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/kaiser.h>
62 #include <linux/delayacct.h>
63 #include <linux/taskstats_kern.h>
64 #include <linux/random.h>
65 #include <linux/tty.h>
66 #include <linux/blkdev.h>
67 #include <linux/fs_struct.h>
68 #include <linux/magic.h>
69 #include <linux/perf_event.h>
70 #include <linux/posix-timers.h>
71 #include <linux/user-return-notifier.h>
72 #include <linux/oom.h>
73 #include <linux/khugepaged.h>
74 #include <linux/signalfd.h>
75 #include <linux/uprobes.h>
76 #include <linux/aio.h>
77 #include <linux/compiler.h>
78 #include <linux/sysctl.h>
79 #include <linux/kcov.h>
81 #include <asm/pgtable.h>
82 #include <asm/pgalloc.h>
83 #include <asm/uaccess.h>
84 #include <asm/mmu_context.h>
85 #include <asm/cacheflush.h>
86 #include <asm/tlbflush.h>
88 #include <trace/events/sched.h>
90 #define CREATE_TRACE_POINTS
91 #include <trace/events/task.h>
93 #ifdef CONFIG_SECURITY_DEFEX
94 #include <linux/defex.h>
98 * Minimum number of threads to boot the kernel
100 #define MIN_THREADS 20
103 * Maximum number of threads
105 #define MAX_THREADS FUTEX_TID_MASK
108 * Protected counters by write_lock_irq(&tasklist_lock)
110 unsigned long total_forks
; /* Handle normal Linux uptimes. */
111 int nr_threads
; /* The idle threads do not count.. */
113 int max_threads
; /* tunable limit on nr_threads */
115 DEFINE_PER_CPU(unsigned long, process_counts
) = 0;
117 __cacheline_aligned
DEFINE_RWLOCK(tasklist_lock
); /* outer */
119 #ifdef CONFIG_PROVE_RCU
120 int lockdep_tasklist_lock_is_held(void)
122 return lockdep_is_held(&tasklist_lock
);
124 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held
);
125 #endif /* #ifdef CONFIG_PROVE_RCU */
127 int nr_processes(void)
132 for_each_possible_cpu(cpu
)
133 total
+= per_cpu(process_counts
, cpu
);
138 void __weak
arch_release_task_struct(struct task_struct
*tsk
)
142 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
143 static struct kmem_cache
*task_struct_cachep
;
145 static inline struct task_struct
*alloc_task_struct_node(int node
)
147 return kmem_cache_alloc_node(task_struct_cachep
, GFP_KERNEL
, node
);
150 static inline void free_task_struct(struct task_struct
*tsk
)
152 kmem_cache_free(task_struct_cachep
, tsk
);
156 void __weak
arch_release_thread_stack(unsigned long *stack
)
160 #ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR
163 * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a
164 * kmemcache based allocator.
166 # if THREAD_SIZE >= PAGE_SIZE
167 static unsigned long *alloc_thread_stack_node(struct task_struct
*tsk
,
170 struct page
*page
= alloc_kmem_pages_node(node
, THREADINFO_GFP
,
173 return page
? page_address(page
) : NULL
;
176 static inline void free_thread_stack(unsigned long *stack
)
178 struct page
*page
= virt_to_page(stack
);
180 kaiser_unmap_thread_stack(stack
);
181 __free_kmem_pages(page
, THREAD_SIZE_ORDER
);
184 static struct kmem_cache
*thread_stack_cache
;
186 static unsigned long *alloc_thread_stack_node(struct task_struct
*tsk
,
189 return kmem_cache_alloc_node(thread_stack_cache
, THREADINFO_GFP
, node
);
192 static void free_thread_stack(unsigned long *stack
)
194 kmem_cache_free(thread_stack_cache
, stack
);
197 void thread_stack_cache_init(void)
199 thread_stack_cache
= kmem_cache_create("thread_stack", THREAD_SIZE
,
200 THREAD_SIZE
, 0, NULL
);
201 BUG_ON(thread_stack_cache
== NULL
);
206 /* SLAB cache for signal_struct structures (tsk->signal) */
207 static struct kmem_cache
*signal_cachep
;
209 /* SLAB cache for sighand_struct structures (tsk->sighand) */
210 struct kmem_cache
*sighand_cachep
;
212 /* SLAB cache for files_struct structures (tsk->files) */
213 struct kmem_cache
*files_cachep
;
215 /* SLAB cache for fs_struct structures (tsk->fs) */
216 struct kmem_cache
*fs_cachep
;
218 /* SLAB cache for vm_area_struct structures */
219 struct kmem_cache
*vm_area_cachep
;
221 /* SLAB cache for mm_struct structures (tsk->mm) */
222 static struct kmem_cache
*mm_cachep
;
224 static void account_kernel_stack(unsigned long *stack
, int account
)
226 struct zone
*zone
= page_zone(virt_to_page(stack
));
228 mod_zone_page_state(zone
, NR_KERNEL_STACK
, account
);
231 void free_task(struct task_struct
*tsk
)
233 account_kernel_stack(tsk
->stack
, -1);
234 arch_release_thread_stack(tsk
->stack
);
235 free_thread_stack(tsk
->stack
);
236 rt_mutex_debug_task_free(tsk
);
237 ftrace_graph_exit_task(tsk
);
238 put_seccomp_filter(tsk
);
239 arch_release_task_struct(tsk
);
240 free_task_struct(tsk
);
242 EXPORT_SYMBOL(free_task
);
244 static inline void free_signal_struct(struct signal_struct
*sig
)
246 taskstats_tgid_free(sig
);
247 sched_autogroup_exit(sig
);
248 kmem_cache_free(signal_cachep
, sig
);
251 static inline void put_signal_struct(struct signal_struct
*sig
)
253 if (atomic_dec_and_test(&sig
->sigcnt
))
254 free_signal_struct(sig
);
257 void __put_task_struct(struct task_struct
*tsk
)
259 WARN_ON(!tsk
->exit_state
);
260 WARN_ON(atomic_read(&tsk
->usage
));
261 WARN_ON(tsk
== current
);
265 security_task_free(tsk
);
267 delayacct_tsk_free(tsk
);
268 put_signal_struct(tsk
->signal
);
270 if (!profile_handoff_task(tsk
))
273 EXPORT_SYMBOL_GPL(__put_task_struct
);
275 void __init __weak
arch_task_cache_init(void) { }
280 static void set_max_threads(unsigned int max_threads_suggested
)
285 * The number of threads shall be limited such that the thread
286 * structures may only consume a small part of the available memory.
288 if (fls64(totalram_pages
) + fls64(PAGE_SIZE
) > 64)
289 threads
= MAX_THREADS
;
291 threads
= div64_u64((u64
) totalram_pages
* (u64
) PAGE_SIZE
,
292 (u64
) THREAD_SIZE
* 8UL);
294 if (threads
> max_threads_suggested
)
295 threads
= max_threads_suggested
;
297 max_threads
= clamp_t(u64
, threads
, MIN_THREADS
, MAX_THREADS
);
300 #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
301 /* Initialized by the architecture: */
302 int arch_task_struct_size __read_mostly
;
305 void __init
fork_init(void)
307 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
308 #ifndef ARCH_MIN_TASKALIGN
309 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
311 /* create a slab on which task_structs can be allocated */
313 kmem_cache_create("task_struct", arch_task_struct_size
,
314 ARCH_MIN_TASKALIGN
, SLAB_PANIC
| SLAB_NOTRACK
, NULL
);
317 /* do the arch specific task caches init */
318 arch_task_cache_init();
320 set_max_threads(MAX_THREADS
);
322 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_cur
= max_threads
/2;
323 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_max
= max_threads
/2;
324 init_task
.signal
->rlim
[RLIMIT_SIGPENDING
] =
325 init_task
.signal
->rlim
[RLIMIT_NPROC
];
328 int __weak
arch_dup_task_struct(struct task_struct
*dst
,
329 struct task_struct
*src
)
335 void set_task_stack_end_magic(struct task_struct
*tsk
)
337 unsigned long *stackend
;
339 stackend
= end_of_stack(tsk
);
340 *stackend
= STACK_END_MAGIC
; /* for overflow detection */
343 static struct task_struct
*dup_task_struct(struct task_struct
*orig
, int node
)
345 struct task_struct
*tsk
;
346 unsigned long *stack
;
349 if (node
== NUMA_NO_NODE
)
350 node
= tsk_fork_get_node(orig
);
351 tsk
= alloc_task_struct_node(node
);
355 stack
= alloc_thread_stack_node(tsk
, node
);
359 err
= arch_dup_task_struct(tsk
, orig
);
365 err
= kaiser_map_thread_stack(tsk
->stack
);
368 #ifdef CONFIG_SECCOMP
370 * We must handle setting up seccomp filters once we're under
371 * the sighand lock in case orig has changed between now and
372 * then. Until then, filter must be NULL to avoid messing up
373 * the usage counts on the error path calling free_task.
