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>
19 #include <linux/vmacache.h>
20 #include <linux/vmalloc.h>
21 #include <linux/completion.h>
22 #include <linux/personality.h>
23 #include <linux/mempolicy.h>
24 #include <linux/sem.h>
25 #include <linux/file.h>
26 #include <linux/fdtable.h>
27 #include <linux/iocontext.h>
28 #include <linux/key.h>
29 #include <linux/binfmts.h>
30 #include <linux/mman.h>
31 #include <linux/mmu_notifier.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>
77 #include <asm/pgtable.h>
78 #include <asm/pgalloc.h>
79 #include <asm/uaccess.h>
80 #include <asm/mmu_context.h>
81 #include <asm/cacheflush.h>
82 #include <asm/tlbflush.h>
84 #include <trace/events/sched.h>
86 #define CREATE_TRACE_POINTS
87 #include <trace/events/task.h>
90 * Protected counters by write_lock_irq(&tasklist_lock)
92 unsigned long total_forks
; /* Handle normal Linux uptimes. */
93 int nr_threads
; /* The idle threads do not count.. */
95 int max_threads
; /* tunable limit on nr_threads */
97 DEFINE_PER_CPU(unsigned long, process_counts
) = 0;
99 __cacheline_aligned
DEFINE_RWLOCK(tasklist_lock
); /* outer */
101 #ifdef CONFIG_PROVE_RCU
102 int lockdep_tasklist_lock_is_held(void)
104 return lockdep_is_held(&tasklist_lock
);
106 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held
);
107 #endif /* #ifdef CONFIG_PROVE_RCU */
109 int nr_processes(void)
114 for_each_possible_cpu(cpu
)
115 total
+= per_cpu(process_counts
, cpu
);
120 void __weak
arch_release_task_struct(struct task_struct
*tsk
)
124 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
125 static struct kmem_cache
*task_struct_cachep
;
127 static inline struct task_struct
*alloc_task_struct_node(int node
)
129 return kmem_cache_alloc_node(task_struct_cachep
, GFP_KERNEL
, node
);
132 static inline void free_task_struct(struct task_struct
*tsk
)
134 kmem_cache_free(task_struct_cachep
, tsk
);
138 void __weak
arch_release_thread_info(struct thread_info
*ti
)
142 #ifndef CONFIG_ARCH_THREAD_INFO_ALLOCATOR
145 * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a
146 * kmemcache based allocator.
148 # if THREAD_SIZE >= PAGE_SIZE
149 static struct thread_info
*alloc_thread_info_node(struct task_struct
*tsk
,
152 struct page
*page
= alloc_pages_node(node
, THREADINFO_GFP_ACCOUNTED
,
155 return page
? page_address(page
) : NULL
;
158 static inline void free_thread_info(struct thread_info
*ti
)
160 free_memcg_kmem_pages((unsigned long)ti
, THREAD_SIZE_ORDER
);
163 static struct kmem_cache
*thread_info_cache
;
165 static struct thread_info
*alloc_thread_info_node(struct task_struct
*tsk
,
168 return kmem_cache_alloc_node(thread_info_cache
, THREADINFO_GFP
, node
);
171 static void free_thread_info(struct thread_info
*ti
)
173 kmem_cache_free(thread_info_cache
, ti
);
176 void thread_info_cache_init(void)
178 thread_info_cache
= kmem_cache_create("thread_info", THREAD_SIZE
,
179 THREAD_SIZE
, 0, NULL
);
180 BUG_ON(thread_info_cache
== NULL
);
185 /* SLAB cache for signal_struct structures (tsk->signal) */
186 static struct kmem_cache
*signal_cachep
;
188 /* SLAB cache for sighand_struct structures (tsk->sighand) */
189 struct kmem_cache
*sighand_cachep
;
191 /* SLAB cache for files_struct structures (tsk->files) */
192 struct kmem_cache
*files_cachep
;
194 /* SLAB cache for fs_struct structures (tsk->fs) */
195 struct kmem_cache
*fs_cachep
;
197 /* SLAB cache for vm_area_struct structures */
198 struct kmem_cache
*vm_area_cachep
;
200 /* SLAB cache for mm_struct structures (tsk->mm) */
201 static struct kmem_cache
*mm_cachep
;
203 /* Notifier list called when a task struct is freed */
204 static ATOMIC_NOTIFIER_HEAD(task_free_notifier
);
206 static void account_kernel_stack(struct thread_info
*ti
, int account
)
208 struct zone
*zone
= page_zone(virt_to_page(ti
));
210 mod_zone_page_state(zone
, NR_KERNEL_STACK
, account
);
213 void free_task(struct task_struct
*tsk
)
215 account_kernel_stack(tsk
->stack
, -1);
216 arch_release_thread_info(tsk
->stack
);
217 free_thread_info(tsk
->stack
);
218 rt_mutex_debug_task_free(tsk
);
219 ftrace_graph_exit_task(tsk
);
220 put_seccomp_filter(tsk
);
221 arch_release_task_struct(tsk
);
222 free_task_struct(tsk
);
224 EXPORT_SYMBOL(free_task
);
226 static inline void free_signal_struct(struct signal_struct
*sig
)
228 taskstats_tgid_free(sig
);
229 sched_autogroup_exit(sig
);
230 kmem_cache_free(signal_cachep
, sig
);
233 static inline void put_signal_struct(struct signal_struct
*sig
)
235 if (atomic_dec_and_test(&sig
->sigcnt
))
236 free_signal_struct(sig
);
239 int task_free_register(struct notifier_block
*n
)
241 return atomic_notifier_chain_register(&task_free_notifier
, n
);
243 EXPORT_SYMBOL(task_free_register
);
245 int task_free_unregister(struct notifier_block
*n
)
247 return atomic_notifier_chain_unregister(&task_free_notifier
, n
);
249 EXPORT_SYMBOL(task_free_unregister
);
251 void __put_task_struct(struct task_struct
*tsk
)
253 WARN_ON(!tsk
->exit_state
);
254 WARN_ON(atomic_read(&tsk
->usage
));
255 WARN_ON(tsk
== current
);
257 security_task_free(tsk
);
259 delayacct_tsk_free(tsk
);
260 put_signal_struct(tsk
->signal
);
262 atomic_notifier_call_chain(&task_free_notifier
, 0, tsk
);
263 if (!profile_handoff_task(tsk
))
266 EXPORT_SYMBOL_GPL(__put_task_struct
);
268 void __init __weak
arch_task_cache_init(void) { }
270 void __init
fork_init(unsigned long mempages
)
272 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
273 #ifndef ARCH_MIN_TASKALIGN
274 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
276 /* create a slab on which task_structs can be allocated */
278 kmem_cache_create("task_struct", sizeof(struct task_struct
),
279 ARCH_MIN_TASKALIGN
, SLAB_PANIC
| SLAB_NOTRACK
, NULL
);
282 /* do the arch specific task caches init */
283 arch_task_cache_init();
286 * The default maximum number of threads is set to a safe
287 * value: the thread structures can take up at most half
290 max_threads
= mempages
/ (8 * THREAD_SIZE
/ PAGE_SIZE
);
293 * we need to allow at least 20 threads to boot a system
295 if (max_threads
< 20)
298 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_cur
= max_threads
/2;
299 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_max
= max_threads
/2;
300 init_task
.signal
->rlim
[RLIMIT_SIGPENDING
] =
301 init_task
.signal
->rlim
[RLIMIT_NPROC
];
304 int __attribute__((weak
)) arch_dup_task_struct(struct task_struct
*dst
,
305 struct task_struct
*src
)
311 static struct task_struct
*dup_task_struct(struct task_struct
*orig
)
313 struct task_struct
*tsk
;
314 struct thread_info
*ti
;
315 unsigned long *stackend
;
316 int node
= tsk_fork_get_node(orig
);
319 tsk
= alloc_task_struct_node(node
);
323 ti
= alloc_thread_info_node(tsk
, node
);
327 err
= arch_dup_task_struct(tsk
, orig
);
331 tsk
->flags
&= ~PF_SU
;
334 #ifdef CONFIG_SECCOMP
336 * We must handle setting up seccomp filters once we're under
337 * the sighand lock in case orig has changed between now and
338 * then. Until then, filter must be NULL to avoid messing up
339 * the usage counts on the error path calling free_task.
