLAB cache for vm_area_struct structures */
199 struct kmem_cache
*vm_area_cachep
;
201 /* SLAB cache for mm_struct structures (tsk->mm) */
202 static struct kmem_cache
*mm_cachep
;
204 static void account_kernel_stack(struct thread_info
*ti
, int account
)
206 struct zone
*zone
= page_zone(virt_to_page(ti
));
208 mod_zone_page_state(zone
, NR_KERNEL_STACK
, account
);
211 void free_task(struct task_struct
*tsk
)
213 account_kernel_stack(tsk
->stack
, -1);
214 arch_release_thread_info(tsk
->stack
);
215 free_thread_info(tsk
->stack
);
216 rt_mutex_debug_task_free(tsk
);
217 ftrace_graph_exit_task(tsk
);
218 put_seccomp_filter(tsk
);
219 arch_release_task_struct(tsk
);
220 free_task_struct(tsk
);
222 EXPORT_SYMBOL(free_task
);
224 static inline void free_signal_struct(struct signal_struct
*sig
)
226 taskstats_tgid_free(sig
);
227 sched_autogroup_exit(sig
);
228 kmem_cache_free(signal_cachep
, sig
);
231 static inline void put_signal_struct(struct signal_struct
*sig
)
233 if (atomic_dec_and_test(&sig
->sigcnt
))
234 free_signal_struct(sig
);
237 void __put_task_struct(struct task_struct
*tsk
)
239 WARN_ON(!tsk
->exit_state
);
240 WARN_ON(atomic_read(&tsk
->usage
));
241 WARN_ON(tsk
== current
);
244 security_task_free(tsk
);
246 delayacct_tsk_free(tsk
);
247 put_signal_struct(tsk
->signal
);
249 if (!profile_handoff_task(tsk
))
252 EXPORT_SYMBOL_GPL(__put_task_struct
);
254 void __init __weak
arch_task_cache_init(void) { }
256 void __init
fork_init(unsigned long mempages
)
258 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
259 #ifndef ARCH_MIN_TASKALIGN
260 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
262 /* create a slab on which task_structs can be allocated */
264 kmem_cache_create("task_struct", sizeof(struct task_struct
),
265 ARCH_MIN_TASKALIGN
, SLAB_PANIC
| SLAB_NOTRACK
, NULL
);
268 /* do the arch specific task caches init */
269 arch_task_cache_init();
272 * The default maximum number of threads is set to a safe
273 * value: the thread structures can take up at most half
276 max_threads
= mempages
/ (8 * THREAD_SIZE
/ PAGE_SIZE
);
279 * we need to allow at least 20 threads to boot a system
281 if (max_threads
< 20)
284 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_cur
= max_threads
/2;
285 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_max
= max_threads
/2;
286 init_task
.signal
->rlim
[RLIMIT_SIGPENDING
] =
287 init_task
.signal
->rlim
[RLIMIT_NPROC
];
290 int __weak
arch_dup_task_struct(struct task_struct
*dst
,
291 struct task_struct
*src
)
297 static struct task_struct
*dup_task_struct(struct task_struct
*orig
)
299 struct task_struct
*tsk
;
300 struct thread_info
*ti
;
301 unsigned long *stackend
;
302 int node
= tsk_fork_get_node(orig
);
305 tsk
= alloc_task_struct_node(node
);
309 ti
= alloc_thread_info_node(tsk
, node
);
313 err
= arch_dup_task_struct(tsk
, orig
);
318 #ifdef CONFIG_SECCOMP
320 * We must handle setting up seccomp filters once we're under
321 * the sighand lock in case orig has changed between now and
322 * then. Until then, filter must be NULL to avoid messing up
323 * the usage counts on the error path calling free_task.
325 tsk
->seccomp
.filter
= NULL
;
328 setup_thread_stack(tsk
, orig
);
329 clear_user_return_notifier(tsk
);
330 clear_tsk_need_resched(tsk
);
331 stackend
= end_of_stack(tsk
);
332 *stackend
= STACK_END_MAGIC
; /* for overflow detection */
334 #ifdef CONFIG_CC_STACKPROTECTOR
335 tsk
->stack_canary
= get_random_int();
339 * One for us, one for whoever does the "release_task()" (usually
342 atomic_set(&tsk
->usage
, 2);
343 #ifdef CONFIG_BLK_DEV_IO_TRACE
346 tsk
->splice_pipe
= NULL
;
347 tsk
->task_frag
.page
= NULL
;
349 account_kernel_stack(ti
, 1);
354 free_thread_info(ti
);
356 free_task_struct(tsk
);
361 static int dup_mmap(struct mm_struct
*mm
, struct mm_struct
*oldmm
)
363 struct vm_area_struct
*mpnt
, *tmp
, *prev
, **pprev
;
364 struct rb_node
**rb_link
, *rb_parent
;
366 unsigned long charge
;
368 uprobe_start_dup_mmap();
369 down_write(&oldmm
->mmap_sem
);
370 flush_cache_dup_mm(oldmm
);
371 uprobe_dup_mmap(oldmm
, mm
);
373 * Not linked in yet - no deadlock potential:
375 down_write_nested(&mm
->mmap_sem
, SINGLE_DEPTH_NESTING
);
377 mm
->total_vm
= oldmm
->total_vm
;
378 mm
->shared_vm
= oldmm
->shared_vm
;
379 mm
->exec_vm
= oldmm
->exec_vm
;
380 mm
->stack_vm
= oldmm
->stack_vm
;
382 rb_link
= &mm
->mm_rb
.rb_node
;
385 retval
= ksm_fork(mm
, oldmm
);
388 retval
= khugepaged_fork(mm
, oldmm
);
393 for (mpnt
= oldmm
->mmap
; mpnt
; mpnt
= mpnt
->vm_next
) {
396 if (mpnt
->vm_flags
& VM_DONTCOPY
) {
397 vm_stat_account(mm
, mpnt
->vm_flags
, mpnt
->vm_file
,
402 if (mpnt
->vm_flags
& VM_ACCOUNT
) {
403 unsigned long len
= vma_pages(mpnt
);
405 if (security_vm_enough_memory_mm(oldmm
, len
)) /* sic */
409 tmp
= kmem_cache_alloc(vm_area_cachep
, GFP_KERNEL
);
413 INIT_LIST_HEAD(&tmp
->anon_vma_chain
);
414 retval
= vma_dup_policy(mpnt
, tmp
);
416 goto fail_nomem_policy
;
418 if (anon_vma_fork(tmp
, mpnt
))
419 goto fail_nomem_anon_vma_fork
;
420 tmp
->vm_flags
&= ~VM_LOCKED
;
421 tmp
->vm_next
= tmp
->vm_prev
= NULL
;
424 struct inode
*inode
= file_inode(file
);
425 struct address_space
*mapping
= file
->f_mapping
;
428 if (tmp
->vm_flags
& VM_DENYWRITE
)
429 atomic_dec(&inode
->i_writecount
);
430 mutex_lock(&mapping
->i_mmap_mutex
);
431 if (tmp
->vm_flags
& VM_SHARED
)
432 atomic_inc(&mapping
->i_mmap_writable
);
433 flush_dcache_mmap_lock(mapping
);
434 /* insert tmp into the share list, just after mpnt */
435 if (unlikely(tmp
->vm_flags
& VM_NONLINEAR
))
436 vma_nonlinear_insert(tmp
,
437 &mapping
->i_mmap_nonlinear
);
439 vma_interval_tree_insert_after(tmp
, mpnt
,
441 flush_dcache_mmap_unlock(mapping
);
442 mutex_unlock(&mapping
->i_mmap_mutex
);
446 * Clear hugetlb-related page reserves for children. This only
447 * affects MAP_PRIVATE mappings. Faults generated by the child
448 * are not guaranteed to succeed, even if read-only
450 if (is_vm_hugetlb_page(tmp
))
451 reset_vma_resv_huge_pages(tmp
);
454 * Link in the new vma and copy the page table entries.