375 tsk
->seccomp
.filter
= NULL
;
378 setup_thread_stack(tsk
, orig
);
379 clear_user_return_notifier(tsk
);
380 clear_tsk_need_resched(tsk
);
381 set_task_stack_end_magic(tsk
);
383 #ifdef CONFIG_CC_STACKPROTECTOR
384 tsk
->stack_canary
= get_random_long();
388 * One for us, one for whoever does the "release_task()" (usually
391 atomic_set(&tsk
->usage
, 2);
392 #ifdef CONFIG_BLK_DEV_IO_TRACE
395 tsk
->splice_pipe
= NULL
;
396 tsk
->task_frag
.page
= NULL
;
397 tsk
->wake_q
.next
= NULL
;
399 account_kernel_stack(stack
, 1);
406 free_thread_stack(stack
);
408 free_task_struct(tsk
);
413 static int dup_mmap(struct mm_struct
*mm
, struct mm_struct
*oldmm
)
415 struct vm_area_struct
*mpnt
, *tmp
, *prev
, **pprev
;
416 struct rb_node
**rb_link
, *rb_parent
;
418 unsigned long charge
;
420 uprobe_start_dup_mmap();
421 down_write(&oldmm
->mmap_sem
);
422 flush_cache_dup_mm(oldmm
);
423 uprobe_dup_mmap(oldmm
, mm
);
425 * Not linked in yet - no deadlock potential:
427 down_write_nested(&mm
->mmap_sem
, SINGLE_DEPTH_NESTING
);
429 /* No ordering required: file already has been exposed. */
430 RCU_INIT_POINTER(mm
->exe_file
, get_mm_exe_file(oldmm
));
432 mm
->total_vm
= oldmm
->total_vm
;
433 mm
->shared_vm
= oldmm
->shared_vm
;
434 mm
->exec_vm
= oldmm
->exec_vm
;
435 mm
->stack_vm
= oldmm
->stack_vm
;
437 rb_link
= &mm
->mm_rb
.rb_node
;
440 retval
= ksm_fork(mm
, oldmm
);
443 retval
= khugepaged_fork(mm
, oldmm
);
448 for (mpnt
= oldmm
->mmap
; mpnt
; mpnt
= mpnt
->vm_next
) {
451 if (mpnt
->vm_flags
& VM_DONTCOPY
) {
452 vm_stat_account(mm
, mpnt
->vm_flags
, mpnt
->vm_file
,
457 if (mpnt
->vm_flags
& VM_ACCOUNT
) {
458 unsigned long len
= vma_pages(mpnt
);
460 if (security_vm_enough_memory_mm(oldmm
, len
)) /* sic */
464 tmp
= kmem_cache_alloc(vm_area_cachep
, GFP_KERNEL
);
468 INIT_LIST_HEAD(&tmp
->anon_vma_chain
);
469 retval
= vma_dup_policy(mpnt
, tmp
);
471 goto fail_nomem_policy
;
473 if (anon_vma_fork(tmp
, mpnt
))
474 goto fail_nomem_anon_vma_fork
;
476 ~(VM_LOCKED
|VM_LOCKONFAULT
|VM_UFFD_MISSING
|VM_UFFD_WP
);
477 tmp
->vm_next
= tmp
->vm_prev
= NULL
;
478 tmp
->vm_userfaultfd_ctx
= NULL_VM_UFFD_CTX
;
481 struct inode
*inode
= file_inode(file
);
482 struct address_space
*mapping
= file
->f_mapping
;
485 if (tmp
->vm_flags
& VM_DENYWRITE
)
486 atomic_dec(&inode
->i_writecount
);
487 i_mmap_lock_write(mapping
);
488 if (tmp
->vm_flags
& VM_SHARED
)
489 atomic_inc(&mapping
->i_mmap_writable
);
490 flush_dcache_mmap_lock(mapping
);
491 /* insert tmp into the share list, just after mpnt */
492 vma_interval_tree_insert_after(tmp
, mpnt
,
494 flush_dcache_mmap_unlock(mapping
);
495 i_mmap_unlock_write(mapping
);
499 * Clear hugetlb-related page reserves for children. This only
500 * affects MAP_PRIVATE mappings. Faults generated by the child
501 * are not guaranteed to succeed, even if read-only
503 if (is_vm_hugetlb_page(tmp
))
504 reset_vma_resv_huge_pages(tmp
);
507 * Link in the new vma and copy the page table entries.
510 pprev
= &tmp
->vm_next
;
514 __vma_link_rb(mm
, tmp
, rb_link
, rb_parent
);
515 rb_link
= &tmp
->vm_rb
.rb_right
;
516 rb_parent
= &tmp
->vm_rb
;
519 retval
= copy_page_range(mm
, oldmm
, mpnt
);
521 if (tmp
->vm_ops
&& tmp
->vm_ops
->open
)
522 tmp
->vm_ops
->open(tmp
);
527 /* a new mm has just been created */
528 arch_dup_mmap(oldmm
, mm
);
531 up_write(&mm
->mmap_sem
);
533 up_write(&oldmm
->mmap_sem
);
534 uprobe_end_dup_mmap();
536 fail_nomem_anon_vma_fork
:
537 mpol_put(vma_policy(tmp
));
539 kmem_cache_free(vm_area_cachep
, tmp
);
542 vm_unacct_memory(charge
);
546 static inline int mm_alloc_pgd(struct mm_struct
*mm
)
548 mm
->pgd
= pgd_alloc(mm
);
549 if (unlikely(!mm
->pgd
))
554 static inline void mm_free_pgd(struct mm_struct
*mm
)
556 pgd_free(mm
, mm
->pgd
);
559 static int dup_mmap(struct mm_struct
*mm
, struct mm_struct
*oldmm
)
561 down_write(&oldmm
->mmap_sem
);
562 RCU_INIT_POINTER(mm
->exe_file
, get_mm_exe_file(oldmm
));
563 up_write(&oldmm
->mmap_sem
);
566 #define mm_alloc_pgd(mm) (0)
567 #define mm_free_pgd(mm)
568 #endif /* CONFIG_MMU */
570 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(mmlist_lock
);
572 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
573 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
575 static unsigned long default_dump_filter
= MMF_DUMP_FILTER_DEFAULT
;
577 static int __init
coredump_filter_setup(char *s
)
579 default_dump_filter
=
580 (simple_strtoul(s
, NULL
, 0) << MMF_DUMP_FILTER_SHIFT
) &
581 MMF_DUMP_FILTER_MASK
;
585 __setup("coredump_filter=", coredump_filter_setup
);
587 #include <linux/init_task.h>
589 static void mm_init_aio(struct mm_struct
*mm
)
592 spin_lock_init(&mm
->ioctx_lock
);
593 mm
->ioctx_table
= NULL
;
597 static void mm_init_owner(struct mm_struct
*mm
, struct task_struct
*p
)
604 static struct mm_struct
*mm_init(struct mm_struct
*mm
, struct task_struct
*p
,
605 struct user_namespace
*user_ns
)
609 mm
->vmacache_seqnum
= 0;
610 atomic_set(&mm
->mm_users
, 1);
611 atomic_set(&mm
->mm_count
, 1);
612 init_rwsem(&mm
->mmap_sem
);
613 INIT_LIST_HEAD(&mm
->mmlist
);
614 mm
->core_state
= NULL
;
615 atomic_long_set(&mm
->nr_ptes
, 0);
620 memset(&mm
->rss_stat
, 0, sizeof(mm
->rss_stat
));
621 spin_lock_init(&mm
->page_table_lock
);
624 mm_init_owner(mm
, p
);
625 mmu_notifier_mm_init(mm
);
626 clear_tlb_flush_pending(mm
);
627 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
628 mm
->pmd_huge_pte
= NULL
;
632 mm
->flags
= current
->mm
->flags
& MMF_INIT_MASK
;
633 mm
->def_flags
= current
->mm
->def_flags
& VM_INIT_DEF_MASK
;
635 mm
->flags
= default_dump_filter
;
639 if (mm_alloc_pgd(mm
))
642 if (init_new_context(p
, mm
))
645 mm
->user_ns
= get_user_ns(user_ns
);
655 static void check_mm(struct mm_struct
*mm
)
659 for (i
= 0; i
< NR_MM_COUNTERS
; i
++) {
660 long x
= atomic_long_read(&mm
->rss_stat
.count
[i
]);
663 printk(KERN_ALERT
"BUG: Bad rss-counter state "
664 "mm:%p idx:%d val:%ld\n", mm
, i
, x
);
667 if (atomic_long_read(&mm
->nr_ptes
))
668 pr_alert("BUG: non-zero nr_ptes on freeing mm: %ld\n",
669 atomic_long_read(&mm
->nr_ptes
));
671 pr_alert("BUG: non-zero nr_pmds on freeing mm: %ld\n",
674 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
675 VM_BUG_ON_MM(mm
->pmd_huge_pte
, mm
);
680 * Allocate and initialize an mm_struct.