341 tsk
->seccomp
.filter
= NULL
;
344 setup_thread_stack(tsk
, orig
);
345 clear_user_return_notifier(tsk
);
346 clear_tsk_need_resched(tsk
);
347 stackend
= end_of_stack(tsk
);
348 *stackend
= STACK_END_MAGIC
; /* for overflow detection */
350 #ifdef CONFIG_CC_STACKPROTECTOR
351 tsk
->stack_canary
= get_random_int();
355 * One for us, one for whoever does the "release_task()" (usually
358 atomic_set(&tsk
->usage
, 2);
359 #ifdef CONFIG_BLK_DEV_IO_TRACE
362 tsk
->splice_pipe
= NULL
;
363 tsk
->task_frag
.page
= NULL
;
365 account_kernel_stack(ti
, 1);
370 free_thread_info(ti
);
372 free_task_struct(tsk
);
377 static int dup_mmap(struct mm_struct
*mm
, struct mm_struct
*oldmm
)
379 struct vm_area_struct
*mpnt
, *tmp
, *prev
, **pprev
;
380 struct rb_node
**rb_link
, *rb_parent
;
382 unsigned long charge
;
383 struct mempolicy
*pol
;
385 uprobe_start_dup_mmap();
386 down_write(&oldmm
->mmap_sem
);
387 flush_cache_dup_mm(oldmm
);
388 uprobe_dup_mmap(oldmm
, mm
);
390 * Not linked in yet - no deadlock potential:
392 down_write_nested(&mm
->mmap_sem
, SINGLE_DEPTH_NESTING
);
396 vmacache_invalidate(mm
);
397 mm
->free_area_cache
= oldmm
->mmap_base
;
398 mm
->cached_hole_size
= ~0UL;
400 cpumask_clear(mm_cpumask(mm
));
402 rb_link
= &mm
->mm_rb
.rb_node
;
405 retval
= ksm_fork(mm
, oldmm
);
408 retval
= khugepaged_fork(mm
, oldmm
);
413 for (mpnt
= oldmm
->mmap
; mpnt
; mpnt
= mpnt
->vm_next
) {
416 if (mpnt
->vm_flags
& VM_DONTCOPY
) {
417 vm_stat_account(mm
, mpnt
->vm_flags
, mpnt
->vm_file
,
422 if (mpnt
->vm_flags
& VM_ACCOUNT
) {
423 unsigned long len
= vma_pages(mpnt
);
425 if (security_vm_enough_memory_mm(oldmm
, len
)) /* sic */
429 tmp
= kmem_cache_alloc(vm_area_cachep
, GFP_KERNEL
);
433 INIT_LIST_HEAD(&tmp
->anon_vma_chain
);
434 pol
= mpol_dup(vma_policy(mpnt
));
435 retval
= PTR_ERR(pol
);
437 goto fail_nomem_policy
;
438 vma_set_policy(tmp
, pol
);
440 if (anon_vma_fork(tmp
, mpnt
))
441 goto fail_nomem_anon_vma_fork
;
442 tmp
->vm_flags
&= ~VM_LOCKED
;
443 tmp
->vm_next
= tmp
->vm_prev
= NULL
;
446 struct inode
*inode
= file_inode(file
);
447 struct address_space
*mapping
= file
->f_mapping
;
450 if (tmp
->vm_flags
& VM_DENYWRITE
)
451 atomic_dec(&inode
->i_writecount
);
452 mutex_lock(&mapping
->i_mmap_mutex
);
453 if (tmp
->vm_flags
& VM_SHARED
)
454 mapping
->i_mmap_writable
++;
455 flush_dcache_mmap_lock(mapping
);
456 /* insert tmp into the share list, just after mpnt */
457 if (unlikely(tmp
->vm_flags
& VM_NONLINEAR
))
458 vma_nonlinear_insert(tmp
,
459 &mapping
->i_mmap_nonlinear
);
461 vma_interval_tree_insert_after(tmp
, mpnt
,
463 flush_dcache_mmap_unlock(mapping
);
464 mutex_unlock(&mapping
->i_mmap_mutex
);
468 * Clear hugetlb-related page reserves for children. This only
469 * affects MAP_PRIVATE mappings. Faults generated by the child
470 * are not guaranteed to succeed, even if read-only
472 if (is_vm_hugetlb_page(tmp
))
473 reset_vma_resv_huge_pages(tmp
);
476 * Link in the new vma and copy the page table entries.
479 pprev
= &tmp
->vm_next
;
483 __vma_link_rb(mm
, tmp
, rb_link
, rb_parent
);
484 rb_link
= &tmp
->vm_rb
.rb_right
;
485 rb_parent
= &tmp
->vm_rb
;
488 retval
= copy_page_range(mm
, oldmm
, mpnt
);
490 if (tmp
->vm_ops
&& tmp
->vm_ops
->open
)
491 tmp
->vm_ops
->open(tmp
);
496 /* a new mm has just been created */
497 arch_dup_mmap(oldmm
, mm
);
500 up_write(&mm
->mmap_sem
);
502 up_write(&oldmm
->mmap_sem
);
503 uprobe_end_dup_mmap();
505 fail_nomem_anon_vma_fork
:
508 kmem_cache_free(vm_area_cachep
, tmp
);
511 vm_unacct_memory(charge
);
515 static inline int mm_alloc_pgd(struct mm_struct
*mm
)
517 mm
->pgd
= pgd_alloc(mm
);
518 if (unlikely(!mm
->pgd
))
523 static inline void mm_free_pgd(struct mm_struct
*mm
)
525 pgd_free(mm
, mm
->pgd
);
528 #define dup_mmap(mm, oldmm) (0)
529 #define mm_alloc_pgd(mm) (0)
530 #define mm_free_pgd(mm)
531 #endif /* CONFIG_MMU */
533 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(mmlist_lock
);
535 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
536 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
538 static unsigned long default_dump_filter
= MMF_DUMP_FILTER_DEFAULT
;
540 static int __init
coredump_filter_setup(char *s
)
542 default_dump_filter
=
543 (simple_strtoul(s
, NULL
, 0) << MMF_DUMP_FILTER_SHIFT
) &
544 MMF_DUMP_FILTER_MASK
;
548 __setup("coredump_filter=", coredump_filter_setup
);
550 #include <linux/init_task.h>
552 static void mm_init_aio(struct mm_struct
*mm
)
555 spin_lock_init(&mm
->ioctx_lock
);
556 INIT_HLIST_HEAD(&mm
->ioctx_list
);
560 static struct mm_struct
*mm_init(struct mm_struct
*mm
, struct task_struct
*p
)
562 atomic_set(&mm
->mm_users
, 1);
563 atomic_set(&mm
->mm_count
, 1);
564 init_rwsem(&mm
->mmap_sem
);
565 INIT_LIST_HEAD(&mm
->mmlist
);
566 mm
->flags
= (current
->mm
) ?