457 pprev
= &tmp
->vm_next
;
461 __vma_link_rb(mm
, tmp
, rb_link
, rb_parent
);
462 rb_link
= &tmp
->vm_rb
.rb_right
;
463 rb_parent
= &tmp
->vm_rb
;
466 retval
= copy_page_range(mm
, oldmm
, mpnt
);
468 if (tmp
->vm_ops
&& tmp
->vm_ops
->open
)
469 tmp
->vm_ops
->open(tmp
);
474 /* a new mm has just been created */
475 arch_dup_mmap(oldmm
, mm
);
478 up_write(&mm
->mmap_sem
);
480 up_write(&oldmm
->mmap_sem
);
481 uprobe_end_dup_mmap();
483 fail_nomem_anon_vma_fork
:
484 mpol_put(vma_policy(tmp
));
486 kmem_cache_free(vm_area_cachep
, tmp
);
489 vm_unacct_memory(charge
);
493 static inline int mm_alloc_pgd(struct mm_struct
*mm
)
495 mm
->pgd
= pgd_alloc(mm
);
496 if (unlikely(!mm
->pgd
))
501 static inline void mm_free_pgd(struct mm_struct
*mm
)
503 pgd_free(mm
, mm
->pgd
);
506 #define dup_mmap(mm, oldmm) (0)
507 #define mm_alloc_pgd(mm) (0)
508 #define mm_free_pgd(mm)
509 #endif /* CONFIG_MMU */
511 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(mmlist_lock
);
513 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
514 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
516 static unsigned long default_dump_filter
= MMF_DUMP_FILTER_DEFAULT
;
518 static int __init
coredump_filter_setup(char *s
)
520 default_dump_filter
=
521 (simple_strtoul(s
, NULL
, 0) << MMF_DUMP_FILTER_SHIFT
) &
522 MMF_DUMP_FILTER_MASK
;
526 __setup("coredump_filter=", coredump_filter_setup
);
528 #include <linux/init_task.h>
530 static void mm_init_aio(struct mm_struct
*mm
)
533 spin_lock_init(&mm
->ioctx_lock
);
534 mm
->ioctx_table
= NULL
;
538 static void mm_init_owner(struct mm_struct
*mm
, struct task_struct
*p
)
545 static struct mm_struct
*mm_init(struct mm_struct
*mm
, struct task_struct
*p
)
549 mm
->vmacache_seqnum
= 0;
550 atomic_set(&mm
->mm_users
, 1);
551 atomic_set(&mm
->mm_count
, 1);
552 init_rwsem(&mm
->mmap_sem
);
553 INIT_LIST_HEAD(&mm
->mmlist
);
554 mm
->core_state
= NULL
;
555 atomic_long_set(&mm
->nr_ptes
, 0);
559 memset(&mm
->rss_stat
, 0, sizeof(mm
->rss_stat
));
560 spin_lock_init(&mm
->page_table_lock
);
563 mm_init_owner(mm
, p
);
564 mmu_notifier_mm_init(mm
);
565 clear_tlb_flush_pending(mm
);
566 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
567 mm
->pmd_huge_pte
= NULL
;
571 mm
->flags
= current
->mm
->flags
& MMF_INIT_MASK
;
572 mm
->def_flags
= current
->mm
->def_flags
& VM_INIT_DEF_MASK
;
574 mm
->flags
= default_dump_filter
;
578 if (mm_alloc_pgd(mm
))
581 if (init_new_context(p
, mm
))
593 static void check_mm(struct mm_struct
*mm
)
597 for (i
= 0; i
< NR_MM_COUNTERS
; i
++) {
598 long x
= atomic_long_read(&mm
->rss_stat
.count
[i
]);
601 printk(KERN_ALERT
"BUG: Bad rss-counter state "
602 "mm:%p idx:%d val:%ld\n", mm
, i
, x
);
605 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
606 VM_BUG_ON(mm
->pmd_huge_pte
);
611 * Allocate and initialize an mm_struct.
613 struct mm_struct
*mm_alloc(void)
615 struct mm_struct
*mm
;
621 memset(mm
, 0, sizeof(*mm
));
622 return mm_init(mm
, current
);
626 * Called when the last reference to the mm
627 * is dropped: either by a lazy thread or by
628 * mmput. Free the page directory and the mm.
630 void __mmdrop(struct mm_struct
*mm
)
632 BUG_ON(mm
== &init_mm
);
635 mmu_notifier_mm_destroy(mm
);
639 EXPORT_SYMBOL_GPL(__mmdrop
);
642 * Decrement the use count and release all resources for an mm.
644 void mmput(struct mm_struct
*mm
)
648 if (atomic_dec_and_test(&mm
->mm_users
)) {
649 uprobe_clear_state(mm
);
652 khugepaged_exit(mm
); /* must run before exit_mmap */
654 set_mm_exe_file(mm
, NULL
);
655 if (!list_empty(&mm
->mmlist
)) {
656 spin_lock(&mmlist_lock
);
657 list_del(&mm
->mmlist
);
658 spin_unlock(&mmlist_lock
);
661 module_put(mm
->binfmt
->module
);
665 EXPORT_SYMBOL_GPL(mmput
);
667 void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
)
670 get_file(new_exe_file
);
673 mm
->exe_file
= new_exe_file
;
676 struct file
*get_mm_exe_file(struct mm_struct
*mm
)
678 struct file
*exe_file
;
680 /* We need mmap_sem to protect against races with removal of exe_file */
681 down_read(&mm
->mmap_sem
);
682 exe_file
= mm
->exe_file
;
685 up_read(&mm
->mmap_sem
);
689 static void dup_mm_exe_file(struct mm_struct
*oldmm
, struct mm_struct
*newmm
)
691 /* It's safe to write the exe_file pointer without exe_file_lock because
692 * this is called during fork when the task is not yet in /proc */
693 newmm
->exe_file
= get_mm_exe_file(oldmm
);
697 * get_task_mm - acquire a reference to the task's mm
699 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
700 * this kernel workthread has transiently adopted a user mm with use_mm,
701 * to do its AIO) is not set and if so returns a reference to it, after
702 * bumping up the use count. User must release the mm via mmput()
703 * after use. Typically used by /proc and ptrace.