682 struct mm_struct
*mm_alloc(void)
684 struct mm_struct
*mm
;
690 memset(mm
, 0, sizeof(*mm
));
691 return mm_init(mm
, current
, current_user_ns());
695 * Called when the last reference to the mm
696 * is dropped: either by a lazy thread or by
697 * mmput. Free the page directory and the mm.
699 void __mmdrop(struct mm_struct
*mm
)
701 BUG_ON(mm
== &init_mm
);
704 mmu_notifier_mm_destroy(mm
);
706 put_user_ns(mm
->user_ns
);
709 EXPORT_SYMBOL_GPL(__mmdrop
);
711 static inline void __mmput(struct mm_struct
*mm
)
713 VM_BUG_ON(atomic_read(&mm
->mm_users
));
715 uprobe_clear_state(mm
);
718 khugepaged_exit(mm
); /* must run before exit_mmap */
720 set_mm_exe_file(mm
, NULL
);
721 if (!list_empty(&mm
->mmlist
)) {
722 spin_lock(&mmlist_lock
);
723 list_del(&mm
->mmlist
);
724 spin_unlock(&mmlist_lock
);
727 module_put(mm
->binfmt
->module
);
732 * Decrement the use count and release all resources for an mm.
734 void mmput(struct mm_struct
*mm
)
738 if (atomic_dec_and_test(&mm
->mm_users
))
741 EXPORT_SYMBOL_GPL(mmput
);
743 static void mmput_async_fn(struct work_struct
*work
)
745 struct mm_struct
*mm
= container_of(work
, struct mm_struct
, async_put_work
);
749 void mmput_async(struct mm_struct
*mm
)
751 if (atomic_dec_and_test(&mm
->mm_users
)) {
752 INIT_WORK(&mm
->async_put_work
, mmput_async_fn
);
753 schedule_work(&mm
->async_put_work
);
758 * set_mm_exe_file - change a reference to the mm's executable file
760 * This changes mm's executable file (shown as symlink /proc/[pid]/exe).
762 * Main users are mmput() and sys_execve(). Callers prevent concurrent
763 * invocations: in mmput() nobody alive left, in execve task is single
764 * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the
765 * mm->exe_file, but does so without using set_mm_exe_file() in order
766 * to do avoid the need for any locks.
768 void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
)
770 struct file
*old_exe_file
;
773 * It is safe to dereference the exe_file without RCU as
774 * this function is only called if nobody else can access
775 * this mm -- see comment above for justification.
777 old_exe_file
= rcu_dereference_raw(mm
->exe_file
);
780 get_file(new_exe_file
);
781 rcu_assign_pointer(mm
->exe_file
, new_exe_file
);
787 * get_mm_exe_file - acquire a reference to the mm's executable file
789 * Returns %NULL if mm has no associated executable file.
790 * User must release file via fput().
792 struct file
*get_mm_exe_file(struct mm_struct
*mm
)
794 struct file
*exe_file
;
797 exe_file
= rcu_dereference(mm
->exe_file
);
798 if (exe_file
&& !get_file_rcu(exe_file
))
803 EXPORT_SYMBOL(get_mm_exe_file
);
806 * get_task_exe_file - acquire a reference to the task's executable file
808 * Returns %NULL if task's mm (if any) has no associated executable file or
809 * this is a kernel thread with borrowed mm (see the comment above get_task_mm).
810 * User must release file via fput().
812 struct file
*get_task_exe_file(struct task_struct
*task
)
814 struct file
*exe_file
= NULL
;
815 struct mm_struct
*mm
;
820 if (!(task
->flags
& PF_KTHREAD
))
821 exe_file
= get_mm_exe_file(mm
);
826 EXPORT_SYMBOL(get_task_exe_file
);
829 * get_task_mm - acquire a reference to the task's mm
831 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
832 * this kernel workthread has transiently adopted a user mm with use_mm,
833 * to do its AIO) is not set and if so returns a reference to it, after
834 * bumping up the use count. User must release the mm via mmput()
835 * after use. Typically used by /proc and ptrace.
837 struct mm_struct
*get_task_mm(struct task_struct
*task
)
839 struct mm_struct
*mm
;
844 if (task
->flags
& PF_KTHREAD
)
847 atomic_inc(&mm
->mm_users
);
852 EXPORT_SYMBOL_GPL(get_task_mm
);
854 struct mm_struct
*mm_access(struct task_struct
*task
, unsigned int mode
)
856 struct mm_struct
*mm
;
859 err
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
863 mm
= get_task_mm(task
);
864 if (mm
&& mm
!= current
->mm
&&
865 !ptrace_may_access(task
, mode
)) {
867 mm
= ERR_PTR(-EACCES
);
869 mutex_unlock(&task
->signal
->cred_guard_mutex
);
874 static void complete_vfork_done(struct task_struct
*tsk
)
876 struct completion
*vfork
;
879 vfork
= tsk
->vfork_done
;
881 tsk
->vfork_done
= NULL
;
887 static int wait_for_vfork_done(struct task_struct
*child
,
888 struct completion
*vfork
)
892 freezer_do_not_count();
893 killed
= wait_for_completion_killable(vfork
);
898 child
->vfork_done
= NULL
;
902 put_task_struct(child
);
906 /* Please note the differences between mmput and mm_release.
907 * mmput is called whenever we stop holding onto a mm_struct,
908 * error success whatever.
910 * mm_release is called after a mm_struct has been removed
911 * from the current process.
913 * This difference is important for error handling, when we
914 * only half set up a mm_struct for a new process and need to restore
915 * the old one. Because we mmput the new mm_struct before
916 * restoring the old one. . .