567 (current
->mm
->flags
& MMF_INIT_MASK
) : default_dump_filter
;
568 mm
->core_state
= NULL
;
570 memset(&mm
->rss_stat
, 0, sizeof(mm
->rss_stat
));
571 spin_lock_init(&mm
->page_table_lock
);
572 mm
->free_area_cache
= TASK_UNMAPPED_BASE
;
573 mm
->cached_hole_size
= ~0UL;
575 mm_init_owner(mm
, p
);
576 clear_tlb_flush_pending(mm
);
578 if (likely(!mm_alloc_pgd(mm
))) {
580 mmu_notifier_mm_init(mm
);
588 static void check_mm(struct mm_struct
*mm
)
592 for (i
= 0; i
< NR_MM_COUNTERS
; i
++) {
593 long x
= atomic_long_read(&mm
->rss_stat
.count
[i
]);
596 printk(KERN_ALERT
"BUG: Bad rss-counter state "
597 "mm:%p idx:%d val:%ld\n", mm
, i
, x
);
600 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
601 VM_BUG_ON(mm
->pmd_huge_pte
);
606 * Allocate and initialize an mm_struct.
608 struct mm_struct
*mm_alloc(void)
610 struct mm_struct
*mm
;
616 memset(mm
, 0, sizeof(*mm
));
618 return mm_init(mm
, current
);
622 * Called when the last reference to the mm
623 * is dropped: either by a lazy thread or by
624 * mmput. Free the page directory and the mm.
626 void __mmdrop(struct mm_struct
*mm
)
628 BUG_ON(mm
== &init_mm
);
631 mmu_notifier_mm_destroy(mm
);
635 EXPORT_SYMBOL_GPL(__mmdrop
);
638 * Decrement the use count and release all resources for an mm.
640 int mmput(struct mm_struct
*mm
)
645 if (atomic_dec_and_test(&mm
->mm_users
)) {
646 uprobe_clear_state(mm
);
649 khugepaged_exit(mm
); /* must run before exit_mmap */
651 set_mm_exe_file(mm
, NULL
);
652 if (!list_empty(&mm
->mmlist
)) {
653 spin_lock(&mmlist_lock
);
654 list_del(&mm
->mmlist
);
655 spin_unlock(&mmlist_lock
);
658 module_put(mm
->binfmt
->module
);
664 EXPORT_SYMBOL_GPL(mmput
);
666 void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
)
669 get_file(new_exe_file
);
672 mm
->exe_file
= new_exe_file
;
675 struct file
*get_mm_exe_file(struct mm_struct
*mm
)
677 struct file
*exe_file
;
679 /* We need mmap_sem to protect against races with removal of exe_file */
680 down_read(&mm
->mmap_sem
);
681 exe_file
= mm
->exe_file
;
684 up_read(&mm
->mmap_sem
);
688 static void dup_mm_exe_file(struct mm_struct
*oldmm
, struct mm_struct
*newmm
)
690 /* It's safe to write the exe_file pointer without exe_file_lock because
691 * this is called during fork when the task is not yet in /proc */
692 newmm
->exe_file
= get_mm_exe_file(oldmm
);
696 * get_task_mm - acquire a reference to the task's mm
698 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
699 * this kernel workthread has transiently adopted a user mm with use_mm,
700 * to do its AIO) is not set and if so returns a reference to it, after
701 * bumping up the use count. User must release the mm via mmput()
702 * after use. Typically used by /proc and ptrace.
704 struct mm_struct
*get_task_mm(struct task_struct
*task
)
706 struct mm_struct
*mm
;
711 if (task
->flags
& PF_KTHREAD
)
714 atomic_inc(&mm
->mm_users
);
719 EXPORT_SYMBOL_GPL(get_task_mm
);
721 struct mm_struct
*mm_access(struct task_struct
*task
, unsigned int mode
)
723 struct mm_struct
*mm
;
726 err
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
730 mm
= get_task_mm(task
);
731 if (mm
&& mm
!= current
->mm
&&
732 !ptrace_may_access(task
, mode
) &&
733 !capable(CAP_SYS_RESOURCE
)) {
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 * Signal userspace if we're not exiting with a core dump
812 * because we want to leave the value intact for debugging
815 if (tsk
->clear_child_tid
) {
816 if (!(tsk
->signal
->flags
& SIGNAL_GROUP_COREDUMP
) &&
817 atomic_read(&mm
->mm_users
) > 1) {
819 * We don't check the error code - if userspace has
820 * not set up a proper pointer then tough luck.
822 put_user(0, tsk
->clear_child_tid
);
823 sys_futex(tsk
->clear_child_tid
, FUTEX_WAKE
,
826 tsk
->clear_child_tid
= NULL
;
830 * All done, finally we can wake up parent and return this mm to him.
831 * Also kthread_stop() uses this completion for synchronization.
834 complete_vfork_done(tsk
);
838 * Allocate a new mm structure and copy contents from the
839 * mm structure of the passed in task structure.
841 struct mm_struct
*dup_mm(struct task_struct
*tsk
)
843 struct mm_struct
*mm
, *oldmm
= current
->mm
;
853 memcpy(mm
, oldmm
, sizeof(*mm
));
856 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
857 mm
->pmd_huge_pte
= NULL
;
859 #ifdef CONFIG_NUMA_BALANCING
860 mm
->first_nid
= NUMA_PTE_SCAN_INIT
;
862 if (!mm_init(mm
, tsk
))
865 if (init_new_context(tsk
, mm
))
868 dup_mm_exe_file(oldmm
, mm
);
870 err
= dup_mmap(mm
, oldmm
);
874 mm
->hiwater_rss
= get_mm_rss(mm
);
875 mm
->hiwater_vm
= mm
->total_vm
;
877 if (mm
->binfmt
&& !try_module_get(mm
->binfmt
->module
))
883 /* don't put binfmt in mmput, we haven't got module yet */
892 * If init_new_context() failed, we cannot use mmput() to free the mm
893 * because it calls destroy_context()
900 static int copy_mm(unsigned long clone_flags
, struct task_struct
*tsk
)
902 struct mm_struct
*mm
, *oldmm
;
905 tsk
->min_flt
= tsk
->maj_flt
= 0;
906 tsk
->nvcsw
= tsk
->nivcsw
= 0;
907 #ifdef CONFIG_DETECT_HUNG_TASK
908 tsk
->last_switch_count
= tsk
->nvcsw
+ tsk
->nivcsw
;
912 tsk
->active_mm
= NULL
;
915 * Are we cloning a kernel thread?
917 * We need to steal a active VM for that..