705 struct mm_struct
*get_task_mm(struct task_struct
*task
)
707 struct mm_struct
*mm
;
712 if (task
->flags
& PF_KTHREAD
)
715 atomic_inc(&mm
->mm_users
);
720 EXPORT_SYMBOL_GPL(get_task_mm
);
722 struct mm_struct
*mm_access(struct task_struct
*task
, unsigned int mode
)
724 struct mm_struct
*mm
;
727 err
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
731 mm
= get_task_mm(task
);
732 if (mm
&& mm
!= current
->mm
&&
733 !ptrace_may_access(task
, mode
)) {
735 mm
= ERR_PTR(-EACCES
);
737 mutex_unlock(&task
->signal
->cred_guard_mutex
);
742 static void complete_vfork_done(struct task_struct
*tsk
)
744 struct completion
*vfork
;
747 vfork
= tsk
->vfork_done
;
749 tsk
->vfork_done
= NULL
;
755 static int wait_for_vfork_done(struct task_struct
*child
,
756 struct completion
*vfork
)
760 freezer_do_not_count();
761 killed
= wait_for_completion_killable(vfork
);
766 child
->vfork_done
= NULL
;
770 put_task_struct(child
);
774 /* Please note the differences between mmput and mm_release.
775 * mmput is called whenever we stop holding onto a mm_struct,
776 * error success whatever.
778 * mm_release is called after a mm_struct has been removed
779 * from the current process.
781 * This difference is important for error handling, when we
782 * only half set up a mm_struct for a new process and need to restore
783 * the old one. Because we mmput the new mm_struct before
784 * restoring the old one. . .
785 * Eric Biederman 10 January 1998
787 void mm_release(struct task_struct
*tsk
, struct mm_struct
*mm
)
789 /* Get rid of any futexes when releasing the mm */
791 if (unlikely(tsk
->robust_list
)) {
792 exit_robust_list(tsk
);
793 tsk
->robust_list
= NULL
;
796 if (unlikely(tsk
->compat_robust_list
)) {
797 compat_exit_robust_list(tsk
);
798 tsk
->compat_robust_list
= NULL
;
801 if (unlikely(!list_empty(&tsk
->pi_state_list
)))
802 exit_pi_state_list(tsk
);
805 uprobe_free_utask(tsk
);
807 /* Get rid of any cached register state */
808 deactivate_mm(tsk
, mm
);
811 * If we're exiting normally, clear a user-space tid field if
812 * requested. We leave this alone when dying by signal, to leave
813 * the value intact in a core dump, and to save the unnecessary
814 * trouble, say, a killed vfork parent shouldn't touch this mm.
815 * Userland only wants this done for a sys_exit.
817 if (tsk
->clear_child_tid
) {
818 if (!(tsk
->flags
& PF_SIGNALED
) &&
819 atomic_read(&mm
->mm_users
) > 1) {
821 * We don't check the error code - if userspace has
822 * not set up a proper pointer then tough luck.
824 put_user(0, tsk
->clear_child_tid
);
825 sys_futex(tsk
->clear_child_tid
, FUTEX_WAKE
,
828 tsk
->clear_child_tid
= NULL
;
832 * All done, finally we can wake up parent and return this mm to him.
833 * Also kthread_stop() uses this completion for synchronization.
836 complete_vfork_done(tsk
);
840 * Allocate a new mm structure and copy contents from the
841 * mm structure of the passed in task structure.
843 static struct mm_struct
*dup_mm(struct task_struct
*tsk
)
845 struct mm_struct
*mm
, *oldmm
= current
->mm
;
852 memcpy(mm
, oldmm
, sizeof(*mm
));
854 if (!mm_init(mm
, tsk
))
857 dup_mm_exe_file(oldmm
, mm
);
859 err
= dup_mmap(mm
, oldmm
);
863 mm
->hiwater_rss
= get_mm_rss(mm
);
864 mm
->hiwater_vm
= mm
->total_vm
;
866 if (mm
->binfmt
&& !try_module_get(mm
->binfmt
->module
))
872 /* don't put binfmt in mmput, we haven't got module yet */
880 static int copy_mm(unsigned long clone_flags
, struct task_struct
*tsk
)
882 struct mm_struct
*mm
, *oldmm
;
885 tsk
->min_flt
= tsk
->maj_flt
= 0;
886 tsk
->nvcsw
= tsk
->nivcsw
= 0;
887 #ifdef CONFIG_DETECT_HUNG_TASK
888 tsk
->last_switch_count
= tsk
->nvcsw
+ tsk
->nivcsw
;
892 tsk
->active_mm
= NULL
;
895 * Are we cloning a kernel thread?
897 * We need to steal a active VM for that..
903 /* initialize the new vmacache entries */
906 if (clone_flags
& CLONE_VM
) {
907 atomic_inc(&oldmm
->mm_users
);
926 static int copy_fs(unsigned long clone_flags
, struct task_struct
*tsk
)
928 struct fs_struct
*fs
= current
->fs
;
929 if (clone_flags
& CLONE_FS
) {
930 /* tsk->fs is already what we want */
931 spin_lock(&fs
->lock
);
933 spin_unlock(&fs
->lock
);
937 spin_unlock(&fs
->lock
);
940 tsk
->fs
= copy_fs_struct(fs
);
946 static int copy_files(unsigned long clone_flags
, struct task_struct
*tsk
)
948 struct files_struct
*oldf
, *newf
;
952 * A background process may not have any files ...
