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/sched/autogroup.h>
16 #include <linux/sched/mm.h>
17 #include <linux/sched/coredump.h>
18 #include <linux/sched/user.h>
19 #include <linux/sched/numa_balancing.h>
20 #include <linux/sched/stat.h>
21 #include <linux/sched/task.h>
22 #include <linux/sched/task_stack.h>
23 #include <linux/sched/cputime.h>
24 #include <linux/rtmutex.h>
25 #include <linux/init.h>
26 #include <linux/unistd.h>
27 #include <linux/module.h>
28 #include <linux/vmalloc.h>
29 #include <linux/completion.h>
30 #include <linux/personality.h>
31 #include <linux/mempolicy.h>
32 #include <linux/sem.h>
33 #include <linux/file.h>
34 #include <linux/fdtable.h>
35 #include <linux/iocontext.h>
36 #include <linux/key.h>
37 #include <linux/binfmts.h>
38 #include <linux/mman.h>
39 #include <linux/mmu_notifier.h>
40 #include <linux/hmm.h>
43 #include <linux/vmacache.h>
44 #include <linux/nsproxy.h>
45 #include <linux/capability.h>
46 #include <linux/cpu.h>
47 #include <linux/cgroup.h>
48 #include <linux/security.h>
49 #include <linux/hugetlb.h>
50 #include <linux/seccomp.h>
51 #include <linux/swap.h>
52 #include <linux/syscalls.h>
53 #include <linux/jiffies.h>
54 #include <linux/futex.h>
55 #include <linux/compat.h>
56 #include <linux/kthread.h>
57 #include <linux/task_io_accounting_ops.h>
58 #include <linux/rcupdate.h>
59 #include <linux/ptrace.h>
60 #include <linux/mount.h>
61 #include <linux/audit.h>
62 #include <linux/memcontrol.h>
63 #include <linux/ftrace.h>
64 #include <linux/proc_fs.h>
65 #include <linux/profile.h>
66 #include <linux/rmap.h>
67 #include <linux/ksm.h>
68 #include <linux/acct.h>
69 #include <linux/userfaultfd_k.h>
70 #include <linux/tsacct_kern.h>
71 #include <linux/cn_proc.h>
72 #include <linux/freezer.h>
73 #include <linux/delayacct.h>
74 #include <linux/taskstats_kern.h>
75 #include <linux/random.h>
76 #include <linux/tty.h>
77 #include <linux/blkdev.h>
78 #include <linux/fs_struct.h>
79 #include <linux/magic.h>
80 #include <linux/sched/mm.h>
81 #include <linux/perf_event.h>
82 #include <linux/posix-timers.h>
83 #include <linux/user-return-notifier.h>
84 #include <linux/oom.h>
85 #include <linux/khugepaged.h>
86 #include <linux/signalfd.h>
87 #include <linux/uprobes.h>
88 #include <linux/aio.h>
89 #include <linux/compiler.h>
90 #include <linux/sysctl.h>
91 #include <linux/kcov.h>
92 #include <linux/livepatch.h>
93 #include <linux/thread_info.h>
94 #include <linux/stackleak.h>
96 #include <asm/pgtable.h>
97 #include <asm/pgalloc.h>
98 #include <linux/uaccess.h>
99 #include <asm/mmu_context.h>
100 #include <asm/cacheflush.h>
101 #include <asm/tlbflush.h>
103 #include <trace/events/sched.h>
105 #define CREATE_TRACE_POINTS
106 #include <trace/events/task.h>
109 * Minimum number of threads to boot the kernel
111 #define MIN_THREADS 20
114 * Maximum number of threads
116 #define MAX_THREADS FUTEX_TID_MASK
119 * Protected counters by write_lock_irq(&tasklist_lock)
121 unsigned long total_forks; /* Handle normal Linux uptimes. */
122 int nr_threads; /* The idle threads do not count.. */
124 int max_threads; /* tunable limit on nr_threads */
126 DEFINE_PER_CPU(unsigned long, process_counts) = 0;
128 __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
130 #ifdef CONFIG_PROVE_RCU
131 int lockdep_tasklist_lock_is_held(void)
133 return lockdep_is_held(&tasklist_lock);
135 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
136 #endif /* #ifdef CONFIG_PROVE_RCU */
138 int nr_processes(void)
143 for_each_possible_cpu(cpu)
144 total += per_cpu(process_counts, cpu);
149 void __weak arch_release_task_struct(struct task_struct *tsk)
153 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
154 static struct kmem_cache *task_struct_cachep;
156 static inline struct task_struct *alloc_task_struct_node(int node)
158 return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node);
161 static inline void free_task_struct(struct task_struct *tsk)
163 kmem_cache_free(task_struct_cachep, tsk);
167 #ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR
170 * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a
171 * kmemcache based allocator.
173 # if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK)
175 #ifdef CONFIG_VMAP_STACK
177 * vmalloc() is a bit slow, and calling vfree() enough times will force a TLB
178 * flush. Try to minimize the number of calls by caching stacks.
180 #define NR_CACHED_STACKS 2
181 static DEFINE_PER_CPU(struct vm_struct *, cached_stacks[NR_CACHED_STACKS]);
183 static int free_vm_stack_cache(unsigned int cpu)
185 struct vm_struct **cached_vm_stacks = per_cpu_ptr(cached_stacks, cpu);
188 for (i = 0; i < NR_CACHED_STACKS; i++) {
189 struct vm_struct *vm_stack = cached_vm_stacks[i];
194 vfree(vm_stack->addr);
195 cached_vm_stacks[i] = NULL;
202 static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node)
204 #ifdef CONFIG_VMAP_STACK
208 for (i = 0; i < NR_CACHED_STACKS; i++) {
211 s = this_cpu_xchg(cached_stacks[i], NULL);
216 /* Clear stale pointers from reused stack. */
217 memset(s->addr, 0, THREAD_SIZE);
219 tsk->stack_vm_area = s;
220 tsk->stack = s->addr;
225 * Allocated stacks are cached and later reused by new threads,
226 * so memcg accounting is performed manually on assigning/releasing
227 * stacks to tasks. Drop __GFP_ACCOUNT.
229 stack = __vmalloc_node_range(THREAD_SIZE, THREAD_ALIGN,
230 VMALLOC_START, VMALLOC_END,
231 THREADINFO_GFP & ~__GFP_ACCOUNT,
233 0, node, __builtin_return_address(0));
236 * We can't call find_vm_area() in interrupt context, and
237 * free_thread_stack() can be called in interrupt context,
238 * so cache the vm_struct.