923 if (clone_flags
& CLONE_VM
) {
924 atomic_inc(&oldmm
->mm_users
);
943 static int copy_fs(unsigned long clone_flags
, struct task_struct
*tsk
)
945 struct fs_struct
*fs
= current
->fs
;
946 if (clone_flags
& CLONE_FS
) {
947 /* tsk->fs is already what we want */
948 spin_lock(&fs
->lock
);
950 spin_unlock(&fs
->lock
);
954 spin_unlock(&fs
->lock
);
957 tsk
->fs
= copy_fs_struct(fs
);
963 static int copy_files(unsigned long clone_flags
, struct task_struct
*tsk
)
965 struct files_struct
*oldf
, *newf
;
969 * A background process may not have any files ...
971 oldf
= current
->files
;
975 if (clone_flags
& CLONE_FILES
) {
976 atomic_inc(&oldf
->count
);
980 newf
= dup_fd(oldf
, &error
);
990 static int copy_io(unsigned long clone_flags
, struct task_struct
*tsk
)
993 struct io_context
*ioc
= current
->io_context
;
994 struct io_context
*new_ioc
;
999 * Share io context with parent, if CLONE_IO is set
1001 if (clone_flags
& CLONE_IO
) {
1003 tsk
->io_context
= ioc
;
1004 } else if (ioprio_valid(ioc
->ioprio
)) {
1005 new_ioc
= get_task_io_context(tsk
, GFP_KERNEL
, NUMA_NO_NODE
);
1006 if (unlikely(!new_ioc
))
1009 new_ioc
->ioprio
= ioc
->ioprio
;
1010 put_io_context(new_ioc
);
1016 static int copy_sighand(unsigned long clone_flags
, struct task_struct
*tsk
)
1018 struct sighand_struct
*sig
;
1020 if (clone_flags
& CLONE_SIGHAND
) {
1021 atomic_inc(¤t
->sighand
->count
);
1024 sig
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
1025 rcu_assign_pointer(tsk
->sighand
, sig
);
1028 atomic_set(&sig
->count
, 1);
1029 memcpy(sig
->action
, current
->sighand
->action
, sizeof(sig
->action
));
1033 void __cleanup_sighand(struct sighand_struct
*sighand
)
1035 if (atomic_dec_and_test(&sighand
->count
)) {
1036 signalfd_cleanup(sighand
);
1037 kmem_cache_free(sighand_cachep
, sighand
);
1043 * Initialize POSIX timer handling for a thread group.
1045 static void posix_cpu_timers_init_group(struct signal_struct
*sig
)
1047 unsigned long cpu_limit
;
1049 /* Thread group counters. */
1050 thread_group_cputime_init(sig
);
1052 cpu_limit
= ACCESS_ONCE(sig
->rlim
[RLIMIT_CPU
].rlim_cur
);
1053 if (cpu_limit
!= RLIM_INFINITY
) {
1054 sig
->cputime_expires
.prof_exp
= secs_to_cputime(cpu_limit
);
1055 sig
->cputimer
.running
= 1;
1058 /* The timer lists. */
1059 INIT_LIST_HEAD(&sig
->cpu_timers
[0]);
1060 INIT_LIST_HEAD(&sig
->cpu_timers
[1]);
1061 INIT_LIST_HEAD(&sig
->cpu_timers
[2]);
1064 static int copy_signal(unsigned long clone_flags
, struct task_struct
*tsk
)
1066 struct signal_struct
*sig
;
1068 if (clone_flags
& CLONE_THREAD
)
1071 sig
= kmem_cache_zalloc(signal_cachep
, GFP_KERNEL
);
1076 sig
->nr_threads
= 1;
1077 atomic_set(&sig
->live
, 1);
1078 atomic_set(&sig
->sigcnt
, 1);
1080 /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */
1081 sig
->thread_head
= (struct list_head
)LIST_HEAD_INIT(tsk
->thread_node
);
1082 tsk
->thread_node
= (struct list_head
)LIST_HEAD_INIT(sig
->thread_head
);
1084 init_waitqueue_head(&sig
->wait_chldexit
);
1085 sig
->curr_target
= tsk
;
1086 init_sigpending(&sig
->shared_pending
);
1087 INIT_LIST_HEAD(&sig
->posix_timers
);
1089 hrtimer_init(&sig
->real_timer
, CLOCK_MONOTONIC
, HRTIMER_MODE_REL
);
1090 sig
->real_timer
.function
= it_real_fn
;
1092 task_lock(current
->group_leader
);
1093 memcpy(sig
->rlim
, current
->signal
->rlim
, sizeof sig
->rlim
);
1094 task_unlock(current
->group_leader
);
1096 posix_cpu_timers_init_group(sig
);
1098 tty_audit_fork(sig
);
1099 sched_autogroup_fork(sig
);
1101 #ifdef CONFIG_CGROUPS
1102 init_rwsem(&sig
->group_rwsem
);
1105 sig
->oom_score_adj
= current
->signal
->oom_score_adj
;
1106 sig
->oom_score_adj_min
= current
->signal
->oom_score_adj_min
;
1108 sig
->has_child_subreaper
= current
->signal
->has_child_subreaper
||
1109 current
->signal
->is_child_subreaper
;
1111 mutex_init(&sig
->cred_guard_mutex
);
1116 static void copy_flags(unsigned long clone_flags
, struct task_struct
*p
)
1118 unsigned long new_flags
= p
->flags
;
1120 new_flags
&= ~(PF_SUPERPRIV
| PF_WQ_WORKER
);
1121 new_flags
|= PF_FORKNOEXEC
;
1122 p
->flags
= new_flags
;
1125 static void copy_seccomp(struct task_struct
*p
)
1127 #ifdef CONFIG_SECCOMP
1129 * Must be called with sighand->lock held, which is common to
1130 * all threads in the group. Holding cred_guard_mutex is not
1131 * needed because this new task is not yet running and cannot
1134 assert_spin_locked(¤t
->sighand
->siglock
);
1136 /* Ref-count the new filter user, and assign it. */
1137 get_seccomp_filter(current
);
1138 p
->seccomp
= current
->seccomp
;
1141 * Explicitly enable no_new_privs here in case it got set
1142 * between the task_struct being duplicated and holding the
1143 * sighand lock. The seccomp state and nnp must be in sync.
1145 if (task_no_new_privs(current
))
1146 task_set_no_new_privs(p
);
1149 * If the parent gained a seccomp mode after copying thread
1150 * flags and between before we held the sighand lock, we have
1151 * to manually enable the seccomp thread flag here.
1153 if (p
->seccomp
.mode
!= SECCOMP_MODE_DISABLED
)
1154 set_tsk_thread_flag(p
, TIF_SECCOMP
);
1158 SYSCALL_DEFINE1(set_tid_address
, int __user
*, tidptr
)
1160 current
->clear_child_tid
= tidptr
;
1162 return task_pid_vnr(current
);
1165 static void rt_mutex_init_task(struct task_struct
*p
)
1167 raw_spin_lock_init(&p
->pi_lock
);
1168 #ifdef CONFIG_RT_MUTEXES
1169 plist_head_init(&p
->pi_waiters
);
1170 p
->pi_blocked_on
= NULL
;
1174 #ifdef CONFIG_MM_OWNER
1175 void mm_init_owner(struct mm_struct
*mm
, struct task_struct
*p
)
1179 #endif /* CONFIG_MM_OWNER */
1182 * Initialize POSIX timer handling for a single task.