954 oldf
= current
->files
;
958 if (clone_flags
& CLONE_FILES
) {
959 atomic_inc(&oldf
->count
);
963 newf
= dup_fd(oldf
, &error
);
973 static int copy_io(unsigned long clone_flags
, struct task_struct
*tsk
)
976 struct io_context
*ioc
= current
->io_context
;
977 struct io_context
*new_ioc
;
982 * Share io context with parent, if CLONE_IO is set
984 if (clone_flags
& CLONE_IO
) {
986 tsk
->io_context
= ioc
;
987 } else if (ioprio_valid(ioc
->ioprio
)) {
988 new_ioc
= get_task_io_context(tsk
, GFP_KERNEL
, NUMA_NO_NODE
);
989 if (unlikely(!new_ioc
))
992 new_ioc
->ioprio
= ioc
->ioprio
;
993 put_io_context(new_ioc
);
999 static int copy_sighand(unsigned long clone_flags
, struct task_struct
*tsk
)
1001 struct sighand_struct
*sig
;
1003 if (clone_flags
& CLONE_SIGHAND
) {
1004 atomic_inc(¤t
->sighand
->count
);
1007 sig
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
1008 rcu_assign_pointer(tsk
->sighand
, sig
);
1011 atomic_set(&sig
->count
, 1);
1012 memcpy(sig
->action
, current
->sighand
->action
, sizeof(sig
->action
));
1016 void __cleanup_sighand(struct sighand_struct
*sighand
)
1018 if (atomic_dec_and_test(&sighand
->count
)) {
1019 signalfd_cleanup(sighand
);
1020 kmem_cache_free(sighand_cachep
, sighand
);
1026 * Initialize POSIX timer handling for a thread group.
1028 static void posix_cpu_timers_init_group(struct signal_struct
*sig
)
1030 unsigned long cpu_limit
;
1032 /* Thread group counters. */
1033 thread_group_cputime_init(sig
);
1035 cpu_limit
= ACCESS_ONCE(sig
->rlim
[RLIMIT_CPU
].rlim_cur
);
1036 if (cpu_limit
!= RLIM_INFINITY
) {
1037 sig
->cputime_expires
.prof_exp
= secs_to_cputime(cpu_limit
);
1038 sig
->cputimer
.running
= 1;
1041 /* The timer lists. */
1042 INIT_LIST_HEAD(&sig
->cpu_timers
[0]);
1043 INIT_LIST_HEAD(&sig
->cpu_timers
[1]);
1044 INIT_LIST_HEAD(&sig
->cpu_timers
[2]);
1047 static int copy_signal(unsigned long clone_flags
, struct task_struct
*tsk
)
1049 struct signal_struct
*sig
;
1051 if (clone_flags
& CLONE_THREAD
)
1054 sig
= kmem_cache_zalloc(signal_cachep
, GFP_KERNEL
);
1059 sig
->nr_threads
= 1;
1060 atomic_set(&sig
->live
, 1);
1061 atomic_set(&sig
->sigcnt
, 1);
1063 /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */
1064 sig
->thread_head
= (struct list_head
)LIST_HEAD_INIT(tsk
->thread_node
);
1065 tsk
->thread_node
= (struct list_head
)LIST_HEAD_INIT(sig
->thread_head
);
1067 init_waitqueue_head(&sig
->wait_chldexit
);
1068 sig
->curr_target
= tsk
;
1069 init_sigpending(&sig
->shared_pending
);
1070 INIT_LIST_HEAD(&sig
->posix_timers
);
1072 hrtimer_init(&sig
->real_timer
, CLOCK_MONOTONIC
, HRTIMER_MODE_REL
);
1073 sig
->real_timer
.function
= it_real_fn
;
1075 task_lock(current
->group_leader
);
1076 memcpy(sig
->rlim
, current
->signal
->rlim
, sizeof sig
->rlim
);
1077 task_unlock(current
->group_leader
);
1079 posix_cpu_timers_init_group(sig
);
1081 tty_audit_fork(sig
);
1082 sched_autogroup_fork(sig
);
1084 #ifdef CONFIG_CGROUPS
1085 init_rwsem(&sig
->group_rwsem
);
1088 sig
->oom_score_adj
= current
->signal
->oom_score_adj
;
1089 sig
->oom_score_adj_min
= current
->signal
->oom_score_adj_min
;
1091 sig
->has_child_subreaper
= current
->signal
->has_child_subreaper
||
1092 current
->signal
->is_child_subreaper
;
1094 mutex_init(&sig
->cred_guard_mutex
);
1099 static void copy_seccomp(struct task_struct
*p
)
1101 #ifdef CONFIG_SECCOMP
1103 * Must be called with sighand->lock held, which is common to
1104 * all threads in the group. Holding cred_guard_mutex is not
1105 * needed because this new task is not yet running and cannot
1108 assert_spin_locked(¤t
->sighand
->siglock
);
1110 /* Ref-count the new filter user, and assign it. */
1111 get_seccomp_filter(current
);
1112 p
->seccomp
= current
->seccomp
;
1115 * Explicitly enable no_new_privs here in case it got set
1116 * between the task_struct being duplicated and holding the
1117 * sighand lock. The seccomp state and nnp must be in sync.
1119 if (task_no_new_privs(current
))
1120 task_set_no_new_privs(p
);
1123 * If the parent gained a seccomp mode after copying thread
1124 * flags and between before we held the sighand lock, we have
1125 * to manually enable the seccomp thread flag here.
1127 if (p
->seccomp
.mode
!= SECCOMP_MODE_DISABLED
)
1128 set_tsk_thread_flag(p
, TIF_SECCOMP
);
1132 SYSCALL_DEFINE1(set_tid_address
, int __user
*, tidptr
)
1134 current
->clear_child_tid
= tidptr
;
1136 return task_pid_vnr(current
);
1139 static void rt_mutex_init_task(struct task_struct
*p
)
1141 raw_spin_lock_init(&p
->pi_lock
);
1142 #ifdef CONFIG_RT_MUTEXES
1143 p
->pi_waiters
= RB_ROOT
;
1144 p
->pi_waiters_leftmost
= NULL
;
1145 p
->pi_blocked_on
= NULL
;
1150 * Initialize POSIX timer handling for a single task.
1152 static void posix_cpu_timers_init(struct task_struct
*tsk
)
1154 tsk
->cputime_expires
.prof_exp
= 0;
1155 tsk
->cputime_expires
.virt_exp
= 0;
1156 tsk
->cputime_expires
.sched_exp
= 0;
1157 INIT_LIST_HEAD(&tsk
->cpu_timers
[0]);
1158 INIT_LIST_HEAD(&tsk
->cpu_timers
[1]);
1159 INIT_LIST_HEAD(&tsk
->cpu_timers
[2]);
1163 init_task_pid(struct task_struct
*task
, enum pid_type type
, struct pid
*pid
)
1165 task
->pids
[type
].pid
= pid
;
1169 * This creates a new process as a copy of the old one,
1170 * but does not actually start it yet.
1172 * It copies the registers, and all the appropriate
1173 * parts of the process environment (as per the clone
1174 * flags). The actual kick-off is left to the caller.