917 * Eric Biederman 10 January 1998
919 void mm_release(struct task_struct
*tsk
, struct mm_struct
*mm
)
921 /* Get rid of any futexes when releasing the mm */
923 if (unlikely(tsk
->robust_list
)) {
924 exit_robust_list(tsk
);
925 tsk
->robust_list
= NULL
;
928 if (unlikely(tsk
->compat_robust_list
)) {
929 compat_exit_robust_list(tsk
);
930 tsk
->compat_robust_list
= NULL
;
933 if (unlikely(!list_empty(&tsk
->pi_state_list
)))
934 exit_pi_state_list(tsk
);
937 uprobe_free_utask(tsk
);
939 /* Get rid of any cached register state */
940 deactivate_mm(tsk
, mm
);
943 * Signal userspace if we're not exiting with a core dump
944 * because we want to leave the value intact for debugging
947 if (tsk
->clear_child_tid
) {
948 if (!(tsk
->signal
->flags
& SIGNAL_GROUP_COREDUMP
) &&
949 atomic_read(&mm
->mm_users
) > 1) {
951 * We don't check the error code - if userspace has
952 * not set up a proper pointer then tough luck.
954 put_user(0, tsk
->clear_child_tid
);
955 sys_futex(tsk
->clear_child_tid
, FUTEX_WAKE
,
958 tsk
->clear_child_tid
= NULL
;
962 * All done, finally we can wake up parent and return this mm to him.
963 * Also kthread_stop() uses this completion for synchronization.
966 complete_vfork_done(tsk
);
970 * Allocate a new mm structure and copy contents from the
971 * mm structure of the passed in task structure.
973 static struct mm_struct
*dup_mm(struct task_struct
*tsk
)
975 struct mm_struct
*mm
, *oldmm
= current
->mm
;
982 memcpy(mm
, oldmm
, sizeof(*mm
));
984 if (!mm_init(mm
, tsk
, mm
->user_ns
))
987 err
= dup_mmap(mm
, oldmm
);
991 mm
->hiwater_rss
= get_mm_rss(mm
);
992 mm
->hiwater_vm
= mm
->total_vm
;
994 if (mm
->binfmt
&& !try_module_get(mm
->binfmt
->module
))
1000 /* don't put binfmt in mmput, we haven't got module yet */
1008 static int copy_mm(unsigned long clone_flags
, struct task_struct
*tsk
)
1010 struct mm_struct
*mm
, *oldmm
;
1013 tsk
->min_flt
= tsk
->maj_flt
= 0;
1014 tsk
->nvcsw
= tsk
->nivcsw
= 0;
1015 #ifdef CONFIG_DETECT_HUNG_TASK
1016 tsk
->last_switch_count
= tsk
->nvcsw
+ tsk
->nivcsw
;
1020 tsk
->active_mm
= NULL
;
1023 * Are we cloning a kernel thread?
1025 * We need to steal a active VM for that..
1027 oldmm
= current
->mm
;
1031 /* initialize the new vmacache entries */
1032 vmacache_flush(tsk
);
1034 if (clone_flags
& CLONE_VM
) {
1035 atomic_inc(&oldmm
->mm_users
);
1047 tsk
->active_mm
= mm
;
1054 static int copy_fs(unsigned long clone_flags
, struct task_struct
*tsk
)
1056 struct fs_struct
*fs
= current
->fs
;
1057 if (clone_flags
& CLONE_FS
) {
1058 /* tsk->fs is already what we want */
1059 spin_lock(&fs
->lock
);
1061 spin_unlock(&fs
->lock
);
1065 spin_unlock(&fs
->lock
);
1068 tsk
->fs
= copy_fs_struct(fs
);
1074 static int copy_files(unsigned long clone_flags
, struct task_struct
*tsk
)
1076 struct files_struct
*oldf
, *newf
;
1080 * A background process may not have any files ...
1082 oldf
= current
->files
;
1086 if (clone_flags
& CLONE_FILES
) {
1087 atomic_inc(&oldf
->count
);
1091 newf
= dup_fd(oldf
, &error
);
1101 static int copy_io(unsigned long clone_flags
, struct task_struct
*tsk
)
1104 struct io_context
*ioc
= current
->io_context
;
1105 struct io_context
*new_ioc
;
1110 * Share io context with parent, if CLONE_IO is set
1112 if (clone_flags
& CLONE_IO
) {
1114 tsk
->io_context
= ioc
;
1115 } else if (ioprio_valid(ioc
->ioprio
)) {
1116 new_ioc
= get_task_io_context(tsk
, GFP_KERNEL
, NUMA_NO_NODE
);
1117 if (unlikely(!new_ioc
))
1120 new_ioc
->ioprio
= ioc
->ioprio
;
1121 put_io_context(new_ioc
);
1127 static int copy_sighand(unsigned long clone_flags
, struct task_struct
*tsk
)
1129 struct sighand_struct
*sig
;
1131 if (clone_flags
& CLONE_SIGHAND
) {
1132 atomic_inc(¤t
->sighand
->count
);
1135 sig
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
1136 rcu_assign_pointer(tsk
->sighand
, sig
);
1140 atomic_set(&sig
->count
, 1);
1141 memcpy(sig
->action
, current
->sighand
->action
, sizeof(sig
->action
));
1145 void __cleanup_sighand(struct sighand_struct
*sighand
)
1147 if (atomic_dec_and_test(&sighand
->count
)) {
1148 signalfd_cleanup(sighand
);
1150 * sighand_cachep is SLAB_DESTROY_BY_RCU so we can free it
1151 * without an RCU grace period, see __lock_task_sighand().
1153 kmem_cache_free(sighand_cachep
, sighand
);
1158 * Initialize POSIX timer handling for a thread group.
1160 static void posix_cpu_timers_init_group(struct signal_struct
*sig
)
1162 unsigned long cpu_limit
;
1164 cpu_limit
= READ_ONCE(sig
->rlim
[RLIMIT_CPU
].rlim_cur
);
1165 if (cpu_limit
!= RLIM_INFINITY
) {
1166 sig
->cputime_expires
.prof_exp
= secs_to_cputime(cpu_limit
);
1167 sig
->cputimer
.running
= true;
1170 /* The timer lists. */
1171 INIT_LIST_HEAD(&sig
->cpu_timers
[0]);
1172 INIT_LIST_HEAD(&sig
->cpu_timers
[1]);
1173 INIT_LIST_HEAD(&sig
->cpu_timers
[2]);
1176 static int copy_signal(unsigned long clone_flags
, struct task_struct
*tsk
)
1178 struct signal_struct
*sig
;
1180 if (clone_flags
& CLONE_THREAD
)
1183 sig
= kmem_cache_zalloc(signal_cachep
, GFP_KERNEL
);
1188 sig
->nr_threads
= 1;
1189 atomic_set(&sig
->live
, 1);
1190 atomic_set(&sig
->sigcnt
, 1);
1192 /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */
1193 sig
->thread_head
= (struct list_head
)LIST_HEAD_INIT(tsk
->thread_node
);
1194 tsk
->thread_node
= (struct list_head
)LIST_HEAD_INIT(sig
->thread_head
);
1196 init_waitqueue_head(&sig
->wait_chldexit
);
1197 sig
->curr_target
= tsk
;
1198 init_sigpending(&sig
->shared_pending
);
1199 INIT_LIST_HEAD(&sig
->posix_timers
);
1200 seqlock_init(&sig
->stats_lock
);
1201 prev_cputime_init(&sig
->prev_cputime
);
1203 hrtimer_init(&sig
->real_timer
, CLOCK_MONOTONIC
, HRTIMER_MODE_REL
);
1204 sig
->real_timer
.function
= it_real_fn
;
1206 task_lock(current
->group_leader
);
1207 memcpy(sig
->rlim
, current
->signal
->rlim
, sizeof sig
->rlim
);
1208 task_unlock(current
->group_leader
);
1210 posix_cpu_timers_init_group(sig
);
1212 tty_audit_fork(sig
);
1213 sched_autogroup_fork(sig
);
1215 sig
->oom_score_adj
= current
->signal
->oom_score_adj
;
1216 sig
->oom_score_adj_min
= current
->signal
->oom_score_adj_min
;
1218 sig
->has_child_subreaper
= current
->signal
->has_child_subreaper
||
1219 current
->signal
->is_child_subreaper
;
1221 mutex_init(&sig
->cred_guard_mutex
);
1226 static void copy_seccomp(struct task_struct
*p
)
1228 #ifdef CONFIG_SECCOMP
1230 * Must be called with sighand->lock held, which is common to
1231 * all threads in the group. Holding cred_guard_mutex is not
1232 * needed because this new task is not yet running and cannot
1235 assert_spin_locked(¤t
->sighand
->siglock
);
1237 /* Ref-count the new filter user, and assign it. */
1238 get_seccomp_filter(current
);
1239 p
->seccomp
= current
->seccomp
;
1242 * Explicitly enable no_new_privs here in case it got set
1243 * between the task_struct being duplicated and holding the
1244 * sighand lock. The seccomp state and nnp must be in sync.