1184 static void posix_cpu_timers_init(struct task_struct
*tsk
)
1186 tsk
->cputime_expires
.prof_exp
= 0;
1187 tsk
->cputime_expires
.virt_exp
= 0;
1188 tsk
->cputime_expires
.sched_exp
= 0;
1189 INIT_LIST_HEAD(&tsk
->cpu_timers
[0]);
1190 INIT_LIST_HEAD(&tsk
->cpu_timers
[1]);
1191 INIT_LIST_HEAD(&tsk
->cpu_timers
[2]);
1194 #ifdef CONFIG_RKP_KDP
1195 void rkp_assign_pgd(struct task_struct
*p
)
1198 pgd
= (u64
)(p
->mm
? p
->mm
->pgd
:swapper_pg_dir
);
1199 rkp_call(RKP_CMDID(0x43),(u64
)p
->cred
, (u64
)pgd
,0,0,0);
1201 #endif /*CONFIG_RKP_KDP*/
1203 * This creates a new process as a copy of the old one,
1204 * but does not actually start it yet.
1206 * It copies the registers, and all the appropriate
1207 * parts of the process environment (as per the clone
1208 * flags). The actual kick-off is left to the caller.
1210 static struct task_struct
*copy_process(unsigned long clone_flags
,
1211 unsigned long stack_start
,
1212 unsigned long stack_size
,
1213 int __user
*child_tidptr
,
1218 struct task_struct
*p
;
1220 if ((clone_flags
& (CLONE_NEWNS
|CLONE_FS
)) == (CLONE_NEWNS
|CLONE_FS
))
1221 return ERR_PTR(-EINVAL
);
1223 if ((clone_flags
& (CLONE_NEWUSER
|CLONE_FS
)) == (CLONE_NEWUSER
|CLONE_FS
))
1224 return ERR_PTR(-EINVAL
);
1227 * Thread groups must share signals as well, and detached threads
1228 * can only be started up within the thread group.
1230 if ((clone_flags
& CLONE_THREAD
) && !(clone_flags
& CLONE_SIGHAND
))
1231 return ERR_PTR(-EINVAL
);
1234 * Shared signal handlers imply shared VM. By way of the above,
1235 * thread groups also imply shared VM. Blocking this case allows
1236 * for various simplifications in other code.
1238 if ((clone_flags
& CLONE_SIGHAND
) && !(clone_flags
& CLONE_VM
))
1239 return ERR_PTR(-EINVAL
);
1242 * Siblings of global init remain as zombies on exit since they are
1243 * not reaped by their parent (swapper). To solve this and to avoid
1244 * multi-rooted process trees, prevent global and container-inits
1245 * from creating siblings.
1247 if ((clone_flags
& CLONE_PARENT
) &&
1248 current
->signal
->flags
& SIGNAL_UNKILLABLE
)
1249 return ERR_PTR(-EINVAL
);
1252 * If the new process will be in a different pid namespace don't
1253 * allow it to share a thread group or signal handlers with the
1256 if ((clone_flags
& (CLONE_SIGHAND
| CLONE_NEWPID
)) &&
1257 (task_active_pid_ns(current
) != current
->nsproxy
->pid_ns
))
1258 return ERR_PTR(-EINVAL
);
1260 retval
= security_task_create(clone_flags
);
1265 p
= dup_task_struct(current
);
1269 ftrace_graph_init_task(p
);
1271 rt_mutex_init_task(p
);
1273 #ifdef CONFIG_PROVE_LOCKING
1274 DEBUG_LOCKS_WARN_ON(!p
->hardirqs_enabled
);
1275 DEBUG_LOCKS_WARN_ON(!p
->softirqs_enabled
);
1278 if (atomic_read(&p
->real_cred
->user
->processes
) >=
1279 task_rlimit(p
, RLIMIT_NPROC
)) {
1280 if (!capable(CAP_SYS_ADMIN
) && !capable(CAP_SYS_RESOURCE
) &&
1281 p
->real_cred
->user
!= INIT_USER
)
1284 current
->flags
&= ~PF_NPROC_EXCEEDED
;
1286 retval
= copy_creds(p
, clone_flags
);
1291 * If multiple threads are within copy_process(), then this check
1292 * triggers too late. This doesn't hurt, the check is only there
1293 * to stop root fork bombs.
1296 if (nr_threads
>= max_threads
)
1297 goto bad_fork_cleanup_count
;
1299 if (!try_module_get(task_thread_info(p
)->exec_domain
->module
))
1300 goto bad_fork_cleanup_count
;
1303 delayacct_tsk_init(p
); /* Must remain after dup_task_struct() */
1304 copy_flags(clone_flags
, p
);
1305 INIT_LIST_HEAD(&p
->children
);
1306 INIT_LIST_HEAD(&p
->sibling
);
1307 rcu_copy_process(p
);
1308 p
->vfork_done
= NULL
;
1309 spin_lock_init(&p
->alloc_lock
);
1311 init_sigpending(&p
->pending
);
1313 p
->utime
= p
->stime
= p
->gtime
= 0;
1314 p
->utimescaled
= p
->stimescaled
= 0;
1316 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
1317 p
->prev_cputime
.utime
= p
->prev_cputime
.stime
= 0;
1319 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1320 seqlock_init(&p
->vtime_seqlock
);
1322 p
->vtime_snap_whence
= VTIME_SLEEPING
;
1325 #if defined(SPLIT_RSS_COUNTING)
1326 memset(&p
->rss_stat
, 0, sizeof(p
->rss_stat
));
1329 p
->default_timer_slack_ns
= current
->timer_slack_ns
;
1331 task_io_accounting_init(&p
->ioac
);
1332 acct_clear_integrals(p
);
1334 posix_cpu_timers_init(p
);
1336 do_posix_clock_monotonic_gettime(&p
->start_time
);
1337 p
->real_start_time
= p
->start_time
;
1338 monotonic_to_bootbased(&p
->real_start_time
);
1339 p
->io_context
= NULL
;
1340 p
->audit_context
= NULL
;
1341 if (clone_flags
& CLONE_THREAD
)
1342 threadgroup_change_begin(current
);
1345 p
->mempolicy
= mpol_dup(p
->mempolicy
);
1346 if (IS_ERR(p
->mempolicy
)) {
1347 retval
= PTR_ERR(p