1176 static struct task_struct
*copy_process(unsigned long clone_flags
,
1177 unsigned long stack_start
,
1178 unsigned long stack_size
,
1179 int __user
*child_tidptr
,
1184 struct task_struct
*p
;
1186 if ((clone_flags
& (CLONE_NEWNS
|CLONE_FS
)) == (CLONE_NEWNS
|CLONE_FS
))
1187 return ERR_PTR(-EINVAL
);
1189 if ((clone_flags
& (CLONE_NEWUSER
|CLONE_FS
)) == (CLONE_NEWUSER
|CLONE_FS
))
1190 return ERR_PTR(-EINVAL
);
1193 * Thread groups must share signals as well, and detached threads
1194 * can only be started up within the thread group.
1196 if ((clone_flags
& CLONE_THREAD
) && !(clone_flags
& CLONE_SIGHAND
))
1197 return ERR_PTR(-EINVAL
);
1200 * Shared signal handlers imply shared VM. By way of the above,
1201 * thread groups also imply shared VM. Blocking this case allows
1202 * for various simplifications in other code.
1204 if ((clone_flags
& CLONE_SIGHAND
) && !(clone_flags
& CLONE_VM
))
1205 return ERR_PTR(-EINVAL
);
1208 * Siblings of global init remain as zombies on exit since they are
1209 * not reaped by their parent (swapper). To solve this and to avoid
1210 * multi-rooted process trees, prevent global and container-inits
1211 * from creating siblings.
1213 if ((clone_flags
& CLONE_PARENT
) &&
1214 current
->signal
->flags
& SIGNAL_UNKILLABLE
)
1215 return ERR_PTR(-EINVAL
);
1218 * If the new process will be in a different pid or user namespace
1219 * do not allow it to share a thread group or signal handlers or
1220 * parent with the forking task.
1222 if (clone_flags
& CLONE_SIGHAND
) {
1223 if ((clone_flags
& (CLONE_NEWUSER
| CLONE_NEWPID
)) ||
1224 (task_active_pid_ns(current
) !=
1225 current
->nsproxy
->pid_ns_for_children
))
1226 return ERR_PTR(-EINVAL
);
1229 retval
= security_task_create(clone_flags
);
1234 p
= dup_task_struct(current
);
1238 ftrace_graph_init_task(p
);
1240 rt_mutex_init_task(p
);
1242 #ifdef CONFIG_PROVE_LOCKING
1243 DEBUG_LOCKS_WARN_ON(!p
->hardirqs_enabled
);
1244 DEBUG_LOCKS_WARN_ON(!p
->softirqs_enabled
);
1247 if (atomic_read(&p
->real_cred
->user
->processes
) >=
1248 task_rlimit(p
, RLIMIT_NPROC
)) {
1249 if (p
->real_cred
->user
!= INIT_USER
&&
1250 !capable(CAP_SYS_RESOURCE
) && !capable(CAP_SYS_ADMIN
))
1253 current
->flags
&= ~PF_NPROC_EXCEEDED
;
1255 retval
= copy_creds(p
, clone_flags
);
1260 * If multiple threads are within copy_process(), then this check
1261 * triggers too late. This doesn't hurt, the check is only there
1262 * to stop root fork bombs.
1265 if (nr_threads
>= max_threads
)
1266 goto bad_fork_cleanup_count
;
1268 if (!try_module_get(task_thread_info(p
)->exec_domain
->module
))
1269 goto bad_fork_cleanup_count
;
1271 delayacct_tsk_init(p
); /* Must remain after dup_task_struct() */
1272 p
->flags
&= ~(PF_SUPERPRIV
| PF_WQ_WORKER
);
1273 p
->flags
|= PF_FORKNOEXEC
;
1274 INIT_LIST_HEAD(&p
->children
);
1275 INIT_LIST_HEAD(&p
->sibling
);
1276 rcu_copy_process(p
);
1277 p
->vfork_done
= NULL
;
1278 spin_lock_init(&p
->alloc_lock
);
1280 init_sigpending(&p
->pending
);
1282 p
->utime
= p
->stime
= p
->gtime
= 0;
1283 p
->utimescaled
= p
->stimescaled
= 0;
1284 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
1285 p
->prev_cputime
.utime
= p
->prev_cputime
.stime
= 0;
1287 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1288 seqlock_init(&p
->vtime_seqlock
);
1290 p
->vtime_snap_whence
= VTIME_SLEEPING
;
1293 #if defined(SPLIT_RSS_COUNTING)
1294 memset(&p
->rss_stat
, 0, sizeof(p
->rss_stat
));
1297 p
->default_timer_slack_ns
= current
->timer_slack_ns
;
1299 task_io_accounting_init(&p
->ioac
);
1300 acct_clear_integrals(p
);
1302 posix_cpu_timers_init(p
);
1304 p
->start_time
= ktime_get_ns();
1305 p
->real_start_time
= ktime_get_boot_ns();
1306 p
->io_context
= NULL
;
1307 p
->audit_context
= NULL
;
1308 if (clone_flags
& CLONE_THREAD
)
1309 threadgroup_change_begin(current
);
1312 p
->mempolicy
= mpol_dup(p
->mempolicy
);
1313 if (IS_ERR(p
->mempolicy
)) {
1314 retval
= PTR_ERR(p
->mempolicy
);
1315 p
->mempolicy
= NULL
;
1316 goto bad_fork_cleanup_threadgroup_lock
;
1319 #ifdef CONFIG_CPUSETS
1320 p
->cpuset_mem_spread_rotor
= NUMA_NO_NODE
;
1321 p
->cpuset_slab_spread_rotor
= NUMA_NO_NODE
;
1322 seqcount_init(&p
->mems_allowed_seq
);
1324 #ifdef CONFIG_TRACE_IRQFLAGS
1326 p
->hardirqs_enabled
= 0;
1327 p
->hardirq_enable_ip
= 0;
1328 p
->hardirq_enable_event
= 0;
1329 p
->hardirq_disable_ip
= _THIS_IP_
;
1330 p
->hardirq_disable_event
= 0;
1331 p
->softirqs_enabled
= 1;
1332 p
->softirq_enable_ip
= _THIS_IP_
;
1333 p
->softirq_enable_event
= 0;
1334 p
->softirq_disable_ip
= 0;
1335 p
->softirq_disable_event
= 0;
1336 p
->hardirq_context
= 0;
1337 p
->softirq_context
= 0;
1339 #ifdef CONFIG_LOCKDEP
1340 p
->lockdep_depth
= 0; /* no locks held yet */
1341 p
->curr_chain_key
= 0;
1342 p
->lockdep_recursion
= 0;
1345 #ifdef CONFIG_DEBUG_MUTEXES
1346 p
->blocked_on
= NULL
; /* not blocked yet */
1348 #ifdef CONFIG_BCACHE
1349 p
->sequential_io
= 0;
1350 p
->sequential_io_avg
= 0;
1353 /* Perform scheduler related setup. Assign this task to a CPU. */
1354 retval
= sched_fork(clone_flags
, p
);
1356 goto bad_fork_cleanup_policy
;
1358 retval
= perf_event_init_task(p
);
1360 goto bad_fork_cleanup_policy
;
1361 retval
= audit_alloc(p
);
1363 goto bad_fork_cleanup_policy
;
1364 /* copy all the process information */
1366 retval
= copy_semundo(clone_flags
, p
);
1368 goto bad_fork_cleanup_audit
;
1369 retval
= copy_files(clone_flags
, p
);
1371 goto bad_fork_cleanup_semundo
;
1372 retval
= copy_fs(clone_flags
, p
);
1374 goto bad_fork_cleanup_files
;
1375 retval
= copy_sighand(clone_flags
, p
);
1377 goto bad_fork_cleanup_fs
;
1378 retval
= copy_signal(clone_flags
, p
);
1380 goto bad_fork_cleanup_sighand
;
1381 retval
= copy_mm(clone_flags
, p
);
1383 goto bad_fork_cleanup_signal
;
1384 retval
= copy_namespaces(clone_flags
, p
);
1386 goto bad_fork_cleanup_mm
;
1387 retval
= copy_io(clone_flags
, p
);
1389 goto bad_fork_cleanup_namespaces
;
1390 retval
= copy_thread(clone_flags
, stack_start
, stack_size
, p
);
1392 goto bad_fork_cleanup_io
;
1394 if (pid
!= &init_struct_pid
) {
1396 pid
= alloc_pid(p
->nsproxy
->pid_ns_for_children
);
1398 goto bad_fork_cleanup_io
;
1401 p
->set_child_tid
= (clone_flags
& CLONE_CHILD_SETTID
) ? child_tidptr
: NULL
;
1403 * Clear TID on mm_release()?