1246 if (task_no_new_privs(current
))
1247 task_set_no_new_privs(p
);
1250 * If the parent gained a seccomp mode after copying thread
1251 * flags and between before we held the sighand lock, we have
1252 * to manually enable the seccomp thread flag here.
1254 if (p
->seccomp
.mode
!= SECCOMP_MODE_DISABLED
)
1255 set_tsk_thread_flag(p
, TIF_SECCOMP
);
1259 SYSCALL_DEFINE1(set_tid_address
, int __user
*, tidptr
)
1261 current
->clear_child_tid
= tidptr
;
1263 return task_pid_vnr(current
);
1266 static void rt_mutex_init_task(struct task_struct
*p
)
1268 raw_spin_lock_init(&p
->pi_lock
);
1269 #ifdef CONFIG_RT_MUTEXES
1270 p
->pi_waiters
= RB_ROOT
;
1271 p
->pi_waiters_leftmost
= NULL
;
1272 p
->pi_blocked_on
= NULL
;
1277 * Initialize POSIX timer handling for a single task.
1279 static void posix_cpu_timers_init(struct task_struct
*tsk
)
1281 tsk
->cputime_expires
.prof_exp
= 0;
1282 tsk
->cputime_expires
.virt_exp
= 0;
1283 tsk
->cputime_expires
.sched_exp
= 0;
1284 INIT_LIST_HEAD(&tsk
->cpu_timers
[0]);
1285 INIT_LIST_HEAD(&tsk
->cpu_timers
[1]);
1286 INIT_LIST_HEAD(&tsk
->cpu_timers
[2]);
1289 #ifdef CONFIG_RKP_KDP
1290 void rkp_assign_pgd(struct task_struct
*p
)
1293 pgd
= (u64
)(p
->mm
? p
->mm
->pgd
:swapper_pg_dir
);
1294 rkp_call(RKP_CMDID(0x43),(u64
)p
->cred
, (u64
)pgd
,0,0,0);
1296 #endif /*CONFIG_RKP_KDP*/
1299 init_task_pid(struct task_struct
*task
, enum pid_type type
, struct pid
*pid
)
1301 task
->pids
[type
].pid
= pid
;
1305 * This creates a new process as a copy of the old one,
1306 * but does not actually start it yet.
1308 * It copies the registers, and all the appropriate
1309 * parts of the process environment (as per the clone
1310 * flags). The actual kick-off is left to the caller.
1312 static struct task_struct
*copy_process(unsigned long clone_flags
,
1313 unsigned long stack_start
,
1314 unsigned long stack_size
,
1315 int __user
*child_tidptr
,
1322 struct task_struct
*p
;
1323 void *cgrp_ss_priv
[CGROUP_CANFORK_COUNT
] = {};
1325 if ((clone_flags
& (CLONE_NEWNS
|CLONE_FS
)) == (CLONE_NEWNS
|CLONE_FS
))
1326 return ERR_PTR(-EINVAL
);
1328 if ((clone_flags
& (CLONE_NEWUSER
|CLONE_FS
)) == (CLONE_NEWUSER
|CLONE_FS
))
1329 return ERR_PTR(-EINVAL
);
1332 * Thread groups must share signals as well, and detached threads
1333 * can only be started up within the thread group.
1335 if ((clone_flags
& CLONE_THREAD
) && !(clone_flags
& CLONE_SIGHAND
))
1336 return ERR_PTR(-EINVAL
);
1339 * Shared signal handlers imply shared VM. By way of the above,
1340 * thread groups also imply shared VM. Blocking this case allows
1341 * for various simplifications in other code.
1343 if ((clone_flags
& CLONE_SIGHAND
) && !(clone_flags
& CLONE_VM
))
1344 return ERR_PTR(-EINVAL
);
1347 * Siblings of global init remain as zombies on exit since they are
1348 * not reaped by their parent (swapper). To solve this and to avoid
1349 * multi-rooted process trees, prevent global and container-inits
1350 * from creating siblings.
1352 if ((clone_flags
& CLONE_PARENT
) &&
1353 current
->signal
->flags
& SIGNAL_UNKILLABLE
)
1354 return ERR_PTR(-EINVAL
);
1357 * If the new process will be in a different pid or user namespace
1358 * do not allow it to share a thread group with the forking task.
1360 if (clone_flags
& CLONE_THREAD
) {
1361 if ((clone_flags
& (CLONE_NEWUSER
| CLONE_NEWPID
)) ||
1362 (task_active_pid_ns(current
) !=
1363 current
->nsproxy
->pid_ns_for_children
))
1364 return ERR_PTR(-EINVAL
);
1367 retval
= security_task_create(clone_flags
);
1372 p
= dup_task_struct(current
, node
);
1376 ftrace_graph_init_task(p
);
1378 rt_mutex_init_task(p
);
1380 #ifdef CONFIG_PROVE_LOCKING
1381 DEBUG_LOCKS_WARN_ON(!p
->hardirqs_enabled
);
1382 DEBUG_LOCKS_WARN_ON(!p
->softirqs_enabled
);
1385 if (atomic_read(&p
->real_cred
->user
->processes
) >=
1386 task_rlimit(p
, RLIMIT_NPROC
)) {
1387 if (p
->real_cred
->user
!= INIT_USER
&&
1388 !capable(CAP_SYS_RESOURCE
) && !capable(CAP_SYS_ADMIN
))
1391 current
->flags
&= ~PF_NPROC_EXCEEDED
;
1393 retval
= copy_creds(p
, clone_flags
);
1398 * If multiple threads are within copy_process(), then this check
1399 * triggers too late. This doesn't hurt, the check is only there
1400 * to stop root fork bombs.