->mempolicy
);
1348 p
->mempolicy
= NULL
;
1349 goto bad_fork_cleanup_cgroup
;
1351 mpol_fix_fork_child_flag(p
);
1353 #ifdef CONFIG_CPUSETS
1354 p
->cpuset_mem_spread_rotor
= NUMA_NO_NODE
;
1355 p
->cpuset_slab_spread_rotor
= NUMA_NO_NODE
;
1356 seqcount_init(&p
->mems_allowed_seq
);
1358 #ifdef CONFIG_TRACE_IRQFLAGS
1360 p
->hardirqs_enabled
= 0;
1361 p
->hardirq_enable_ip
= 0;
1362 p
->hardirq_enable_event
= 0;
1363 p
->hardirq_disable_ip
= _THIS_IP_
;
1364 p
->hardirq_disable_event
= 0;
1365 p
->softirqs_enabled
= 1;
1366 p
->softirq_enable_ip
= _THIS_IP_
;
1367 p
->softirq_enable_event
= 0;
1368 p
->softirq_disable_ip
= 0;
1369 p
->softirq_disable_event
= 0;
1370 p
->hardirq_context
= 0;
1371 p
->softirq_context
= 0;
1373 #ifdef CONFIG_LOCKDEP
1374 p
->lockdep_depth
= 0; /* no locks held yet */
1375 p
->curr_chain_key
= 0;
1376 p
->lockdep_recursion
= 0;
1379 #ifdef CONFIG_DEBUG_MUTEXES
1380 p
->blocked_on
= NULL
; /* not blocked yet */
1383 p
->memcg_batch
.do_batch
= 0;
1384 p
->memcg_batch
.memcg
= NULL
;
1386 #ifdef CONFIG_BCACHE
1387 p
->sequential_io
= 0;
1388 p
->sequential_io_avg
= 0;
1391 /* Perform scheduler related setup. Assign this task to a CPU. */
1394 retval
= perf_event_init_task(p
);
1396 goto bad_fork_cleanup_policy
;
1397 retval
= audit_alloc(p
);
1399 goto bad_fork_cleanup_perf
;
1400 /* copy all the process information */
1401 retval
= copy_semundo(clone_flags
, p
);
1403 goto bad_fork_cleanup_audit
;
1404 retval
= copy_files(clone_flags
, p
);
1406 goto bad_fork_cleanup_semundo
;
1407 retval
= copy_fs(clone_flags
, p
);
1409 goto bad_fork_cleanup_files
;
1410 retval
= copy_sighand(clone_flags
, p
);
1412 goto bad_fork_cleanup_fs
;
1413 retval
= copy_signal(clone_flags
, p
);
1415 goto bad_fork_cleanup_sighand
;
1416 retval
= copy_mm(clone_flags
, p
);
1418 goto bad_fork_cleanup_signal
;
1419 retval
= copy_namespaces(clone_flags
, p
);
1421 goto bad_fork_cleanup_mm
;
1422 retval
= copy_io(clone_flags
, p
);
1424 goto bad_fork_cleanup_namespaces
;
1425 retval
= copy_thread(clone_flags
, stack_start
, stack_size
, p
);
1427 goto bad_fork_cleanup_io
;
1429 if (pid
!= &init_struct_pid
) {
1431 pid
= alloc_pid(p
->nsproxy
->pid_ns
);
1433 goto bad_fork_cleanup_io
;
1436 p
->pid
= pid_nr(pid
);
1438 if (clone_flags
& CLONE_THREAD
)
1439 p
->tgid
= current
->tgid
;
1441 p
->set_child_tid
= (clone_flags
& CLONE_CHILD_SETTID
) ? child_tidptr
: NULL
;
1443 * Clear TID on mm_release()?
1445 p
->clear_child_tid
= (clone_flags
& CLONE_CHILD_CLEARTID
) ? child_tidptr
: NULL
;
1450 p
->robust_list
= NULL
;
1451 #ifdef CONFIG_COMPAT
1452 p
->compat_robust_list
= NULL
;
1454 INIT_LIST_HEAD(&p
->pi_state_list
);
1455 p
->pi_state_cache
= NULL
;
1457 uprobe_copy_process(p
);
1459 * sigaltstack should be cleared when sharing the same VM
1461 if ((clone_flags
& (CLONE_VM
|CLONE_VFORK
)) == CLONE_VM
)
1462 p
->sas_ss_sp
= p
->sas_ss_size
= 0;
1465 * Syscall tracing and stepping should be turned off in the
1466 * child regardless of CLONE_PTRACE.
1468 user_disable_single_step(p
);
1469 clear_tsk_thread_flag(p
, TIF_SYSCALL_TRACE
);
1470 #ifdef TIF_SYSCALL_EMU
1471 clear_tsk_thread_flag(p
, TIF_SYSCALL_EMU
);
1473 clear_all_latency_tracing(p
);
1475 /* ok, now we should be set up.. */
1476 if (clone_flags
& CLONE_THREAD
)
1477 p
->exit_signal
= -1;
1478 else if (clone_flags
& CLONE_PARENT
)
1479 p
->exit_signal
= current
->group_leader
->exit_signal
;
1481 p
->exit_signal
= (clone_flags
& CSIGNAL
);
1483 p
->pdeath_signal
= 0;
1487 p
->nr_dirtied_pause
= 128 >> (PAGE_SHIFT
- 10);
1488 p
->dirty_paused_when
= 0;
1491 * Ok, make it visible to the rest of the system.
1492 * We dont wake it up yet.
1494 p
->group_leader
= p
;
1495 INIT_LIST_HEAD(&p
->thread_group
);
1496 p
->task_works
= NULL
;
1498 /* Need tasklist lock for parent etc handling! */
1499 write_lock_irq(&tasklist_lock
);
1501 /* CLONE_PARENT re-uses the old parent */
1502 if (clone_flags
& (CLONE_PARENT
|CLONE_THREAD
)) {
1503 p
->real_parent
= current
->real_parent
;
1504 p
->parent_exec_id
= current
->parent_exec_id
;
1506 p
->real_parent
= current
;
1507 p
->parent_exec_id
= current
->self_exec_id
;
1510 spin_lock(¤t
->sighand
->siglock
);
1513 * Copy seccomp details explicitly here, in case they were changed
1514 * before holding sighand lock.
1519 * Process group and session signals need to be delivered to just the
1520 * parent before the fork or both the parent and the child after the
1521 * fork. Restart if a signal comes in before we add the new process to
1522 * it's process group.
1523 * A fatal signal pending means that current will exit, so the new
1524 * thread can't slip out of an OOM kill (or normal SIGKILL).