1405 p
->clear_child_tid
= (clone_flags
& CLONE_CHILD_CLEARTID
) ? child_tidptr
: NULL
;
1410 p
->robust_list
= NULL
;
1411 #ifdef CONFIG_COMPAT
1412 p
->compat_robust_list
= NULL
;
1414 INIT_LIST_HEAD(&p
->pi_state_list
);
1415 p
->pi_state_cache
= NULL
;
1418 * sigaltstack should be cleared when sharing the same VM
1420 if ((clone_flags
& (CLONE_VM
|CLONE_VFORK
)) == CLONE_VM
)
1421 p
->sas_ss_sp
= p
->sas_ss_size
= 0;
1424 * Syscall tracing and stepping should be turned off in the
1425 * child regardless of CLONE_PTRACE.
1427 user_disable_single_step(p
);
1428 clear_tsk_thread_flag(p
, TIF_SYSCALL_TRACE
);
1429 #ifdef TIF_SYSCALL_EMU
1430 clear_tsk_thread_flag(p
, TIF_SYSCALL_EMU
);
1432 clear_all_latency_tracing(p
);
1434 /* ok, now we should be set up.. */
1435 p
->pid
= pid_nr(pid
);
1436 if (clone_flags
& CLONE_THREAD
) {
1437 p
->exit_signal
= -1;
1438 p
->group_leader
= current
->group_leader
;
1439 p
->tgid
= current
->tgid
;
1441 if (clone_flags
& CLONE_PARENT
)
1442 p
->exit_signal
= current
->group_leader
->exit_signal
;
1444 p
->exit_signal
= (clone_flags
& CSIGNAL
);
1445 p
->group_leader
= p
;
1450 p
->nr_dirtied_pause
= 128 >> (PAGE_SHIFT
- 10);
1451 p
->dirty_paused_when
= 0;
1453 p
->pdeath_signal
= 0;
1454 INIT_LIST_HEAD(&p
->thread_group
);
1455 p
->task_works
= NULL
;
1458 * Make it visible to the rest of the system, but dont wake it up yet.
1459 * Need tasklist lock for parent etc handling!
1461 write_lock_irq(&tasklist_lock
);
1463 /* CLONE_PARENT re-uses the old parent */
1464 if (clone_flags
& (CLONE_PARENT
|CLONE_THREAD
)) {
1465 p
->real_parent
= current
->real_parent
;
1466 p
->parent_exec_id
= current
->parent_exec_id
;
1468 p
->real_parent
= current
;
1469 p
->parent_exec_id
= current
->self_exec_id
;
1472 spin_lock(¤t
->sighand
->siglock
);
1475 * Copy seccomp details explicitly here, in case they were changed
1476 * before holding sighand lock.
1481 * Process group and session signals need to be delivered to just the
1482 * parent before the fork or both the parent and the child after the
1483 * fork. Restart if a signal comes in before we add the new process to
1484 * it's process group.
1485 * A fatal signal pending means that current will exit, so the new
1486 * thread can't slip out of an OOM kill (or normal SIGKILL).
1488 recalc_sigpending();
1489 if (signal_pending(current
)) {
1490 spin_unlock(¤t
->sighand
->siglock
);
1491 write_unlock_irq(&tasklist_lock
);
1492 retval
= -ERESTARTNOINTR
;
1493 goto bad_fork_free_pid
;
1496 if (likely(p
->pid
)) {
1497 ptrace_init_task(p
, (clone_flags
& CLONE_PTRACE
) || trace
);
1499 init_task_pid(p
, PIDTYPE_PID
, pid
);
1500 if (thread_group_leader(p
)) {
1501 init_task_pid(p
, PIDTYPE_PGID
, task_pgrp(current
));
1502 init_task_pid(p
, PIDTYPE_SID
, task_session(current
));
1504 if (is_child_reaper(pid
)) {
1505 ns_of_pid(pid
)->child_reaper
= p
;
1506 p
->signal
->flags
|= SIGNAL_UNKILLABLE
;
1509 p
->signal
->leader_pid
= pid
;
1510 p
->signal
->tty
= tty_kref_get(current
->signal
->tty
);
1511 list_add_tail(&p
->sibling
, &p
->real_parent
->children
);
1512 list_add_tail_rcu(&p
->tasks
, &init_task
.tasks
);
1513 attach_pid(p
, PIDTYPE_PGID
);
1514 attach_pid(p
, PIDTYPE_SID
);
1515 __this_cpu_inc(process_counts
);
1517 current
->signal
->nr_threads
++;
1518 atomic_inc(¤t
->signal
->live
);
1519 atomic_inc(¤t
->signal
->sigcnt
);
1520 list_add_tail_rcu(&p
->thread_group
,
1521 &p
->group_leader
->thread_group
);
1522 list_add_tail_rcu(&p
->thread_node
,
1523 &p
->signal
->thread_head
);
1525 attach_pid(p
, PIDTYPE_PID
);
1530 spin_unlock(¤t
->sighand
->siglock
);
1531 syscall_tracepoint_update(p
);
1532 write_unlock_irq(&tasklist_lock
);
1534 proc_fork_connector(p
);
1535 cgroup_post_fork(p
);
1536 if (clone_flags
& CLONE_THREAD
)
1537 threadgroup_change_end(current
);
1540 trace_task_newtask(p
, clone_flags
);
1541 uprobe_copy_process(p
, clone_flags
);
1546 if (pid
!= &init_struct_pid
)
1548 bad_fork_cleanup_io
:
1551 bad_fork_cleanup_namespaces
:
1552 exit_task_namespaces(p
);
1553 bad_fork_cleanup_mm
:
1556 bad_fork_cleanup_signal
:
1557 if (!(clone_flags
& CLONE_THREAD
))
1558 free_signal_struct(p
->signal
);
1559 bad_fork_cleanup_sighand
:
1560 __cleanup_sighand(p
->sighand
);
1561 bad_fork_cleanup_fs
:
1562 exit_fs(p
); /* blocking */
1563 bad_fork_cleanup_files
:
1564 exit_files(p
); /* blocking */
1565 bad_fork_cleanup_semundo
:
1567 bad_fork_cleanup_audit
:
1569 bad_fork_cleanup_policy
:
1570 perf_event_free_task(p
);
1572 mpol_put(p
->mempolicy
);
1573 bad_fork_cleanup_threadgroup_lock
:
1575 if (clone_flags
& CLONE_THREAD
)
1576 threadgroup_change_end(current
);
1577 delayacct_tsk_free(p
);
1578 module_put(task_thread_info(p
)->exec_domain
->module
);
1579 bad_fork_cleanup_count
:
1580 atomic_dec(&p
->cred
->user
->processes
);
1585 return ERR_PTR(retval
);