241 tsk->stack_vm_area = find_vm_area(stack);
246 struct page *page = alloc_pages_node(node, THREADINFO_GFP,
249 return page ? page_address(page) : NULL;
253 static inline void free_thread_stack(struct task_struct *tsk)
255 #ifdef CONFIG_VMAP_STACK
256 struct vm_struct *vm = task_stack_vm_area(tsk);
261 for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) {
262 mod_memcg_page_state(vm->pages[i],
263 MEMCG_KERNEL_STACK_KB,
264 -(int)(PAGE_SIZE / 1024));
266 memcg_kmem_uncharge(vm->pages[i], 0);
269 for (i = 0; i < NR_CACHED_STACKS; i++) {
270 if (this_cpu_cmpxchg(cached_stacks[i],
271 NULL, tsk->stack_vm_area) != NULL)
277 vfree_atomic(tsk->stack);
282 __free_pages(virt_to_page(tsk->stack), THREAD_SIZE_ORDER);
285 static struct kmem_cache *thread_stack_cache;
287 static unsigned long *alloc_thread_stack_node(struct task_struct *tsk,
290 unsigned long *stack;
291 stack = kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node);
296 static void free_thread_stack(struct task_struct *tsk)
298 kmem_cache_free(thread_stack_cache, tsk->stack);
301 void thread_stack_cache_init(void)
303 thread_stack_cache = kmem_cache_create_usercopy("thread_stack",
304 THREAD_SIZE, THREAD_SIZE, 0, 0,
306 BUG_ON(thread_stack_cache == NULL);
311 /* SLAB cache for signal_struct structures (tsk->signal) */
312 static struct kmem_cache *signal_cachep;
314 /* SLAB cache for sighand_struct structures (tsk->sighand) */
315 struct kmem_cache *sighand_cachep;
317 /* SLAB cache for files_struct structures (tsk->files) */
318 struct kmem_cache *files_cachep;
320 /* SLAB cache for fs_struct structures (tsk->fs) */
321 struct kmem_cache *fs_cachep;
323 /* SLAB cache for vm_area_struct structures */
324 static struct kmem_cache *vm_area_cachep;
326 /* SLAB cache for mm_struct structures (tsk->mm) */
327 static struct kmem_cache *mm_cachep;
329 struct vm_area_struct *vm_area_alloc(struct mm_struct *mm)
331 struct vm_area_struct *vma;
333 vma = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
339 struct vm_area_struct *vm_area_dup(struct vm_area_struct *orig)
341 struct vm_area_struct *new = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
345 INIT_LIST_HEAD(&new->anon_vma_chain);
350 void vm_area_free(struct vm_area_struct *vma)
352 kmem_cache_free(vm_area_cachep, vma);
355 static void account_kernel_stack(struct task_struct *tsk, int account)
357 void *stack = task_stack_page(tsk);
358 struct vm_struct *vm = task_stack_vm_area(tsk);
360 BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK) && PAGE_SIZE % 1024 != 0);
365 BUG_ON(vm->nr_pages != THREAD_SIZE / PAGE_SIZE);
367 for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) {
368 mod_zone_page_state(page_zone(vm->pages[i]),
370 PAGE_SIZE / 1024 * account);
374 * All stack pages are in the same zone and belong to the
377 struct page *first_page = virt_to_page(stack);
379 mod_zone_page_state(page_zone(first_page), NR_KERNEL_STACK_KB,
380 THREAD_SIZE / 1024 * account);
382 mod_memcg_page_state(first_page, MEMCG_KERNEL_STACK_KB,
383 account * (THREAD_SIZE / 1024));
387 static int memcg_charge_kernel_stack(struct task_struct *tsk)
389 #ifdef CONFIG_VMAP_STACK
390 struct vm_struct *vm = task_stack_vm_area(tsk);
396 for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) {
398 * If memcg_kmem_charge() fails, page->mem_cgroup
399 * pointer is NULL, and both memcg_kmem_uncharge()
400 * and mod_memcg_page_state() in free_thread_stack()
401 * will ignore this page. So it's safe.
403 ret = memcg_kmem_charge(vm->pages[i], GFP_KERNEL, 0);
407 mod_memcg_page_state(vm->pages[i],
408 MEMCG_KERNEL_STACK_KB,
416 static void release_task_stack(struct task_struct *tsk)
418 if (WARN_ON(tsk->state != TASK_DEAD))
419 return; /* Better to leak the stack than to free prematurely */
421 account_kernel_stack(tsk, -1);
422 free_thread_stack(tsk);
424 #ifdef CONFIG_VMAP_STACK
425 tsk->stack_vm_area = NULL;
429 #ifdef CONFIG_THREAD_INFO_IN_TASK
430 void put_task_stack(struct task_struct *tsk)
432 if (atomic_dec_and_test(&tsk->stack_refcount))
433 release_task_stack(tsk);
437 void free_task(struct task_struct *tsk)
439 #ifndef CONFIG_THREAD_INFO_IN_TASK
441 * The task is finally done with both the stack and thread_info,
444 release_task_stack(tsk);
447 * If the task had a separate stack allocation, it should be gone
450 WARN_ON_ONCE(atomic_read(&tsk->stack_refcount) != 0);
452 rt_mutex_debug_task_free(tsk);
453 ftrace_graph_exit_task(tsk);
454 put_seccomp_filter(tsk);
455 arch_release_task_struct(tsk);
456 if (tsk->flags & PF_KTHREAD)
457 free_kthread_struct(tsk);
458 free_task_struct(tsk);
460 EXPORT_SYMBOL(free_task);
463 static __latent_entropy int dup_mmap(struct mm_struct *mm,
464 struct mm_struct *oldmm)
466 struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
467 struct rb_node **rb_link, *rb_parent;
469 unsigned long charge;
472 uprobe_start_dup_mmap();
473 if (down_write_killable(&oldmm->mmap_sem)) {
475 goto fail_uprobe_end;
477 flush_cache_dup_mm(oldmm);
478 uprobe_dup_mmap(oldmm, mm);
480 * Not linked in yet - no deadlock potential:
482 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
484 /* No ordering required: file already has been exposed. */
485 RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
487 mm->total_vm = oldmm->total_vm;
488 mm->data_vm = oldmm->data_vm;
489 mm->exec_vm = oldmm->exec_vm;
490 mm->stack_vm = oldmm->stack_vm;
492 rb_link = &mm->mm_rb.rb_node;
495 retval = ksm_fork(mm, oldmm);
498 retval = khugepaged_fork(mm, oldmm);
503 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
506 if (mpnt->vm_flags & VM_DONTCOPY) {
507 vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt));
512 * Don't duplicate many vmas if we've been oom-killed (for
515 if (fatal_signal_pending(current)) {
519 if (mpnt->vm_flags & VM_ACCOUNT) {
520 unsigned long len = vma_pages(mpnt);
522 if (security_vm_enough_memory_mm(oldmm, len)) /* sic */
526 tmp = vm_area_dup(mpnt);
529 retval = vma_dup_policy(mpnt, tmp);
531 goto fail_nomem_policy;
533 retval = dup_userfaultfd(tmp, &uf);
535 goto fail_nomem_anon_vma_fork;
536 if (tmp->vm_flags & VM_WIPEONFORK) {
537 /* VM_WIPEONFORK gets a clean slate in the child. */
538 tmp->anon_vma = NULL;
539 if (anon_vma_prepare(tmp))
540 goto fail_nomem_anon_vma_fork;
541 } else if (anon_vma_fork(tmp, mpnt))
542 goto fail_nomem_anon_vma_fork;
543 tmp->vm_flags &= ~(VM_LOCKED | VM_LOCKONFAULT);
544 tmp->vm_next = tmp->vm_prev = NULL;
547 struct inode *inode = file_inode(file);
548 struct address_space *mapping = file->f_mapping;
551 if (tmp->vm_flags & VM_DENYWRITE)
552 atomic_dec(&inode->i_writecount);
553 i_mmap_lock_write(mapping);
554 if (tmp->vm_flags & VM_SHARED)
555 atomic_inc(&mapping->i_mmap_writable);
556 flush_dcache_mmap_lock(mapping);
557 /* insert tmp into the share list, just after mpnt */
558 vma_interval_tree_insert_after(tmp, mpnt,
560 flush_dcache_mmap_unlock(mapping);
561 i_mmap_unlock_write(mapping);
565 * Clear hugetlb-related page reserves for children. This only
566 * affects MAP_PRIVATE mappings. Faults generated by the child
567 * are not guaranteed to succeed, even if read-only
569 if (is_vm_hugetlb_page(tmp))
570 reset_vma_resv_huge_pages(tmp);
573 * Link in the new vma and copy the page table entries.