1403 if (nr_threads
>= max_threads
)
1404 goto bad_fork_cleanup_count
;
1406 delayacct_tsk_init(p
); /* Must remain after dup_task_struct() */
1407 p
->flags
&= ~(PF_SUPERPRIV
| PF_WQ_WORKER
);
1408 p
->flags
|= PF_FORKNOEXEC
;
1409 INIT_LIST_HEAD(&p
->children
);
1410 INIT_LIST_HEAD(&p
->sibling
);
1411 rcu_copy_process(p
);
1412 p
->vfork_done
= NULL
;
1413 spin_lock_init(&p
->alloc_lock
);
1415 init_sigpending(&p
->pending
);
1417 p
->utime
= p
->stime
= p
->gtime
= 0;
1418 p
->utimescaled
= p
->stimescaled
= 0;
1419 prev_cputime_init(&p
->prev_cputime
);
1422 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1423 seqlock_init(&p
->vtime_seqlock
);
1425 p
->vtime_snap_whence
= VTIME_SLEEPING
;
1428 #if defined(SPLIT_RSS_COUNTING)
1429 memset(&p
->rss_stat
, 0, sizeof(p
->rss_stat
));
1432 p
->default_timer_slack_ns
= current
->timer_slack_ns
;
1434 task_io_accounting_init(&p
->ioac
);
1435 acct_clear_integrals(p
);
1437 posix_cpu_timers_init(p
);
1439 p
->start_time
= ktime_get_ns();
1440 p
->real_start_time
= ktime_get_boot_ns();
1441 p
->io_context
= NULL
;
1442 p
->audit_context
= NULL
;
1445 p
->mempolicy
= mpol_dup(p
->mempolicy
);
1446 if (IS_ERR(p
->mempolicy
)) {
1447 retval
= PTR_ERR(p
->mempolicy
);
1448 p
->mempolicy
= NULL
;
1449 goto bad_fork_cleanup_threadgroup_lock
;
1452 #ifdef CONFIG_CPUSETS
1453 p
->cpuset_mem_spread_rotor
= NUMA_NO_NODE
;
1454 p
->cpuset_slab_spread_rotor
= NUMA_NO_NODE
;
1455 seqcount_init(&p
->mems_allowed_seq
);
1457 #ifdef CONFIG_TRACE_IRQFLAGS
1459 p
->hardirqs_enabled
= 0;
1460 p
->hardirq_enable_ip
= 0;
1461 p
->hardirq_enable_event
= 0;
1462 p
->hardirq_disable_ip
= _THIS_IP_
;
1463 p
->hardirq_disable_event
= 0;
1464 p
->softirqs_enabled
= 1;
1465 p
->softirq_enable_ip
= _THIS_IP_
;
1466 p
->softirq_enable_event
= 0;
1467 p
->softirq_disable_ip
= 0;
1468 p
->softirq_disable_event
= 0;
1469 p
->hardirq_context
= 0;
1470 p
->softirq_context
= 0;
1473 p
->pagefault_disabled
= 0;
1475 #ifdef CONFIG_LOCKDEP
1476 p
->lockdep_depth
= 0; /* no locks held yet */
1477 p
->curr_chain_key
= 0;
1478 p
->lockdep_recursion
= 0;
1481 #ifdef CONFIG_DEBUG_MUTEXES
1482 p
->blocked_on
= NULL
; /* not blocked yet */
1484 #ifdef CONFIG_BCACHE
1485 p
->sequential_io
= 0;
1486 p
->sequential_io_avg
= 0;
1489 /* Perform scheduler related setup. Assign this task to a CPU. */
1490 retval
= sched_fork(clone_flags
, p
);
1492 goto bad_fork_cleanup_policy
;
1494 retval
= perf_event_init_task(p
);
1496 goto bad_fork_cleanup_policy
;
1497 retval
= audit_alloc(p
);
1499 goto bad_fork_cleanup_perf
;
1500 /* copy all the process information */
1502 retval
= copy_semundo(clone_flags
, p
);
1504 goto bad_fork_cleanup_audit
;
1505 retval
= copy_files(clone_flags
, p
);
1507 goto bad_fork_cleanup_semundo
;
1508 retval
= copy_fs(clone_flags
, p
);
1510 goto bad_fork_cleanup_files
;
1511 retval
= copy_sighand(clone_flags
, p
);
1513 goto bad_fork_cleanup_fs
;
1514 retval
= copy_signal(clone_flags
, p
);
1516 goto bad_fork_cleanup_sighand
;
1517 retval
= copy_mm(clone_flags
, p
);
1519 goto bad_fork_cleanup_signal
;
1520 retval
= copy_namespaces(clone_flags
, p
);
1522 goto bad_fork_cleanup_mm
;
1523 retval
= copy_io(clone_flags
, p
);
1525 goto bad_fork_cleanup_namespaces
;
1526 retval
= copy_thread_tls(clone_flags
, stack_start
, stack_size
, p
, tls
);
1528 goto bad_fork_cleanup_io
;
1530 if (pid
!= &init_struct_pid
) {
1531 pid
= alloc_pid(p
->nsproxy
->pid_ns_for_children
);
1533 retval
= PTR_ERR(pid
);
1534 goto bad_fork_cleanup_io
;
1538 p
->set_child_tid
= (clone_flags
& CLONE_CHILD_SETTID
) ? child_tidptr
: NULL
;
1540 * Clear TID on mm_release()?
1542 p
->clear_child_tid
= (clone_flags
& CLONE_CHILD_CLEARTID
) ? child_tidptr
: NULL
;
1547 p
->robust_list
= NULL
;
1548 #ifdef CONFIG_COMPAT
1549 p
->compat_robust_list
= NULL
;
1551 INIT_LIST_HEAD(&p
->pi_state_list
);
1552 p
->pi_state_cache
= NULL
;
1555 * sigaltstack should be cleared when sharing the same VM
1557 if ((clone_flags
& (CLONE_VM
|CLONE_VFORK
)) == CLONE_VM
)
1558 p
->sas_ss_sp
= p
->sas_ss_size
= 0;
1561 * Syscall tracing and stepping should be turned off in the
1562 * child regardless of CLONE_PTRACE.
1564 user_disable_single_step(p
);
1565 clear_tsk_thread_flag(p
, TIF_SYSCALL_TRACE
);
1566 #ifdef TIF_SYSCALL_EMU
1567 clear_tsk_thread_flag(p
, TIF_SYSCALL_EMU
);
1569 clear_all_latency_tracing(p
);
1571 /* ok, now we should be set up.. */
1572 p
->pid
= pid_nr(pid
);
1573 if (clone_flags
& CLONE_THREAD
) {
1574 p
->exit_signal
= -1;
1575 p
->group_leader
= current
->group_leader
;
1576 p
->tgid
= current
->tgid
;
1578 if (clone_flags
& CLONE_PARENT
)
1579 p
->exit_signal
= current
->group_leader
->exit_signal
;
1581 p
->exit_signal
= (clone_flags
& CSIGNAL
);
1582 p
->group_leader
= p
;
1587 p
->nr_dirtied_pause
= 128 >> (PAGE_SHIFT
- 10);
1588 p
->dirty_paused_when
= 0;
1590 p
->pdeath_signal
= 0;
1591 INIT_LIST_HEAD(&p
->thread_group
);
1592 p
->task_works
= NULL
;
1594 threadgroup_change_begin(current
);
1596 * Ensure that the cgroup subsystem policies allow the new process to be
1597 * forked. It should be noted the the new process's css_set can be changed
1598 * between here and cgroup_post_fork() if an organisation operation is in
1601 retval
= cgroup_can_fork(p
, cgrp_ss_priv
);
1603 goto bad_fork_free_pid
;
1606 * Make it visible to the rest of the system, but dont wake it up yet.
1607 * Need tasklist lock for parent etc handling!
1609 write_lock_irq(&tasklist_lock
);
1611 /* CLONE_PARENT re-uses the old parent */
1612 if (clone_flags
& (CLONE_PARENT
|CLONE_THREAD
)) {
1613 p
->real_parent
= current
->real_parent
;
1614 p
->parent_exec_id
= current
->parent_exec_id
;
1616 p
->real_parent
= current
;
1617 p
->parent_exec_id
= current
->self_exec_id
;
1620 spin_lock(¤t
->sighand
->siglock
);
1623 * Copy seccomp details explicitly here, in case they were changed
1624 * before holding sighand lock.
1629 * Process group and session signals need to be delivered to just the
1630 * parent before the fork or both the parent and the child after the
1631 * fork. Restart if a signal comes in before we add the new process to
1632 * it's process group.
1633 * A fatal signal pending means that current will exit, so the new
1634 * thread can't slip out of an OOM kill (or normal SIGKILL).