1526 recalc_sigpending();
1527 if (signal_pending(current
)) {
1528 spin_unlock(¤t
->sighand
->siglock
);
1529 write_unlock_irq(&tasklist_lock
);
1530 retval
= -ERESTARTNOINTR
;
1531 goto bad_fork_free_pid
;
1534 if (likely(p
->pid
)) {
1535 ptrace_init_task(p
, (clone_flags
& CLONE_PTRACE
) || trace
);
1537 if (thread_group_leader(p
)) {
1538 if (is_child_reaper(pid
)) {
1539 ns_of_pid(pid
)->child_reaper
= p
;
1540 p
->signal
->flags
|= SIGNAL_UNKILLABLE
;
1543 p
->signal
->leader_pid
= pid
;
1544 p
->signal
->tty
= tty_kref_get(current
->signal
->tty
);
1545 attach_pid(p
, PIDTYPE_PGID
, task_pgrp(current
));
1546 attach_pid(p
, PIDTYPE_SID
, task_session(current
));
1547 list_add_tail(&p
->sibling
, &p
->real_parent
->children
);
1548 list_add_tail_rcu(&p
->tasks
, &init_task
.tasks
);
1549 __this_cpu_inc(process_counts
);
1551 current
->signal
->nr_threads
++;
1552 atomic_inc(¤t
->signal
->live
);
1553 atomic_inc(¤t
->signal
->sigcnt
);
1554 p
->group_leader
= current
->group_leader
;
1555 list_add_tail_rcu(&p
->thread_group
,
1556 &p
->group_leader
->thread_group
);
1557 list_add_tail_rcu(&p
->thread_node
,
1558 &p
->signal
->thread_head
);
1560 attach_pid(p
, PIDTYPE_PID
, pid
);
1565 spin_unlock(¤t
->sighand
->siglock
);
1566 syscall_tracepoint_update(p
);
1567 write_unlock_irq(&tasklist_lock
);
1569 proc_fork_connector(p
);
1570 cgroup_post_fork(p
);
1571 if (clone_flags
& CLONE_THREAD
)
1572 threadgroup_change_end(current
);
1575 trace_task_newtask(p
, clone_flags
);
1576 #ifdef CONFIG_RKP_KDP
1579 #endif/*CONFIG_RKP_KDP*/
1584 if (pid
!= &init_struct_pid
)
1586 bad_fork_cleanup_io
:
1589 bad_fork_cleanup_namespaces
:
1590 exit_task_namespaces(p
);
1591 bad_fork_cleanup_mm
:
1594 bad_fork_cleanup_signal
:
1595 if (!(clone_flags
& CLONE_THREAD
))
1596 free_signal_struct(p
->signal
);
1597 bad_fork_cleanup_sighand
:
1598 __cleanup_sighand(p
->sighand
);
1599 bad_fork_cleanup_fs
:
1600 exit_fs(p
); /* blocking */
1601 bad_fork_cleanup_files
:
1602 exit_files(p
); /* blocking */
1603 bad_fork_cleanup_semundo
:
1605 bad_fork_cleanup_audit
:
1607 bad_fork_cleanup_perf
:
1608 perf_event_free_task(p
);
1609 bad_fork_cleanup_policy
:
1611 mpol_put(p
->mempolicy
);
1612 bad_fork_cleanup_cgroup
:
1614 if (clone_flags
& CLONE_THREAD
)
1615 threadgroup_change_end(current
);
1617 delayacct_tsk_free(p
);
1618 module_put(task_thread_info(p
)->exec_domain
->module
);
1619 bad_fork_cleanup_count
:
1620 atomic_dec(&p
->cred
->user
->processes
);
1625 return ERR_PTR(retval
);
1628 static inline void init_idle_pids(struct pid_link
*links
)
1632 for (type
= PIDTYPE_PID
; type
< PIDTYPE_MAX
; ++type
) {
1633 INIT_HLIST_NODE(&links
[type
].node
); /* not really needed */
1634 links
[type
].pid
= &init_struct_pid
;
1638 struct task_struct
* __cpuinit
fork_idle(int cpu
)
1640 struct task_struct
*task
;
1641 task
= copy_process(CLONE_VM
, 0, 0, NULL
, &init_struct_pid
, 0);
1642 if (!IS_ERR(task
)) {
1643 init_idle_pids(task
->pids
);
1644 init_idle(task
, cpu
);
1651 * Ok, this is the main fork-routine.
1653 * It copies the process, and if successful kick-starts
1654 * it and waits for it to finish using the VM if required.
1656 long do_fork(unsigned long clone_flags
,
1657 unsigned long stack_start
,
1658 unsigned long stack_size
,
1659 int __user
*parent_tidptr
,
1660 int __user
*child_tidptr
)
1662 struct task_struct
*p
;
1667 * Do some preliminary argument and permissions checking before we
1668 * actually start allocating stuff
1670 if (clone_flags
& (CLONE_NEWUSER
| CLONE_NEWPID
)) {
1671 if (clone_flags
& (CLONE_THREAD
|CLONE_PARENT
))
1676 * Determine whether and which event to report to ptracer. When
1677 * called from kernel_thread or CLONE_UNTRACED is explicitly
1678 * requested, no event is reported; otherwise, report if the event
1679 * for the type of forking is enabled.
1681 if (!(clone_flags
& CLONE_UNTRACED
)) {
1682 if (clone_flags
& CLONE_VFORK
)
1683 trace
= PTRACE_EVENT_VFORK
;
1684 else if ((clone_flags
& CSIGNAL
) != SIGCHLD
)
1685 trace
= PTRACE_EVENT_CLONE
;
1687 trace
= PTRACE_EVENT_FORK
;
1689 if (likely(!ptrace_event_enabled(current
, trace
)))
1693 p
= copy_process(clone_flags
, stack_start
, stack_size
,
1694 child_tidptr
, NULL
, trace
);
1696 * Do this prior waking up the new thread - the thread pointer
1697 * might get invalid after that point, if the thread exits quickly.
1700 struct completion vfork
;
1703 trace_sched_process_fork(current
, p
);
1705 pid
= get_task_pid(p
, PIDTYPE_PID
);
1708 if (clone_flags
& CLONE_PARENT_SETTID
)
1709 put_user(nr
, parent_tidptr
);
1711 if (clone_flags
& CLONE_VFORK
) {
1712 p
->vfork_done
= &vfork
;
1713 init_completion(&vfork
);
1717 wake_up_new_task(p
);
1719 /* forking complete and child started to run, tell ptracer */
1720 if (unlikely(trace
))
1721 ptrace_event_pid(trace
, pid
);
1723 if (clone_flags
& CLONE_VFORK
) {
1724 if (!wait_for_vfork_done(p
, &vfork
))
1725 ptrace_event_pid(PTRACE_EVENT_VFORK_DONE
, pid
);
1736 * Create a kernel thread.
1738 pid_t
kernel_thread(int (*fn
)(void *), void *arg
, unsigned long flags
)
1740 return do_fork(flags
|CLONE_VM
|CLONE_UNTRACED
, (unsigned long)fn
,
1741 (unsigned long)arg
, NULL
, NULL
);
1744 #ifdef __ARCH_WANT_SYS_FORK
1745 SYSCALL_DEFINE0(fork
)
1748 return do_fork(SIGCHLD
, 0, 0, NULL
, NULL
);
1750 /* can not support in nommu mode */
1756 #ifdef __ARCH_WANT_SYS_VFORK
1757 SYSCALL_DEFINE0(vfork
)
1759 return do_fork(CLONE_VFORK
| CLONE_VM
| SIGCHLD
, 0,
1764 #ifdef __ARCH_WANT_SYS_CLONE
1765 #ifdef CONFIG_CLONE_BACKWARDS
1766 SYSCALL_DEFINE5(clone
, unsigned long, clone_flags
, unsigned long, newsp
,
1767 int __user
*, parent_tidptr
,
1769 int __user
*, child_tidptr
)
1770 #elif defined(CONFIG_CLONE_BACKWARDS2)
1771 SYSCALL_DEFINE5(clone
, unsigned long, newsp
, unsigned long, clone_flags
,
1772 int __user
*, parent_tidptr
,
1773 int __user
*, child_tidptr
,
1775 #elif defined(CONFIG_CLONE_BACKWARDS3)
1776 SYSCALL_DEFINE6(clone
, unsigned long, clone_flags
, unsigned long, newsp
,
1778 int __user
*, parent_tidptr
,
1779 int __user
*, child_tidptr
,
1782 SYSCALL_DEFINE5(clone
, unsigned long, clone_flags
, unsigned long, newsp
,
1783 int __user
*, parent_tidptr
,
1784 int __user
*, child_tidptr
,
1788 return do_fork(clone_flags
, newsp
, 0, parent_tidptr
, child_tidptr
);
1792 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1793 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1796 static void sighand_ctor(void *data
)
1798 struct sighand_struct
*sighand
= data
;
1800 spin_lock_init(&sighand
->siglock
);
1801 init_waitqueue_head(&sighand
->signalfd_wqh
);
1804 void __init
proc_caches_init(void)
1806 sighand_cachep
= kmem_cache_create("sighand_cache",
1807 sizeof(struct sighand_struct
), 0,
1808 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_DESTROY_BY_RCU
|
1809 SLAB_NOTRACK
, sighand_ctor
);
1810 signal_cachep
= kmem_cache_create("signal_cache",
1811 sizeof(struct signal_struct
), 0,
1812 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1813 files_cachep
= kmem_cache_create("files_cache",
1814 sizeof(struct files_struct
), 0,
1815 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1816 fs_cachep
= kmem_cache_create("fs_cache",
1817 sizeof(struct fs_struct
), 0,
1818 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1820 * FIXME! The "sizeof(struct mm_struct)" currently includes the
1821 * whole struct cpumask for the OFFSTACK case. We could change
1822 * this to *only* allocate as much of it as required by the
1823 * maximum number of CPU's we can ever have. The cpumask_allocation
1824 * is at the end of the structure, exactly for that reason.