576 pprev = &tmp->vm_next;
580 __vma_link_rb(mm, tmp, rb_link, rb_parent);
581 rb_link = &tmp->vm_rb.rb_right;
582 rb_parent = &tmp->vm_rb;
585 if (!(tmp->vm_flags & VM_WIPEONFORK))
586 retval = copy_page_range(mm, oldmm, mpnt);
588 if (tmp->vm_ops && tmp->vm_ops->open)
589 tmp->vm_ops->open(tmp);
594 /* a new mm has just been created */
595 retval = arch_dup_mmap(oldmm, mm);
597 up_write(&mm->mmap_sem);
599 up_write(&oldmm->mmap_sem);
600 dup_userfaultfd_complete(&uf);
602 uprobe_end_dup_mmap();
604 fail_nomem_anon_vma_fork:
605 mpol_put(vma_policy(tmp));
610 vm_unacct_memory(charge);
614 static inline int mm_alloc_pgd(struct mm_struct *mm)
616 mm->pgd = pgd_alloc(mm);
617 if (unlikely(!mm->pgd))
622 static inline void mm_free_pgd(struct mm_struct *mm)
624 pgd_free(mm, mm->pgd);
627 static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
629 down_write(&oldmm->mmap_sem);
630 RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
631 up_write(&oldmm->mmap_sem);
634 #define mm_alloc_pgd(mm) (0)
635 #define mm_free_pgd(mm)
636 #endif /* CONFIG_MMU */
638 static void check_mm(struct mm_struct *mm)
642 for (i = 0; i < NR_MM_COUNTERS; i++) {
643 long x = atomic_long_read(&mm->rss_stat.count[i]);
646 printk(KERN_ALERT "BUG: Bad rss-counter state "
647 "mm:%p idx:%d val:%ld\n", mm, i, x);
650 if (mm_pgtables_bytes(mm))
651 pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n",
652 mm_pgtables_bytes(mm));
654 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
655 VM_BUG_ON_MM(mm->pmd_huge_pte, mm);
659 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
660 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
663 * Called when the last reference to the mm
664 * is dropped: either by a lazy thread or by
665 * mmput. Free the page directory and the mm.
667 void __mmdrop(struct mm_struct *mm)
669 BUG_ON(mm == &init_mm);
670 WARN_ON_ONCE(mm == current->mm);
671 WARN_ON_ONCE(mm == current->active_mm);
675 mmu_notifier_mm_destroy(mm);
677 put_user_ns(mm->user_ns);
680 EXPORT_SYMBOL_GPL(__mmdrop);
682 static void mmdrop_async_fn(struct work_struct *work)
684 struct mm_struct *mm;
686 mm = container_of(work, struct mm_struct, async_put_work);
690 static void mmdrop_async(struct mm_struct *mm)
692 if (unlikely(atomic_dec_and_test(&mm->mm_count))) {
693 INIT_WORK(&mm->async_put_work, mmdrop_async_fn);
694 schedule_work(&mm->async_put_work);
698 static inline void free_signal_struct(struct signal_struct *sig)
700 taskstats_tgid_free(sig);
701 sched_autogroup_exit(sig);
703 * __mmdrop is not safe to call from softirq context on x86 due to
704 * pgd_dtor so postpone it to the async context
707 mmdrop_async(sig->oom_mm);
708 kmem_cache_free(signal_cachep, sig);
711 static inline void put_signal_struct(struct signal_struct *sig)
713 if (atomic_dec_and_test(&sig->sigcnt))
714 free_signal_struct(sig);
717 void __put_task_struct(struct task_struct *tsk)
719 WARN_ON(!tsk->exit_state);
720 WARN_ON(atomic_read(&tsk->usage));
721 WARN_ON(tsk == current);
725 security_task_free(tsk);
727 delayacct_tsk_free(tsk);
728 put_signal_struct(tsk->signal);
730 if (!profile_handoff_task(tsk))
733 EXPORT_SYMBOL_GPL(__put_task_struct);
735 void __init __weak arch_task_cache_init(void) { }
740 static void set_max_threads(unsigned int max_threads_suggested)
743 unsigned long nr_pages = totalram_pages();
746 * The number of threads shall be limited such that the thread
747 * structures may only consume a small part of the available memory.
749 if (fls64(nr_pages) + fls64(PAGE_SIZE) > 64)
750 threads = MAX_THREADS;
752 threads = div64_u64((u64) nr_pages * (u64) PAGE_SIZE,
753 (u64) THREAD_SIZE * 8UL);
755 if (threads > max_threads_suggested)
756 threads = max_threads_suggested;
758 max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS);
761 #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
762 /* Initialized by the architecture: */
763 int arch_task_struct_size __read_mostly;
766 static void task_struct_whitelist(unsigned long *offset, unsigned long *size)
768 /* Fetch thread_struct whitelist for the architecture. */
769 arch_thread_struct_whitelist(offset, size);
772 * Handle zero-sized whitelist or empty thread_struct, otherwise
773 * adjust offset to position of thread_struct in task_struct.
775 if (unlikely(*size == 0))
778 *offset += offsetof(struct task_struct, thread);
781 void __init fork_init(void)
784 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
785 #ifndef ARCH_MIN_TASKALIGN
786 #define ARCH_MIN_TASKALIGN 0
788 int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN);
789 unsigned long useroffset, usersize;
791 /* create a slab on which task_structs can be allocated */
792 task_struct_whitelist(&useroffset, &usersize);
793 task_struct_cachep = kmem_cache_create_usercopy("task_struct",
794 arch_task_struct_size, align,
795 SLAB_PANIC|SLAB_ACCOUNT,
796 useroffset, usersize, NULL);
799 /* do the arch specific task caches init */
800 arch_task_cache_init();
802 set_max_threads(MAX_THREADS);
804 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
805 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
806 init_task.signal->rlim[RLIMIT_SIGPENDING] =
807 init_task.signal->rlim[RLIMIT_NPROC];
809 for (i = 0; i < UCOUNT_COUNTS; i++) {
810 init_user_ns.ucount_max[i] = max_threads/2;
813 #ifdef CONFIG_VMAP_STACK
814 cpuhp_setup_state(CPUHP_BP_PREPARE_DYN, "fork:vm_stack_cache",
815 NULL, free_vm_stack_cache);
818 lockdep_init_task(&init_task);
821 int __weak arch_dup_task_struct(struct task_struct *dst,
822 struct task_struct *src)
828 void set_task_stack_end_magic(struct task_struct *tsk)
830 unsigned long *stackend;
832 stackend = end_of_stack(tsk);
833 *stackend = STACK_END_MAGIC; /* for overflow detection */
836 static struct task_struct *dup_task_struct(struct task_struct *orig, int node)
838 struct task_struct *tsk;
839 unsigned long *stack;
840 struct vm_struct *stack_vm_area __maybe_unused;
843 if (node == NUMA_NO_NODE)
844 node = tsk_fork_get_node(orig);
845 tsk = alloc_task_struct_node(node);
849 stack = alloc_thread_stack_node(tsk, node);
853 if (memcg_charge_kernel_stack(tsk))
856 stack_vm_area = task_stack_vm_area(tsk);
858 err = arch_dup_task_struct(tsk, orig);
861 * arch_dup_task_struct() clobbers the stack-related fields. Make
862 * sure they're properly initialized before using any stack-related
866 #ifdef CONFIG_VMAP_STACK
867 tsk->stack_vm_area = stack_vm_area;
869 #ifdef CONFIG_THREAD_INFO_IN_TASK
870 atomic_set(&tsk->stack_refcount, 1);
876 #ifdef CONFIG_SECCOMP
878 * We must handle setting up seccomp filters once we're under
879 * the sighand lock in case orig has changed between now and
880 * then. Until then, filter must be NULL to avoid messing up
881 * the usage counts on the error path calling free_task.