1636 recalc_sigpending();
1637 if (signal_pending(current
)) {
1638 retval
= -ERESTARTNOINTR
;
1639 goto bad_fork_cancel_cgroup
;
1641 if (unlikely(!(ns_of_pid(pid
)->nr_hashed
& PIDNS_HASH_ADDING
))) {
1643 goto bad_fork_cancel_cgroup
;
1646 if (likely(p
->pid
)) {
1647 ptrace_init_task(p
, (clone_flags
& CLONE_PTRACE
) || trace
);
1649 init_task_pid(p
, PIDTYPE_PID
, pid
);
1650 if (thread_group_leader(p
)) {
1651 init_task_pid(p
, PIDTYPE_PGID
, task_pgrp(current
));
1652 init_task_pid(p
, PIDTYPE_SID
, task_session(current
));
1654 if (is_child_reaper(pid
)) {
1655 ns_of_pid(pid
)->child_reaper
= p
;
1656 p
->signal
->flags
|= SIGNAL_UNKILLABLE
;
1659 p
->signal
->leader_pid
= pid
;
1660 p
->signal
->tty
= tty_kref_get(current
->signal
->tty
);
1661 list_add_tail(&p
->sibling
, &p
->real_parent
->children
);
1662 list_add_tail_rcu(&p
->tasks
, &init_task
.tasks
);
1663 attach_pid(p
, PIDTYPE_PGID
);
1664 attach_pid(p
, PIDTYPE_SID
);
1665 __this_cpu_inc(process_counts
);
1667 current
->signal
->nr_threads
++;
1668 atomic_inc(¤t
->signal
->live
);
1669 atomic_inc(¤t
->signal
->sigcnt
);
1670 list_add_tail_rcu(&p
->thread_group
,
1671 &p
->group_leader
->thread_group
);
1672 list_add_tail_rcu(&p
->thread_node
,
1673 &p
->signal
->thread_head
);
1675 attach_pid(p
, PIDTYPE_PID
);
1680 spin_unlock(¤t
->sighand
->siglock
);
1681 syscall_tracepoint_update(p
);
1682 write_unlock_irq(&tasklist_lock
);
1684 proc_fork_connector(p
);
1685 cgroup_post_fork(p
, cgrp_ss_priv
);
1686 threadgroup_change_end(current
);
1689 trace_task_newtask(p
, clone_flags
);
1690 uprobe_copy_process(p
, clone_flags
);
1691 #ifdef CONFIG_RKP_KDP
1694 #endif/*CONFIG_RKP_KDP*/
1697 bad_fork_cancel_cgroup
:
1698 spin_unlock(¤t
->sighand
->siglock
);
1699 write_unlock_irq(&tasklist_lock
);
1700 cgroup_cancel_fork(p
, cgrp_ss_priv
);
1702 threadgroup_change_end(current
);
1703 if (pid
!= &init_struct_pid
)
1705 bad_fork_cleanup_io
:
1708 bad_fork_cleanup_namespaces
:
1709 exit_task_namespaces(p
);
1710 bad_fork_cleanup_mm
:
1713 bad_fork_cleanup_signal
:
1714 if (!(clone_flags
& CLONE_THREAD
))
1715 free_signal_struct(p
->signal
);
1716 bad_fork_cleanup_sighand
:
1717 __cleanup_sighand(p
->sighand
);
1718 bad_fork_cleanup_fs
:
1719 exit_fs(p
); /* blocking */
1720 bad_fork_cleanup_files
:
1721 exit_files(p
); /* blocking */
1722 bad_fork_cleanup_semundo
:
1724 bad_fork_cleanup_audit
:
1726 bad_fork_cleanup_perf
:
1727 perf_event_free_task(p
);
1728 bad_fork_cleanup_policy
:
1730 mpol_put(p
->mempolicy
);
1731 bad_fork_cleanup_threadgroup_lock
:
1733 delayacct_tsk_free(p
);
1734 bad_fork_cleanup_count
:
1735 atomic_dec(&p
->cred
->user
->processes
);
1740 return ERR_PTR(retval
);
1743 static inline void init_idle_pids(struct pid_link
*links
)
1747 for (type
= PIDTYPE_PID
; type
< PIDTYPE_MAX
; ++type
) {
1748 INIT_HLIST_NODE(&links
[type
].node
); /* not really needed */
1749 links
[type
].pid
= &init_struct_pid
;
1753 struct task_struct
*fork_idle(int cpu
)
1755 struct task_struct
*task
;
1756 task
= copy_process(CLONE_VM
, 0, 0, NULL
, &init_struct_pid
, 0, 0,
1758 if (!IS_ERR(task
)) {
1759 init_idle_pids(task
->pids
);
1760 init_idle(task
, cpu
);
1767 * Ok, this is the main fork-routine.
1769 * It copies the process, and if successful kick-starts
1770 * it and waits for it to finish using the VM if required.
1772 long _do_fork(unsigned long clone_flags
,
1773 unsigned long stack_start
,
1774 unsigned long stack_size
,
1775 int __user
*parent_tidptr
,
1776 int __user
*child_tidptr
,
1779 struct task_struct
*p
;
1784 * Determine whether and which event to report to ptracer. When
1785 * called from kernel_thread or CLONE_UNTRACED is explicitly
1786 * requested, no event is reported; otherwise, report if the event
1787 * for the type of forking is enabled.
1789 if (!(clone_flags
& CLONE_UNTRACED
)) {
1790 if (clone_flags
& CLONE_VFORK
)
1791 trace
= PTRACE_EVENT_VFORK
;
1792 else if ((clone_flags
& CSIGNAL
) != SIGCHLD
)
1793 trace
= PTRACE_EVENT_CLONE
;
1795 trace
= PTRACE_EVENT_FORK
;
1797 if (likely(!ptrace_event_enabled(current
, trace
)))
1801 p
= copy_process(clone_flags
, stack_start
, stack_size
,
1802 child_tidptr
, NULL
, trace
, tls
, NUMA_NO_NODE
);
1804 * Do this prior waking up the new thread - the thread pointer
1805 * might get invalid after that point, if the thread exits quickly.
1808 struct completion vfork
;
1811 trace_sched_process_fork(current
, p
);
1813 pid
= get_task_pid(p
, PIDTYPE_PID
);
1816 #ifdef CONFIG_SECURITY_DEFEX
1817 task_defex_zero_creds(p
);
1820 if (clone_flags
& CLONE_PARENT_SETTID
)
1821 put_user(nr
, parent_tidptr
);
1823 if (clone_flags
& CLONE_VFORK
) {
1824 p
->vfork_done
= &vfork
;
1825 init_completion(&vfork
);
1829 wake_up_new_task(p
);
1831 /* forking complete and child started to run, tell ptracer */
1832 if (unlikely(trace
))
1833 ptrace_event_pid(trace
, pid
);
1835 if (clone_flags
& CLONE_VFORK
) {
1836 if (!wait_for_vfork_done(p
, &vfork
))
1837 ptrace_event_pid(PTRACE_EVENT_VFORK_DONE
, pid
);
1847 #ifndef CONFIG_HAVE_COPY_THREAD_TLS
1848 /* For compatibility with architectures that call do_fork directly rather than
1849 * using the syscall entry points below. */
1850 long do_fork(unsigned long clone_flags
,
1851 unsigned long stack_start
,
1852 unsigned long stack_size
,
1853 int __user
*parent_tidptr
,
1854 int __user
*child_tidptr
)
1856 return _do_fork(clone_flags
, stack_start
, stack_size
,
1857 parent_tidptr
, child_tidptr
, 0);
1862 * Create a kernel thread.
1864 pid_t
kernel_thread(int (*fn
)(void *), void *arg
, unsigned long flags
)
1866 return _do_fork(flags
|CLONE_VM
|CLONE_UNTRACED
, (unsigned long)fn
,
1867 (unsigned long)arg
, NULL
, NULL
, 0);
1870 #ifdef __ARCH_WANT_SYS_FORK
1871 SYSCALL_DEFINE0(fork
)
1874 return _do_fork(SIGCHLD
, 0, 0, NULL
, NULL
, 0);
1876 /* can not support in nommu mode */
1882 #ifdef __ARCH_WANT_SYS_VFORK
1883 SYSCALL_DEFINE0(vfork
)
1885 return _do_fork(CLONE_VFORK
| CLONE_VM
| SIGCHLD
, 0,
1890 #ifdef __ARCH_WANT_SYS_CLONE
1891 #ifdef CONFIG_CLONE_BACKWARDS
1892 SYSCALL_DEFINE5(clone
, unsigned long, clone_flags
, unsigned long, newsp
,
1893 int __user
*, parent_tidptr
,
1895 int __user
*, child_tidptr
)
1896 #elif defined(CONFIG_CLONE_BACKWARDS2)
1897 SYSCALL_DEFINE5(clone
, unsigned long, newsp
, unsigned long, clone_flags
,
1898 int __user
*, parent_tidptr
,
1899 int __user
*, child_tidptr
,
1901 #elif defined(CONFIG_CLONE_BACKWARDS3)
1902 SYSCALL_DEFINE6(clone
, unsigned long, clone_flags
, unsigned long, newsp
,
1904 int __user
*, parent_tidptr
,
1905 int __user
*, child_tidptr
,
1908 SYSCALL_DEFINE5(clone
, unsigned long, clone_flags
, unsigned long, newsp
,
1909 int __user
*, parent_tidptr
,
1910 int __user
*, child_tidptr
,
1914 return _do_fork(clone_flags
, newsp
, 0, parent_tidptr
, child_tidptr
, tls
);
1918 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1919 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1922 static void sighand_ctor(void *data
)
1924 struct sighand_struct
*sighand
= data
;
1926 spin_lock_init(&sighand
->siglock
);
1927 init_waitqueue_head(&sighand
->signalfd_wqh
);
1930 void __init
proc_caches_init(void)
1932 sighand_cachep
= kmem_cache_create("sighand_cache",
1933 sizeof(struct sighand_struct
), 0,
1934 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_DESTROY_BY_RCU
|
1935 SLAB_NOTRACK
, sighand_ctor
);
1936 signal_cachep
= kmem_cache_create("signal_cache",
1937 sizeof(struct signal_struct
), 0,
1938 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1939 files_cachep
= kmem_cache_create("files_cache",
1940 sizeof(struct files_struct
), 0,
1941 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1942 fs_cachep
= kmem_cache_create("fs_cache",
1943 sizeof(struct fs_struct
), 0,
1944 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1946 * FIXME! The "sizeof(struct mm_struct)" currently includes the
1947 * whole struct cpumask for the OFFSTACK case. We could change
1948 * this to *only* allocate as much of it as required by the
1949 * maximum number of CPU's we can ever have. The cpumask_allocation
1950 * is at the end of the structure, exactly for that reason.