1826 mm_cachep
= kmem_cache_create("mm_struct",
1827 sizeof(struct mm_struct
), ARCH_MIN_MMSTRUCT_ALIGN
,
1828 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1829 vm_area_cachep
= KMEM_CACHE(vm_area_struct
, SLAB_PANIC
);
1831 nsproxy_cache_init();
1835 * Check constraints on flags passed to the unshare system call.
1837 static int check_unshare_flags(unsigned long unshare_flags
)
1839 if (unshare_flags
& ~(CLONE_THREAD
|CLONE_FS
|CLONE_NEWNS
|CLONE_SIGHAND
|
1840 CLONE_VM
|CLONE_FILES
|CLONE_SYSVSEM
|
1841 CLONE_NEWUTS
|CLONE_NEWIPC
|CLONE_NEWNET
|
1842 CLONE_NEWUSER
|CLONE_NEWPID
))
1845 * Not implemented, but pretend it works if there is nothing
1846 * to unshare. Note that unsharing the address space or the
1847 * signal handlers also need to unshare the signal queues (aka
1850 if (unshare_flags
& (CLONE_THREAD
| CLONE_SIGHAND
| CLONE_VM
)) {
1851 if (!thread_group_empty(current
))
1854 if (unshare_flags
& (CLONE_SIGHAND
| CLONE_VM
)) {
1855 if (atomic_read(¤t
->sighand
->count
) > 1)
1858 if (unshare_flags
& CLONE_VM
) {
1859 if (!current_is_single_threaded())
1867 * Unshare the filesystem structure if it is being shared
1869 static int unshare_fs(unsigned long unshare_flags
, struct fs_struct
**new_fsp
)
1871 struct fs_struct
*fs
= current
->fs
;
1873 if (!(unshare_flags
& CLONE_FS
) || !fs
)
1876 /* don't need lock here; in the worst case we'll do useless copy */
1880 *new_fsp
= copy_fs_struct(fs
);
1888 * Unshare file descriptor table if it is being shared
1890 static int unshare_fd(unsigned long unshare_flags
, struct files_struct
**new_fdp
)
1892 struct files_struct
*fd
= current
->files
;
1895 if ((unshare_flags
& CLONE_FILES
) &&
1896 (fd
&& atomic_read(&fd
->count
) > 1)) {
1897 *new_fdp
= dup_fd(fd
, &error
);
1906 * unshare allows a process to 'unshare' part of the process
1907 * context which was originally shared using clone. copy_*
1908 * functions used by do_fork() cannot be used here directly
1909 * because they modify an inactive task_struct that is being
1910 * constructed. Here we are modifying the current, active,
1913 SYSCALL_DEFINE1(unshare
, unsigned long, unshare_flags
)
1915 struct fs_struct
*fs
, *new_fs
= NULL
;
1916 struct files_struct
*fd
, *new_fd
= NULL
;
1917 struct cred
*new_cred
= NULL
;
1918 struct nsproxy
*new_nsproxy
= NULL
;
1923 * If unsharing a user namespace must also unshare the thread.
1925 if (unshare_flags
& CLONE_NEWUSER
)
1926 unshare_flags
|= CLONE_THREAD
| CLONE_FS
;
1928 * If unsharing a pid namespace must also unshare the thread.
1930 if (unshare_flags
& CLONE_NEWPID
)
1931 unshare_flags
|= CLONE_THREAD
;
1933 * If unsharing vm, must also unshare signal handlers.
1935 if (unshare_flags
& CLONE_VM
)
1936 unshare_flags
|= CLONE_SIGHAND
;
1938 * If unsharing a signal handlers, must also unshare the signal queues.
1940 if (unshare_flags
& CLONE_SIGHAND
)
1941 unshare_flags
|= CLONE_THREAD
;
1943 * If unsharing namespace, must also unshare filesystem information.
1945 if (unshare_flags
& CLONE_NEWNS
)
1946 unshare_flags
|= CLONE_FS
;
1948 err
= check_unshare_flags(unshare_flags
);
1950 goto bad_unshare_out
;
1952 * CLONE_NEWIPC must also detach from the undolist: after switching
1953 * to a new ipc namespace, the semaphore arrays from the old
1954 * namespace are unreachable.
1956 if (unshare_flags
& (CLONE_NEWIPC
|CLONE_SYSVSEM
))
1958 err
= unshare_fs(unshare_flags
, &new_fs
);
1960 goto bad_unshare_out
;
1961 err
= unshare_fd(unshare_flags
, &new_fd
);
1963 goto bad_unshare_cleanup_fs
;
1964 err
= unshare_userns(unshare_flags
, &new_cred
);
1966 goto bad_unshare_cleanup_fd
;
1967 err
= unshare_nsproxy_namespaces(unshare_flags
, &new_nsproxy
,
1970 goto bad_unshare_cleanup_cred
;
1972 if (new_fs
|| new_fd
|| do_sysvsem
|| new_cred
|| new_nsproxy
) {
1975 * CLONE_SYSVSEM is equivalent to sys_exit().
1981 switch_task_namespaces(current
, new_nsproxy
);
1987 spin_lock(&fs
->lock
);
1988 current
->fs
= new_fs
;
1993 spin_unlock(&fs
->lock
);
1997 fd
= current
->files
;
1998 current
->files
= new_fd
;
2002 task_unlock(current
);
2005 /* Install the new user namespace */
2006 commit_creds(new_cred
);
2011 bad_unshare_cleanup_cred
:
2014 bad_unshare_cleanup_fd
:
2016 put_files_struct(new_fd
);
2018 bad_unshare_cleanup_fs
:
2020 free_fs_struct(new_fs
);
2027 * Helper to unshare the files of the current task.
2028 * We don't want to expose copy_files internals to
2029 * the exec layer of the kernel.
2032 int unshare_files(struct files_struct
**displaced
)
2034 struct task_struct
*task
= current
;
2035 struct files_struct
*copy
= NULL
;
2038 error
= unshare_fd(CLONE_FILES
, ©
);
2039 if (error
|| !copy
) {
2043 *displaced
= task
->files
;