883 tsk->seccomp.filter = NULL;
886 setup_thread_stack(tsk, orig);
887 clear_user_return_notifier(tsk);
888 clear_tsk_need_resched(tsk);
889 set_task_stack_end_magic(tsk);
891 #ifdef CONFIG_STACKPROTECTOR
892 tsk->stack_canary = get_random_canary();
896 * One for us, one for whoever does the "release_task()" (usually
899 atomic_set(&tsk->usage, 2);
900 #ifdef CONFIG_BLK_DEV_IO_TRACE
903 tsk->splice_pipe = NULL;
904 tsk->task_frag.page = NULL;
905 tsk->wake_q.next = NULL;
907 account_kernel_stack(tsk, 1);
911 #ifdef CONFIG_FAULT_INJECTION
915 #ifdef CONFIG_BLK_CGROUP
916 tsk->throttle_queue = NULL;
917 tsk->use_memdelay = 0;
921 tsk->active_memcg = NULL;
922 tsk->memcg_high_reclaim = NULL;
927 free_thread_stack(tsk);
929 free_task_struct(tsk);
933 __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
935 static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
937 static int __init coredump_filter_setup(char *s)
939 default_dump_filter =
940 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
941 MMF_DUMP_FILTER_MASK;
945 __setup("coredump_filter=", coredump_filter_setup);
947 #include <linux/init_task.h>
949 static void mm_init_aio(struct mm_struct *mm)
952 spin_lock_init(&mm->ioctx_lock);
953 mm->ioctx_table = NULL;
957 static void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
964 static void mm_init_uprobes_state(struct mm_struct *mm)
966 #ifdef CONFIG_UPROBES
967 mm->uprobes_state.xol_area = NULL;
971 static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p,
972 struct user_namespace *user_ns)
976 mm->vmacache_seqnum = 0;
977 atomic_set(&mm->mm_users, 1);
978 atomic_set(&mm->mm_count, 1);
979 init_rwsem(&mm->mmap_sem);
980 INIT_LIST_HEAD(&mm->mmlist);
981 mm->core_state = NULL;
982 mm_pgtables_bytes_init(mm);
985 atomic64_set(&mm->pinned_vm, 0);
986 memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
987 spin_lock_init(&mm->page_table_lock);
988 spin_lock_init(&mm->arg_lock);
991 mm_init_owner(mm, p);
992 RCU_INIT_POINTER(mm->exe_file, NULL);
993 mmu_notifier_mm_init(mm);
995 init_tlb_flush_pending(mm);
996 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
997 mm->pmd_huge_pte = NULL;
999 mm_init_uprobes_state(mm);
1002 mm->flags = current->mm->flags & MMF_INIT_MASK;
1003 mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK;
1005 mm->flags = default_dump_filter;
1009 if (mm_alloc_pgd(mm))
1012 if (init_new_context(p, mm))
1013 goto fail_nocontext;
1015 mm->user_ns = get_user_ns(user_ns);
1026 * Allocate and initialize an mm_struct.
1028 struct mm_struct *mm_alloc(void)
1030 struct mm_struct *mm;
1036 memset(mm, 0, sizeof(*mm));
1037 return mm_init(mm, current, current_user_ns());
1040 static inline void __mmput(struct mm_struct *mm)
1042 VM_BUG_ON(atomic_read(&mm->mm_users));
1044 uprobe_clear_state(mm);
1047 khugepaged_exit(mm); /* must run before exit_mmap */
1049 mm_put_huge_zero_page(mm);
1050 set_mm_exe_file(mm, NULL);
1051 if (!list_empty(&mm->mmlist)) {
1052 spin_lock(&mmlist_lock);
1053 list_del(&mm->mmlist);
1054 spin_unlock(&mmlist_lock);
1057 module_put(mm->binfmt->module);
1062 * Decrement the use count and release all resources for an mm.
1064 void mmput(struct mm_struct *mm)
1068 if (atomic_dec_and_test(&mm->mm_users))
1071 EXPORT_SYMBOL_GPL(mmput);
1074 static void mmput_async_fn(struct work_struct *work)
1076 struct mm_struct *mm = container_of(work, struct mm_struct,
1082 void mmput_async(struct mm_struct *mm)
1084 if (atomic_dec_and_test(&mm->mm_users)) {
1085 INIT_WORK(&mm->async_put_work, mmput_async_fn);
1086 schedule_work(&mm->async_put_work);
1092 * set_mm_exe_file - change a reference to the mm's executable file
1094 * This changes mm's executable file (shown as symlink /proc/[pid]/exe).
1096 * Main users are mmput() and sys_execve(). Callers prevent concurrent
1097 * invocations: in mmput() nobody alive left, in execve task is single
1098 * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the
1099 * mm->exe_file, but does so without using set_mm_exe_file() in order
1100 * to do avoid the need for any locks.
1102 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1104 struct file *old_exe_file;
1107 * It is safe to dereference the exe_file without RCU as
1108 * this function is only called if nobody else can access
1109 * this mm -- see comment above for justification.
1111 old_exe_file = rcu_dereference_raw(mm->exe_file);
1114 get_file(new_exe_file);
1115 rcu_assign_pointer(mm->exe_file, new_exe_file);
1121 * get_mm_exe_file - acquire a reference to the mm's executable file
1123 * Returns %NULL if mm has no associated executable file.
1124 * User must release file via fput().
1126 struct file *get_mm_exe_file(struct mm_struct *mm)
1128 struct file *exe_file;
1131 exe_file = rcu_dereference(mm->exe_file);
1132 if (exe_file && !get_file_rcu(exe_file))
1137 EXPORT_SYMBOL(get_mm_exe_file);
1140 * get_task_exe_file - acquire a reference to the task's executable file
1142 * Returns %NULL if task's mm (if any) has no associated executable file or
1143 * this is a kernel thread with borrowed mm (see the comment above get_task_mm).
1144 * User must release file via fput().
1146 struct file *get_task_exe_file(struct task_struct *task)
1148 struct file *exe_file = NULL;
1149 struct mm_struct *mm;
1154 if (!(task->flags & PF_KTHREAD))
1155 exe_file = get_mm_exe_file(mm);
1160 EXPORT_SYMBOL(get_task_exe_file);
1163 * get_task_mm - acquire a reference to the task's mm
1165 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
1166 * this kernel workthread has transiently adopted a user mm with use_mm,
1167 * to do its AIO) is not set and if so returns a reference to it, after
1168 * bumping up the use count. User must release the mm via mmput()
1169 * after use. Typically used by /proc and ptrace.
1171 struct mm_struct *get_task_mm(struct task_struct *task)
1173 struct mm_struct *mm;
1178 if (task->flags & PF_KTHREAD)
1186 EXPORT_SYMBOL_GPL(get_task_mm);
1188 struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
1190 struct mm_struct *mm;
1193 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
1195 return ERR_PTR(err);
1197 mm = get_task_mm(task);
1198 if (mm && mm != current->mm &&
1199 !ptrace_may_access(task, mode)) {
1201 mm = ERR_PTR(-EACCES);
1203 mutex_unlock(&task->signal->cred_guard_mutex);
1208 static void complete_vfork_done(struct task_struct *tsk)
1210 struct completion *vfork;
1213 vfork = tsk->vfork_done;
1214 if (likely(vfork)) {
1215 tsk->vfork_done = NULL;
1221 static int wait_for_vfork_done(struct task_struct *child,
1222 struct completion *vfork)
1226 freezer_do_not_count();
1227 killed = wait_for_completion_killable(vfork);
1232 child->vfork_done = NULL;
1236 put_task_struct(child);
1240 /* Please note the differences between mmput and mm_release.
1241 * mmput is called whenever we stop holding onto a mm_struct,
1242 * error success whatever.
1244 * mm_release is called after a mm_struct has been removed
1245 * from the current process.
1247 * This difference is important for error handling, when we
1248 * only half set up a mm_struct for a new process and need to restore
1249 * the old one. Because we mmput the new mm_struct before
1250 * restoring the old one. . .
1251 * Eric Biederman 10 January 1998
1253 void mm_release(struct task_struct *tsk, struct mm_struct *mm)
1255 /* Get rid of any futexes when releasing the mm */
1257 if (unlikely(tsk->robust_list)) {
1258 exit_robust_list(tsk);
1259 tsk->robust_list = NULL;
1261 #ifdef CONFIG_COMPAT
1262 if (unlikely(tsk->compat_robust_list)) {
1263 compat_exit_robust_list(tsk);
1264 tsk->compat_robust_list = NULL;
1267 if (unlikely(!list_empty(&tsk->pi_state_list)))
1268 exit_pi_state_list(tsk);
1271 uprobe_free_utask(tsk);
1273 /* Get rid of any cached register state */
1274 deactivate_mm(tsk, mm);
1277 * Signal userspace if we're not exiting with a core dump
1278 * because we want to leave the value intact for debugging
1281 if (tsk->clear_child_tid) {
1282 if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) &&
1283 atomic_read(&mm->mm_users) > 1) {
1285 * We don't check the error code - if userspace has
1286 * not set up a proper pointer then tough luck.