1952 mm_cachep
= kmem_cache_create("mm_struct",
1953 sizeof(struct mm_struct
), ARCH_MIN_MMSTRUCT_ALIGN
,
1954 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
, NULL
);
1955 vm_area_cachep
= KMEM_CACHE(vm_area_struct
, SLAB_PANIC
);
1957 nsproxy_cache_init();
1961 * Check constraints on flags passed to the unshare system call.
1963 static int check_unshare_flags(unsigned long unshare_flags
)
1965 if (unshare_flags
& ~(CLONE_THREAD
|CLONE_FS
|CLONE_NEWNS
|CLONE_SIGHAND
|
1966 CLONE_VM
|CLONE_FILES
|CLONE_SYSVSEM
|
1967 CLONE_NEWUTS
|CLONE_NEWIPC
|CLONE_NEWNET
|
1968 CLONE_NEWUSER
|CLONE_NEWPID
))
1971 * Not implemented, but pretend it works if there is nothing
1972 * to unshare. Note that unsharing the address space or the
1973 * signal handlers also need to unshare the signal queues (aka
1976 if (unshare_flags
& (CLONE_THREAD
| CLONE_SIGHAND
| CLONE_VM
)) {
1977 if (!thread_group_empty(current
))
1980 if (unshare_flags
& (CLONE_SIGHAND
| CLONE_VM
)) {
1981 if (atomic_read(¤t
->sighand
->count
) > 1)
1984 if (unshare_flags
& CLONE_VM
) {
1985 if (!current_is_single_threaded())
1993 * Unshare the filesystem structure if it is being shared
1995 static int unshare_fs(unsigned long unshare_flags
, struct fs_struct
**new_fsp
)
1997 struct fs_struct
*fs
= current
->fs
;
1999 if (!(unshare_flags
& CLONE_FS
) || !fs
)
2002 /* don't need lock here; in the worst case we'll do useless copy */
2006 *new_fsp
= copy_fs_struct(fs
);
2014 * Unshare file descriptor table if it is being shared
2016 static int unshare_fd(unsigned long unshare_flags
, struct files_struct
**new_fdp
)
2018 struct files_struct
*fd
= current
->files
;
2021 if ((unshare_flags
& CLONE_FILES
) &&
2022 (fd
&& atomic_read(&fd
->count
) > 1)) {
2023 *new_fdp
= dup_fd(fd
, &error
);
2032 * unshare allows a process to 'unshare' part of the process
2033 * context which was originally shared using clone. copy_*
2034 * functions used by do_fork() cannot be used here directly
2035 * because they modify an inactive task_struct that is being
2036 * constructed. Here we are modifying the current, active,
2039 SYSCALL_DEFINE1(unshare
, unsigned long, unshare_flags
)
2041 struct fs_struct
*fs
, *new_fs
= NULL
;
2042 struct files_struct
*fd
, *new_fd
= NULL
;
2043 struct cred
*new_cred
= NULL
;
2044 struct nsproxy
*new_nsproxy
= NULL
;
2049 * If unsharing a user namespace must also unshare the thread group
2050 * and unshare the filesystem root and working directories.
2052 if (unshare_flags
& CLONE_NEWUSER
)
2053 unshare_flags
|= CLONE_THREAD
| CLONE_FS
;
2055 * If unsharing vm, must also unshare signal handlers.
2057 if (unshare_flags
& CLONE_VM
)
2058 unshare_flags
|= CLONE_SIGHAND
;
2060 * If unsharing a signal handlers, must also unshare the signal queues.
2062 if (unshare_flags
& CLONE_SIGHAND
)
2063 unshare_flags
|= CLONE_THREAD
;
2065 * If unsharing namespace, must also unshare filesystem information.
2067 if (unshare_flags
& CLONE_NEWNS
)
2068 unshare_flags
|= CLONE_FS
;
2070 err
= check_unshare_flags(unshare_flags
);
2072 goto bad_unshare_out
;
2074 * CLONE_NEWIPC must also detach from the undolist: after switching
2075 * to a new ipc namespace, the semaphore arrays from the old
2076 * namespace are unreachable.
2078 if (unshare_flags
& (CLONE_NEWIPC
|CLONE_SYSVSEM
))
2080 err
= unshare_fs(unshare_flags
, &new_fs
);
2082 goto bad_unshare_out
;
2083 err
= unshare_fd(unshare_flags
, &new_fd
);
2085 goto bad_unshare_cleanup_fs
;
2086 err
= unshare_userns(unshare_flags
, &new_cred
);
2088 goto bad_unshare_cleanup_fd
;
2089 err
= unshare_nsproxy_namespaces(unshare_flags
, &new_nsproxy
,
2092 goto bad_unshare_cleanup_cred
;
2094 if (new_fs
|| new_fd
|| do_sysvsem
|| new_cred
|| new_nsproxy
) {
2097 * CLONE_SYSVSEM is equivalent to sys_exit().
2101 if (unshare_flags
& CLONE_NEWIPC
) {
2102 /* Orphan segments in old ns (see sem above). */
2104 shm_init_task(current
);
2108 switch_task_namespaces(current
, new_nsproxy
);
2114 spin_lock(&fs
->lock
);
2115 current
->fs
= new_fs
;
2120 spin_unlock(&fs
->lock
);
2124 fd
= current
->files
;
2125 current
->files
= new_fd
;
2129 task_unlock(current
);
2132 /* Install the new user namespace */
2133 commit_creds(new_cred
);
2138 bad_unshare_cleanup_cred
:
2141 bad_unshare_cleanup_fd
:
2143 put_files_struct(new_fd
);
2145 bad_unshare_cleanup_fs
:
2147 free_fs_struct(new_fs
);
2154 * Helper to unshare the files of the current task.
2155 * We don't want to expose copy_files internals to
2156 * the exec layer of the kernel.
2159 int unshare_files(struct files_struct
**displaced
)
2161 struct task_struct
*task
= current
;
2162 struct files_struct
*copy
= NULL
;
2165 error
= unshare_fd(CLONE_FILES
, ©
);
2166 if (error
|| !copy
) {
2170 *displaced
= task
->files
;
2177 int sysctl_max_threads(struct ctl_table
*table
, int write
,
2178 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
2182 int threads
= max_threads
;
2183 int min
= MIN_THREADS
;
2184 int max
= MAX_THREADS
;
2191 ret
= proc_dointvec_minmax(&t
, write
, buffer
, lenp
, ppos
);
2195 set_max_threads(threads
);