1588 static inline void init_idle_pids(struct pid_link
*links
)
1592 for (type
= PIDTYPE_PID
; type
< PIDTYPE_MAX
; ++type
) {
1593 INIT_HLIST_NODE(&links
[type
].node
); /* not really needed */
1594 links
[type
].pid
= &init_struct_pid
;
1598 struct task_struct
*fork_idle(int cpu
)
1600 struct task_struct
*task
;
1601 task
= copy_process(CLONE_VM
, 0, 0, NULL
, &init_struct_pid
, 0);
1602 if (!IS_ERR(task
)) {
1603 init_idle_pids(task
->pids
);
1604 init_idle(task
, cpu
);
1611 * Ok, this is the main fork-routine.
1613 * It copies the process, and if successful kick-starts
1614 * it and waits for it to finish using the VM if required.
1616 long do_fork(unsigned long clone_flags
,
1617 unsigned long stack_start
,
1618 unsigned long stack_size
,
1619 int __user
*parent_tidptr
,
1620 int __user
*child_tidptr
)
1622 struct task_struct
*p
;
1627 * Determine whether and which event to report to ptracer. When
1628 * called from kernel_thread or CLONE_UNTRACED is explicitly
1629 * requested, no event is reported; otherwise, report if the event
1630 * for the type of forking is enabled.
1632 if (!(clone_flags
& CLONE_UNTRACED
)) {
1633 if (clone_flags
& CLONE_VFORK
)
1634 trace
= PTRACE_EVENT_VFORK
;
1635 else if ((clone_flags
& CSIGNAL
) != SIGCHLD
)
1636 trace
= PTRACE_EVENT_CLONE
;
1638 trace
= PTRACE_EVENT_FORK
;
1640 if (likely(!ptrace_event_enabled(current
, trace
)))
1644 p
= copy_process(clone_flags
, stack_start
, stack_size
,
1645 child_tidptr
, NULL
, trace
);
1647 * Do this prior waking up the new thread - the thread pointer
1648 * might get invalid after that point, if the thread exits quickly.
1651 struct completion vfork
;
1654 trace_sched_process_fork(current
, p
);
1656 pid
= get_task_pid(p
, PIDTYPE_PID
);
1659 if (clone_flags
& CLONE_PARENT_SETTID
)
1660 put_user(nr
, parent_tidptr
);
1662 if (clone_flags
& CLONE_VFORK
) {
1663 p
->vfork_done
= &vfork
;
1664 init_completion(&vfork
);
1668 wake_up_new_task(p
);
1670 /* forking complete and child started to run, tell ptracer */
1671 if (unlikely(trace
))
1672 ptrace_event_pid(trace
, pid
);
1674 if (clone_flags
& CLONE_VFORK
) {
1675 if (!wait_for_vfork_done(p
, &vfork
))
1676 ptrace_event_pid(PTRACE_EVENT_VFORK_DONE
, pid
);
1687 * Create a kernel thread.
1689 pid_t
kernel_thread(int (*fn
)(void *), void *arg
, unsigned long flags
)
1691 return do_fork(flags
|CLONE_VM
|CLONE_UNTRACED
, (unsigned long)fn
,
1692 (unsigned long)arg
, NULL
, NULL
);
1695 #ifdef __ARCH_WANT_SYS_FORK
1696 SYSCALL_DEFINE0(fork
)
1699 return do_fork(SIGCHLD
, 0, 0, NULL
, NULL
);
1701 /* can not support in nommu mode */
1707 #ifdef __ARCH_WANT_SYS_VFORK
1708 SYSCALL_DEFINE0(vfork
)
1710 return do_fork(CLONE_VFORK
| CLONE_VM
| SIGCHLD
, 0,
1715 #ifdef __ARCH_WANT_SYS_CLONE
1716 #ifdef CONFIG_CLONE_BACKWARDS
1717 SYSCALL_DEFINE5(clone
, unsigned long, clone_flags
, unsigned long, newsp
,
1718 int __user
*, parent_tidptr
,
1720 int __user
*, child_tidptr
)
1721 #elif defined(CONFIG_CLONE_BACKWARDS2)
1722 SYSCALL_DEFINE5(clone
, unsigned long, newsp
, unsigned long, clone_flags
,
1723 int __user
*, parent_tidptr
,
1724 int __user
*, child_tidptr
,
1726 #elif defined(CONFIG_CLONE_BACKWARDS3)
1727 SYSCALL_DEFINE6(clone
, unsigned long, clone_flags
, unsigned long, newsp
,
1729 int __user
*, parent_tidptr
,
1730 int __user
*, child_tidptr
,
1733 SYSCALL_DEFINE5(clone
, unsigned long, clone_flags
, unsigned long, newsp
,
1734 int __user
*, parent_tidptr
,
1735 int __user
*, child_tidptr
,
1739 return do_fork(clone_flags
, newsp
, 0, parent_tidptr
, child_tidptr
);
1743 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1744 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1747 static void sighand_ctor(void *data
)
1749 struct sighand_struct
*sighand
= data
;
1751 spin_lock_init(&sighand
->siglock
);
1752 init_waitqueue_head(&sighand
->signalfd_wqh
);
1755 void __init
proc_caches_init(void)
1757 sighand_cachep
= kmem_cache_create("sighand_cache",
1758 sizeof(struct sighand_struct
), 0,
1759 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_DESTROY_BY_RCU
|
1760 SLAB_NOTRACK
, sighand_ctor
);
1761 signal_cachep
= kmem_cache_create("signal_cache",
1762 sizeof(struct signal_struct
), 0,
1763 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1764 files_cachep
= kmem_cache_create("files_cache",
1765 sizeof(struct files_struct
), 0,
1766 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1767 fs_cachep
= kmem_cache_create("fs_cache",
1768 sizeof(struct fs_struct
), 0,
1769 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1771 * FIXME! The "sizeof(struct mm_struct)" currently includes the
1772 * whole struct cpumask for the OFFSTACK case. We could change
1773 * this to *only* allocate as much of it as required by the
1774 * maximum number of CPU's we can ever have. The cpumask_allocation
1775 * is at the end of the structure, exactly for that reason.