1288 put_user(0, tsk->clear_child_tid);
1289 do_futex(tsk->clear_child_tid, FUTEX_WAKE,
1290 1, NULL, NULL, 0, 0);
1292 tsk->clear_child_tid = NULL;
1296 * All done, finally we can wake up parent and return this mm to him.
1297 * Also kthread_stop() uses this completion for synchronization.
1299 if (tsk->vfork_done)
1300 complete_vfork_done(tsk);
1304 * Allocate a new mm structure and copy contents from the
1305 * mm structure of the passed in task structure.
1307 static struct mm_struct *dup_mm(struct task_struct *tsk)
1309 struct mm_struct *mm, *oldmm = current->mm;
1316 memcpy(mm, oldmm, sizeof(*mm));
1318 if (!mm_init(mm, tsk, mm->user_ns))
1321 err = dup_mmap(mm, oldmm);
1325 mm->hiwater_rss = get_mm_rss(mm);
1326 mm->hiwater_vm = mm->total_vm;
1328 if (mm->binfmt && !try_module_get(mm->binfmt->module))
1334 /* don't put binfmt in mmput, we haven't got module yet */
1342 static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
1344 struct mm_struct *mm, *oldmm;
1347 tsk->min_flt = tsk->maj_flt = 0;
1348 tsk->nvcsw = tsk->nivcsw = 0;
1349 #ifdef CONFIG_DETECT_HUNG_TASK
1350 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
1351 tsk->last_switch_time = 0;
1355 tsk->active_mm = NULL;
1358 * Are we cloning a kernel thread?
1360 * We need to steal a active VM for that..
1362 oldmm = current->mm;
1366 /* initialize the new vmacache entries */
1367 vmacache_flush(tsk);
1369 if (clone_flags & CLONE_VM) {
1382 tsk->active_mm = mm;
1389 static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
1391 struct fs_struct *fs = current->fs;
1392 if (clone_flags & CLONE_FS) {
1393 /* tsk->fs is already what we want */
1394 spin_lock(&fs->lock);
1396 spin_unlock(&fs->lock);
1400 spin_unlock(&fs->lock);
1403 tsk->fs = copy_fs_struct(fs);
1409 static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
1411 struct files_struct *oldf, *newf;
1415 * A background process may not have any files ...
1417 oldf = current->files;
1421 if (clone_flags & CLONE_FILES) {
1422 atomic_inc(&oldf->count);
1426 newf = dup_fd(oldf, &error);
1436 static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
1439 struct io_context *ioc = current->io_context;
1440 struct io_context *new_ioc;
1445 * Share io context with parent, if CLONE_IO is set
1447 if (clone_flags & CLONE_IO) {
1449 tsk->io_context = ioc;
1450 } else if (ioprio_valid(ioc->ioprio)) {
1451 new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE);
1452 if (unlikely(!new_ioc))
1455 new_ioc->ioprio = ioc->ioprio;
1456 put_io_context(new_ioc);
1462 static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
1464 struct sighand_struct *sig;
1466 if (clone_flags & CLONE_SIGHAND) {
1467 atomic_inc(¤t->sighand->count);
1470 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
1471 rcu_assign_pointer(tsk->sighand, sig);
1475 atomic_set(&sig->count, 1);
1476 spin_lock_irq(¤t->sighand->siglock);
1477 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
1478 spin_unlock_irq(¤t->sighand->siglock);
1482 void __cleanup_sighand(struct sighand_struct *sighand)
1484 if (atomic_dec_and_test(&sighand->count)) {
1485 signalfd_cleanup(sighand);
1487 * sighand_cachep is SLAB_TYPESAFE_BY_RCU so we can free it
1488 * without an RCU grace period, see __lock_task_sighand().
1490 kmem_cache_free(sighand_cachep, sighand);
1494 #ifdef CONFIG_POSIX_TIMERS
1496 * Initialize POSIX timer handling for a thread group.
1498 static void posix_cpu_timers_init_group(struct signal_struct *sig)
1500 unsigned long cpu_limit;
1502 cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
1503 if (cpu_limit != RLIM_INFINITY) {
1504 sig->cputime_expires.prof_exp = cpu_limit * NSEC_PER_SEC;
1505 sig->cputimer.running = true;
1508 /* The timer lists. */
1509 INIT_LIST_HEAD(&sig->cpu_timers[0]);
1510 INIT_LIST_HEAD(&sig->cpu_timers[1]);
1511 INIT_LIST_HEAD(&sig->cpu_timers[2]);
1514 static inline void posix_cpu_timers_init_group(struct signal_struct *sig) { }
1517 static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
1519 struct signal_struct *sig;
1521 if (clone_flags & CLONE_THREAD)
1524 sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
1529 sig->nr_threads = 1;
1530 atomic_set(&sig->live, 1);
1531 atomic_set(&sig->sigcnt, 1);
1533 /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */
1534 sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node);
1535 tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head);
1537 init_waitqueue_head(&sig->wait_chldexit);
1538 sig->curr_target = tsk;
1539 init_sigpending(&sig->shared_pending);
1540 INIT_HLIST_HEAD(&sig->multiprocess);
1541 seqlock_init(&sig->stats_lock);
1542 prev_cputime_init(&sig->prev_cputime);
1544 #ifdef CONFIG_POSIX_TIMERS
1545 INIT_LIST_HEAD(&sig->posix_timers);
1546 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
1547 sig->real_timer.function = it_real_fn;
1550 task_lock(current->group_leader);
1551 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
1552 task_unlock(current->group_leader);
1554 posix_cpu_timers_init_group(sig);
1556 tty_audit_fork(sig);
1557 sched_autogroup_fork(sig);
1559 sig->oom_score_adj = current->signal->oom_score_adj;
1560 sig->oom_score_adj_min = current->signal->oom_score_adj_min;
1562 mutex_init(&sig->cred_guard_mutex);
1567 static void copy_seccomp(struct task_struct *p)
1569 #ifdef CONFIG_SECCOMP
1571 * Must be called with sighand->lock held, which is common to
1572 * all threads in the group. Holding cred_guard_mutex is not
1573 * needed because this new task is not yet running and cannot
1576 assert_spin_locked(¤t->sighand->siglock);
1578 /* Ref-count the new filter user, and assign it. */
1579 get_seccomp_filter(current);
1580 p->seccomp = current->seccomp;
1583 * Explicitly enable no_new_privs here in case it got set
1584 * between the task_struct being duplicated and holding the
1585 * sighand lock. The seccomp state and nnp must be in sync.
1587 if (task_no_new_privs(current))
1588 task_set_no_new_privs(p);
1591 * If the parent gained a seccomp mode after copying thread
1592 * flags and between before we held the sighand lock, we have
1593 * to manually enable the seccomp thread flag here.
1595 if (p->seccomp.mode != SECCOMP_MODE_DISABLED)
1596 set_tsk_thread_flag(p, TIF_SECCOMP);
1600 SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
1602 current->clear_child_tid = tidptr;
1604 return task_pid_vnr(current);
1607 static void rt_mutex_init_task(struct task_struct *p)
1609 raw_spin_lock_init(&p->pi_lock);
1610 #ifdef CONFIG_RT_MUTEXES
1611 p->pi_waiters = RB_ROOT_CACHED;
1612 p->pi_top_task = NULL;
1613 p->pi_blocked_on = NULL;
1617 #ifdef CONFIG_POSIX_TIMERS
1619 * Initialize POSIX timer handling for a single task.
1621 static void posix_cpu_timers_init(struct task_struct *tsk)
1623 tsk->cputime_expires.prof_exp = 0;
1624 tsk->cputime_expires.virt_exp = 0;
1625 tsk->cputime_expires.sched_exp = 0;
1626 INIT_LIST_HEAD(&tsk->cpu_timers[0]);
1627 INIT_LIST_HEAD(&tsk->cpu_timers[1]);
1628 INIT_LIST_HEAD(&tsk->cpu_timers[2]);
1631 static inline void posix_cpu_timers_init(struct task_struct *tsk) { }
1634 static inline void init_task_pid_links(struct task_struct *task)
1638 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1639 INIT_HLIST_NODE(&task->pid_links[type]);
1644 init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid)
1646 if (type == PIDTYPE_PID)
1647 task->thread_pid = pid;
1649 task->signal->pids[type] = pid;
1652 static inline void rcu_copy_process(struct task_struct *p)
1654 #ifdef CONFIG_PREEMPT_RCU
1655 p->rcu_read_lock_nesting = 0;
1656 p->rcu_read_unlock_special.s = 0;
1657 p->rcu_blocked_node = NULL;
1658 INIT_LIST_HEAD(&p->rcu_node_entry);
1659 #endif /* #ifdef CONFIG_PREEMPT_RCU */
1660 #ifdef CONFIG_TASKS_RCU
1661 p->rcu_tasks_holdout = false;
1662 INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
1663 p->rcu_tasks_idle_cpu = -1;
1664 #endif /* #ifdef CONFIG_TASKS_RCU */
1668 * This creates a new process as a copy of the old one,
1669 * but does not actually start it yet.
1671 * It copies the registers, and all the appropriate
1672 * parts of the process environment (as per the clone
1673 * flags). The actual kick-off is left to the caller.
1675 static __latent_entropy struct task_struct *copy_process(
1676 unsigned long clone_flags,
1677 unsigned long stack_start,
1678 unsigned long stack_size,
1679 int __user *child_tidptr,
1686 struct task_struct *p;
1687 struct multiprocess_signals delayed;
1690 * Don't allow sharing the root directory with processes in a different
1693 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
1694 return ERR_PTR(-EINVAL);
1696 if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS))
1697 return ERR_PTR(-EINVAL);
1700 * Thread groups must share signals as well, and detached threads
1701 * can only be started up within the thread group.
1703 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1704 return ERR_PTR(-EINVAL);
1707 * Shared signal handlers imply shared VM. By way of the above,
1708 * thread groups also imply shared VM. Blocking this case allows
1709 * for various simplifications in other code.
1711 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1712 return ERR_PTR(-EINVAL);
1715 * Siblings of global init remain as zombies on exit since they are
1716 * not reaped by their parent (swapper). To solve this and to avoid
1717 * multi-rooted process trees, prevent global and container-inits
1718 * from creating siblings.
1720 if ((clone_flags & CLONE_PARENT) &&
1721 current->signal->flags & SIGNAL_UNKILLABLE)
1722 return ERR_PTR(-EINVAL);
1725 * If the new process will be in a different pid or user namespace
1726 * do not allow it to share a thread group with the forking task.
1728 if (clone_flags & CLONE_THREAD) {
1729 if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) ||
1730 (task_active_pid_ns(current) !=
1731 current->nsproxy->pid_ns_for_children))
1732 return ERR_PTR(-EINVAL);
1736 * Force any signals received before this point to be delivered
1737 * before the fork happens. Collect up signals sent to multiple
1738 * processes that happen during the fork and delay them so that
1739 * they appear to happen after the fork.
1741 sigemptyset(&delayed.signal);
1742 INIT_HLIST_NODE(&delayed.node);
1744 spin_lock_irq(¤t->sighand->siglock);
1745 if (!(clone_flags & CLONE_THREAD))
1746 hlist_add_head(&delayed.node, ¤t->signal->multiprocess);
1747 recalc_sigpending();
1748 spin_unlock_irq(¤t->sighand->siglock);
1749 retval = -ERESTARTNOINTR;
1750 if (signal_pending(current))
1754 p = dup_task_struct(current, node);
1759 * This _must_ happen before we call free_task(), i.e. before we jump
1760 * to any of the bad_fork_* labels. This is to avoid freeing
1761 * p->set_child_tid which is (ab)used as a kthread's data pointer for
1762 * kernel threads (PF_KTHREAD).
1764 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1766 * Clear TID on mm_release()?
1768 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL;
1770 ftrace_graph_init_task(p);
1772 rt_mutex_init_task(p);
1774 #ifdef CONFIG_PROVE_LOCKING
1775 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1776 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1779 if (atomic_read(&p->real_cred->user->processes) >=
1780 task_rlimit(p, RLIMIT_NPROC)) {
1781 if (p->real_cred->user != INIT_USER &&
1782 !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN))
1785 current->flags &= ~PF_NPROC_EXCEEDED;
1787 retval = copy_creds(p, clone_flags);
1792 * If multiple threads are within copy_process(), then this check
1793 * triggers too late. This doesn't hurt, the check is only there
1794 * to stop root fork bombs.
1797 if (nr_threads >= max_threads)
1798 goto bad_fork_cleanup_count;
1800 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
1801 p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER | PF_IDLE);
1802 p->flags |= PF_FORKNOEXEC;
1803 INIT_LIST_HEAD(&p->children);
1804 INIT_LIST_HEAD(&p->sibling);
1805 rcu_copy_process(p);
1806 p->vfork_done = NULL;
1807 spin_lock_init(&p->alloc_lock);
1809 init_sigpending(&p->pending);
1811 p->utime = p->stime = p->gtime = 0;
1812 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1813 p->utimescaled = p->stimescaled = 0;
1815 prev_cputime_init(&p->prev_cputime);
1817 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1818 seqcount_init(&p->vtime.seqcount);
1819 p->vtime.starttime = 0;
1820 p->vtime.state = VTIME_INACTIVE;
1823 #if defined(SPLIT_RSS_COUNTING)
1824 memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1827 p->default_timer_slack_ns = current->timer_slack_ns;
1833 task_io_accounting_init(&p->ioac);
1834 acct_clear_integrals(p);
1836 posix_cpu_timers_init(p);
1838 p->io_context = NULL;
1839 audit_set_context(p, NULL);
1842 p->mempolicy = mpol_dup(p->mempolicy);
1843 if (IS_ERR(p->mempolicy)) {
1844 retval = PTR_ERR(p->mempolicy);
1845 p->mempolicy = NULL;
1846 goto bad_fork_cleanup_threadgroup_lock;
1849 #ifdef CONFIG_CPUSETS
1850 p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
1851 p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
1852 seqcount_init(&p->mems_allowed_seq);
1854 #ifdef CONFIG_TRACE_IRQFLAGS
1856 p->hardirqs_enabled = 0;
1857 p->hardirq_enable_ip = 0;
1858 p->hardirq_enable_event = 0;
1859 p->hardirq_disable_ip = _THIS_IP_;
1860 p->hardirq_disable_event = 0;
1861 p->softirqs_enabled = 1;
1862 p->softirq_enable_ip = _THIS_IP_;
1863 p->softirq_enable_event = 0;
1864 p->softirq_disable_ip = 0;
1865 p->softirq_disable_event = 0;
1866 p->hardirq_context = 0;
1867 p->softirq_context = 0;
1870 p->pagefault_disabled = 0;
1872 #ifdef CONFIG_LOCKDEP
1873 p->lockdep_depth = 0; /* no locks held yet */
1874 p->curr_chain_key = 0;
1875 p->lockdep_recursion = 0;
1876 lockdep_init_task(p);
1879 #ifdef CONFIG_DEBUG_MUTEXES
1880 p->blocked_on = NULL; /* not blocked yet */
1882 #ifdef CONFIG_BCACHE
1883 p->sequential_io = 0;
1884 p->sequential_io_avg = 0;
1887 /* Perform scheduler related setup. Assign this task to a CPU. */
1888 retval = sched_fork(clone_flags, p);
1890 goto bad_fork_cleanup_policy;
1892 retval = perf_event_init_task(p);
1894 goto bad_fork_cleanup_policy;
1895 retval = audit_alloc(p);
1897 goto bad_fork_cleanup_perf;
1898 /* copy all the process information */
1900 retval = security_task_alloc(p, clone_flags);
1902 goto bad_fork_cleanup_audit;
1903 retval = copy_semundo(clone_flags, p);
1905 goto bad_fork_cleanup_security;
1906 retval = copy_files(clone_flags, p);
1908 goto bad_fork_cleanup_semundo;
1909 retval = copy_fs(clone_flags, p);
1911 goto bad_fork_cleanup_files;
1912 retval = copy_sighand(clone_flags, p);
1914 goto bad_fork_cleanup_fs;
1915 retval = copy_signal(clone_flags, p);
1917 goto bad_fork_cleanup_sighand;
1918 retval = copy_mm(clone_flags, p);
1920 goto bad_fork_cleanup_signal;
1921 retval = copy_namespaces(clone_flags, p);
1923 goto bad_fork_cleanup_mm;
1924 retval = copy_io(clone_flags, p);
1926 goto bad_fork_cleanup_namespaces;
1927 retval = copy_thread_tls(clone_flags, stack_start, stack_size, p, tls);
1929 goto bad_fork_cleanup_io;
1931 stackleak_task_init(p);
1933 if (pid != &init_struct_pid) {
1934 pid = alloc_pid(p->nsproxy->pid_ns_for_children);
1936 retval = PTR_ERR(pid);
1937 goto bad_fork_cleanup_thread;
1945 p->robust_list = NULL;
1946 #ifdef CONFIG_COMPAT
1947 p->compat_robust_list = NULL;
1949 INIT_LIST_HEAD(&p->pi_state_list);
1950 p->pi_state_cache = NULL;
1953 * sigaltstack should be cleared when sharing the same VM
1955 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1959 * Syscall tracing and stepping should be turned off in the
1960 * child regardless of CLONE_PTRACE.