1777 mm_cachep
= kmem_cache_create("mm_struct",
1778 sizeof(struct mm_struct
), ARCH_MIN_MMSTRUCT_ALIGN
,
1779 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1780 vm_area_cachep
= KMEM_CACHE(vm_area_struct
, SLAB_PANIC
);
1782 nsproxy_cache_init();
1786 * Check constraints on flags passed to the unshare system call.
1788 static int check_unshare_flags(unsigned long unshare_flags
)
1790 if (unshare_flags
& ~(CLONE_THREAD
|CLONE_FS
|CLONE_NEWNS
|CLONE_SIGHAND
|
1791 CLONE_VM
|CLONE_FILES
|CLONE_SYSVSEM
|
1792 CLONE_NEWUTS
|CLONE_NEWIPC
|CLONE_NEWNET
|
1793 CLONE_NEWUSER
|CLONE_NEWPID
))
1796 * Not implemented, but pretend it works if there is nothing to
1797 * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND
1798 * needs to unshare vm.
1800 if (unshare_flags
& (CLONE_THREAD
| CLONE_SIGHAND
| CLONE_VM
)) {
1801 /* FIXME: get_task_mm() increments ->mm_users */
1802 if (atomic_read(¤t
->mm
->mm_users
) > 1)
1810 * Unshare the filesystem structure if it is being shared
1812 static int unshare_fs(unsigned long unshare_flags
, struct fs_struct
**new_fsp
)
1814 struct fs_struct
*fs
= current
->fs
;
1816 if (!(unshare_flags
& CLONE_FS
) || !fs
)
1819 /* don't need lock here; in the worst case we'll do useless copy */
1823 *new_fsp
= copy_fs_struct(fs
);
1831 * Unshare file descriptor table if it is being shared
1833 static int unshare_fd(unsigned long unshare_flags
, struct files_struct
**new_fdp
)
1835 struct files_struct
*fd
= current
->files
;
1838 if ((unshare_flags
& CLONE_FILES
) &&
1839 (fd
&& atomic_read(&fd
->count
) > 1)) {
1840 *new_fdp
= dup_fd(fd
, &error
);
1849 * unshare allows a process to 'unshare' part of the process
1850 * context which was originally shared using clone. copy_*
1851 * functions used by do_fork() cannot be used here directly
1852 * because they modify an inactive task_struct that is being
1853 * constructed. Here we are modifying the current, active,
1856 SYSCALL_DEFINE1(unshare
, unsigned long, unshare_flags
)
1858 struct fs_struct
*fs
, *new_fs
= NULL
;
1859 struct files_struct
*fd
, *new_fd
= NULL
;
1860 struct cred
*new_cred
= NULL
;
1861 struct nsproxy
*new_nsproxy
= NULL
;
1866 * If unsharing a user namespace must also unshare the thread.
1868 if (unshare_flags
& CLONE_NEWUSER
)
1869 unshare_flags
|= CLONE_THREAD
| CLONE_FS
;
1871 * If unsharing a thread from a thread group, must also unshare vm.
1873 if (unshare_flags
& CLONE_THREAD
)
1874 unshare_flags
|= CLONE_VM
;
1876 * If unsharing vm, must also unshare signal handlers.
1878 if (unshare_flags
& CLONE_VM
)
1879 unshare_flags
|= CLONE_SIGHAND
;
1881 * If unsharing namespace, must also unshare filesystem information.
1883 if (unshare_flags
& CLONE_NEWNS
)
1884 unshare_flags
|= CLONE_FS
;
1886 err
= check_unshare_flags(unshare_flags
);
1888 goto bad_unshare_out
;
1890 * CLONE_NEWIPC must also detach from the undolist: after switching
1891 * to a new ipc namespace, the semaphore arrays from the old
1892 * namespace are unreachable.
1894 if (unshare_flags
& (CLONE_NEWIPC
|CLONE_SYSVSEM
))
1896 err
= unshare_fs(unshare_flags
, &new_fs
);
1898 goto bad_unshare_out
;
1899 err
= unshare_fd(unshare_flags
, &new_fd
);
1901 goto bad_unshare_cleanup_fs
;
1902 err
= unshare_userns(unshare_flags
, &new_cred
);
1904 goto bad_unshare_cleanup_fd
;
1905 err
= unshare_nsproxy_namespaces(unshare_flags
, &new_nsproxy
,
1908 goto bad_unshare_cleanup_cred
;
1910 if (new_fs
|| new_fd
|| do_sysvsem
|| new_cred
|| new_nsproxy
) {
1913 * CLONE_SYSVSEM is equivalent to sys_exit().
1917 if (unshare_flags
& CLONE_NEWIPC
) {
1918 /* Orphan segments in old ns (see sem above). */
1920 shm_init_task(current
);
1924 switch_task_namespaces(current
, new_nsproxy
);
1930 spin_lock(&fs
->lock
);
1931 current
->fs
= new_fs
;
1936 spin_unlock(&fs
->lock
);
1940 fd
= current
->files
;
1941 current
->files
= new_fd
;
1945 task_unlock(current
);
1948 /* Install the new user namespace */
1949 commit_creds(new_cred
);
1954 bad_unshare_cleanup_cred
:
1957 bad_unshare_cleanup_fd
:
1959 put_files_struct(new_fd
);
1961 bad_unshare_cleanup_fs
:
1963 free_fs_struct(new_fs
);
1970 * Helper to unshare the files of the current task.
1971 * We don't want to expose copy_files internals to
1972 * the exec layer of the kernel.
1975 int unshare_files(struct files_struct
**displaced
)
1977 struct task_struct
*task
= current
;
1978 struct files_struct
*copy
= NULL
;
1981 error
= unshare_fd(CLONE_FILES
, ©
);
1982 if (error
|| !copy
) {
1986 *displaced
= task
->files
;