1962 user_disable_single_step(p);
1963 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
1964 #ifdef TIF_SYSCALL_EMU
1965 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1967 clear_all_latency_tracing(p);
1969 /* ok, now we should be set up.. */
1970 p->pid = pid_nr(pid);
1971 if (clone_flags & CLONE_THREAD) {
1972 p->exit_signal = -1;
1973 p->group_leader = current->group_leader;
1974 p->tgid = current->tgid;
1976 if (clone_flags & CLONE_PARENT)
1977 p->exit_signal = current->group_leader->exit_signal;
1979 p->exit_signal = (clone_flags & CSIGNAL);
1980 p->group_leader = p;
1985 p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
1986 p->dirty_paused_when = 0;
1988 p->pdeath_signal = 0;
1989 INIT_LIST_HEAD(&p->thread_group);
1990 p->task_works = NULL;
1992 cgroup_threadgroup_change_begin(current);
1994 * Ensure that the cgroup subsystem policies allow the new process to be
1995 * forked. It should be noted the the new process's css_set can be changed
1996 * between here and cgroup_post_fork() if an organisation operation is in
1999 retval = cgroup_can_fork(p);
2001 goto bad_fork_free_pid;
2004 * From this point on we must avoid any synchronous user-space
2005 * communication until we take the tasklist-lock. In particular, we do
2006 * not want user-space to be able to predict the process start-time by
2007 * stalling fork(2) after we recorded the start_time but before it is
2008 * visible to the system.
2011 p->start_time = ktime_get_ns();
2012 p->real_start_time = ktime_get_boot_ns();
2015 * Make it visible to the rest of the system, but dont wake it up yet.
2016 * Need tasklist lock for parent etc handling!
2018 write_lock_irq(&tasklist_lock);
2020 /* CLONE_PARENT re-uses the old parent */
2021 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
2022 p->real_parent = current->real_parent;
2023 p->parent_exec_id = current->parent_exec_id;
2025 p->real_parent = current;
2026 p->parent_exec_id = current->self_exec_id;
2029 klp_copy_process(p);
2031 spin_lock(¤t->sighand->siglock);
2034 * Copy seccomp details explicitly here, in case they were changed
2035 * before holding sighand lock.
2039 rseq_fork(p, clone_flags);
2041 /* Don't start children in a dying pid namespace */
2042 if (unlikely(!(ns_of_pid(pid)->pid_allocated & PIDNS_ADDING))) {
2044 goto bad_fork_cancel_cgroup;
2047 /* Let kill terminate clone/fork in the middle */
2048 if (fatal_signal_pending(current)) {
2050 goto bad_fork_cancel_cgroup;
2054 init_task_pid_links(p);
2055 if (likely(p->pid)) {
2056 ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
2058 init_task_pid(p, PIDTYPE_PID, pid);
2059 if (thread_group_leader(p)) {
2060 init_task_pid(p, PIDTYPE_TGID, pid);
2061 init_task_pid(p, PIDTYPE_PGID, task_pgrp(current));
2062 init_task_pid(p, PIDTYPE_SID, task_session(current));
2064 if (is_child_reaper(pid)) {
2065 ns_of_pid(pid)->child_reaper = p;
2066 p->signal->flags |= SIGNAL_UNKILLABLE;
2068 p->signal->shared_pending.signal = delayed.signal;
2069 p->signal->tty = tty_kref_get(current->signal->tty);
2071 * Inherit has_child_subreaper flag under the same
2072 * tasklist_lock with adding child to the process tree
2073 * for propagate_has_child_subreaper optimization.
2075 p->signal->has_child_subreaper = p->real_parent->signal->has_child_subreaper ||
2076 p->real_parent->signal->is_child_subreaper;
2077 list_add_tail(&p->sibling, &p->real_parent->children);
2078 list_add_tail_rcu(&p->tasks, &init_task.tasks);
2079 attach_pid(p, PIDTYPE_TGID);
2080 attach_pid(p, PIDTYPE_PGID);
2081 attach_pid(p, PIDTYPE_SID);
2082 __this_cpu_inc(process_counts);
2084 current->signal->nr_threads++;
2085 atomic_inc(¤t->signal->live);
2086 atomic_inc(¤t->signal->sigcnt);
2087 task_join_group_stop(p);
2088 list_add_tail_rcu(&p->thread_group,
2089 &p->group_leader->thread_group);
2090 list_add_tail_rcu(&p->thread_node,
2091 &p->signal->thread_head);
2093 attach_pid(p, PIDTYPE_PID);
2097 hlist_del_init(&delayed.node);
2098 spin_unlock(¤t->sighand->siglock);
2099 syscall_tracepoint_update(p);
2100 write_unlock_irq(&tasklist_lock);
2102 proc_fork_connector(p);
2103 cgroup_post_fork(p);
2104 cgroup_threadgroup_change_end(current);
2107 trace_task_newtask(p, clone_flags);
2108 uprobe_copy_process(p, clone_flags);
2112 bad_fork_cancel_cgroup:
2113 spin_unlock(¤t->sighand->siglock);
2114 write_unlock_irq(&tasklist_lock);
2115 cgroup_cancel_fork(p);
2117 cgroup_threadgroup_change_end(current);
2118 if (pid != &init_struct_pid)
2120 bad_fork_cleanup_thread:
2122 bad_fork_cleanup_io:
2125 bad_fork_cleanup_namespaces:
2126 exit_task_namespaces(p);
2127 bad_fork_cleanup_mm:
2130 bad_fork_cleanup_signal:
2131 if (!(clone_flags & CLONE_THREAD))
2132 free_signal_struct(p->signal);
2133 bad_fork_cleanup_sighand:
2134 __cleanup_sighand(p->sighand);
2135 bad_fork_cleanup_fs:
2136 exit_fs(p); /* blocking */
2137 bad_fork_cleanup_files:
2138 exit_files(p); /* blocking */
2139 bad_fork_cleanup_semundo:
2141 bad_fork_cleanup_security:
2142 security_task_free(p);
2143 bad_fork_cleanup_audit:
2145 bad_fork_cleanup_perf:
2146 perf_event_free_task(p);
2147 bad_fork_cleanup_policy:
2148 lockdep_free_task(p);
2150 mpol_put(p->mempolicy);
2151 bad_fork_cleanup_threadgroup_lock:
2153 delayacct_tsk_free(p);
2154 bad_fork_cleanup_count:
2155 atomic_dec(&p->cred->user->processes);
2158 p->state = TASK_DEAD;
2162 spin_lock_irq(¤t->sighand->siglock);
2163 hlist_del_init(&delayed.node);
2164 spin_unlock_irq(¤t->sighand->siglock);
2165 return ERR_PTR(retval);
2168 static inline void init_idle_pids(struct task_struct *idle)
2172 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
2173 INIT_HLIST_NODE(&idle->pid_links[type]); /* not really needed */
2174 init_task_pid(idle, type, &init_struct_pid);
2178 struct task_struct *fork_idle(int cpu)
2180 struct task_struct *task;
2181 task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0, 0,
2183 if (!IS_ERR(task)) {
2184 init_idle_pids(task);
2185 init_idle(task, cpu);
2192 * Ok, this is the main fork-routine.
2194 * It copies the process, and if successful kick-starts
2195 * it and waits for it to finish using the VM if required.
2197 long _do_fork(unsigned long clone_flags,
2198 unsigned long stack_start,
2199 unsigned long stack_size,
2200 int __user *parent_tidptr,
2201 int __user *child_tidptr,
2204 struct completion vfork;
2206 struct task_struct *p;
2211 * Determine whether and which event to report to ptracer. When
2212 * called from kernel_thread or CLONE_UNTRACED is explicitly
2213 * requested, no event is reported; otherwise, report if the event
2214 * for the type of forking is enabled.
2216 if (!(clone_flags & CLONE_UNTRACED)) {
2217 if (clone_flags & CLONE_VFORK)
2218 trace = PTRACE_EVENT_VFORK;
2219 else if ((clone_flags & CSIGNAL) != SIGCHLD)
2220 trace = PTRACE_EVENT_CLONE;
2222 trace = PTRACE_EVENT_FORK;
2224 if (likely(!ptrace_event_enabled(current, trace)))
2228 p = copy_process(clone_flags, stack_start, stack_size,
2229 child_tidptr, NULL, trace, tls, NUMA_NO_NODE);
2230 add_latent_entropy();
2236 * Do this prior waking up the new thread - the thread pointer
2237 * might get invalid after that point, if the thread exits quickly.
2239 trace_sched_process_fork(current, p);
2241 pid = get_task_pid(p, PIDTYPE_PID);
2244 if (clone_flags & CLONE_PARENT_SETTID)
2245 put_user(nr, parent_tidptr);
2247 if (clone_flags & CLONE_VFORK) {
2248 p->vfork_done = &vfork;
2249 init_completion(&vfork);
2253 wake_up_new_task(p);
2255 /* forking complete and child started to run, tell ptracer */
2256 if (unlikely(trace))
2257 ptrace_event_pid(trace, pid);
2259 if (clone_flags & CLONE_VFORK) {
2260 if (!wait_for_vfork_done(p, &vfork))
2261 ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid);
2268 #ifndef CONFIG_HAVE_COPY_THREAD_TLS
2269 /* For compatibility with architectures that call do_fork directly rather than
2270 * using the syscall entry points below. */
2271 long do_fork(unsigned long clone_flags,
2272 unsigned long stack_start,
2273 unsigned long stack_size,
2274 int __user *parent_tidptr,
2275 int __user *child_tidptr)
2277 return _do_fork(clone_flags, stack_start, stack_size,
2278 parent_tidptr, child_tidptr, 0);
2283 * Create a kernel thread.
2285 pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
2287 return _do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn,
2288 (unsigned long)arg, NULL, NULL, 0);
2291 #ifdef __ARCH_WANT_SYS_FORK
2292 SYSCALL_DEFINE0(fork)
2295 return _do_fork(SIGCHLD, 0, 0, NULL, NULL, 0);
2297 /* can not support in nommu mode */
2303 #ifdef __ARCH_WANT_SYS_VFORK
2304 SYSCALL_DEFINE0(vfork)
2306 return _do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0,
2311 #ifdef __ARCH_WANT_SYS_CLONE
2312 #ifdef CONFIG_CLONE_BACKWARDS
2313 SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
2314 int __user *, parent_tidptr,
2316 int __user *, child_tidptr)
2317 #elif defined(CONFIG_CLONE_BACKWARDS2)
2318 SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags,
2319 int __user *, parent_tidptr,
2320 int __user *, child_tidptr,
2322 #elif defined(CONFIG_CLONE_BACKWARDS3)
2323 SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp,
2325 int __user *, parent_tidptr,
2326 int __user *, child_tidptr,
2329 SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
2330 int __user *, parent_tidptr,
2331 int __user *, child_tidptr,
2335 return _do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr, tls);
2339 void walk_process_tree(struct task_struct *top, proc_visitor visitor, void *data)
2341 struct task_struct *leader, *parent, *child;
2344 read_lock(&tasklist_lock);
2345 leader = top = top->group_leader;
2347 for_each_thread(leader, parent) {
2348 list_for_each_entry(child, &parent->children, sibling) {
2349 res = visitor(child, data);
2361 if (leader != top) {
2363 parent = child->real_parent;
2364 leader = parent->group_leader;
2368 read_unlock(&tasklist_lock);
2371 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
2372 #define ARCH_MIN_MMSTRUCT_ALIGN 0
2375 static void sighand_ctor(void *data)
2377 struct sighand_struct *sighand = data;
2379 spin_lock_init(&sighand->siglock);
2380 init_waitqueue_head(&sighand->signalfd_wqh);
2383 void __init proc_caches_init(void)
2385 unsigned int mm_size;
2387 sighand_cachep = kmem_cache_create("sighand_cache",
2388 sizeof(struct sighand_struct), 0,
2389 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_TYPESAFE_BY_RCU|
2390 SLAB_ACCOUNT, sighand_ctor);
2391 signal_cachep = kmem_cache_create("signal_cache",
2392 sizeof(struct signal_struct), 0,
2393 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT,
2395 files_cachep = kmem_cache_create("files_cache",
2396 sizeof(struct files_struct), 0,
2397 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT,
2399 fs_cachep = kmem_cache_create("fs_cache",
2400 sizeof(struct fs_struct), 0,
2401 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT,
2405 * The mm_cpumask is located at the end of mm_struct, and is
2406 * dynamically sized based on the maximum CPU number this system
2407 * can have, taking hotplug into account (nr_cpu_ids).
2409 mm_size = sizeof(struct mm_struct) + cpumask_size();
2411 mm_cachep = kmem_cache_create_usercopy("mm_struct",
2412 mm_size, ARCH_MIN_MMSTRUCT_ALIGN,
2413 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT,
2414 offsetof(struct mm_struct, saved_auxv),
2415 sizeof_field(struct mm_struct, saved_auxv),
2417 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT);
2419 nsproxy_cache_init();
2423 * Check constraints on flags passed to the unshare system call.
2425 static int check_unshare_flags(unsigned long unshare_flags)
2427 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
2428 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
2429 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET|
2430 CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP))
2433 * Not implemented, but pretend it works if there is nothing
2434 * to unshare. Note that unsharing the address space or the
2435 * signal handlers also need to unshare the signal queues (aka
2438 if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
2439 if (!thread_group_empty(current))
2442 if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) {
2443 if (atomic_read(¤t->sighand->count) > 1)
2446 if (unshare_flags & CLONE_VM) {
2447 if (!current_is_single_threaded())
2455 * Unshare the filesystem structure if it is being shared
2457 static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
2459 struct fs_struct *fs = current->fs;
2461 if (!(unshare_flags & CLONE_FS) || !fs)
2464 /* don't need lock here; in the worst case we'll do useless copy */
2468 *new_fsp = copy_fs_struct(fs);
2476 * Unshare file descriptor table if it is being shared
2478 static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
2480 struct files_struct *fd = current->files;
2483 if ((unshare_flags & CLONE_FILES) &&
2484 (fd && atomic_read(&fd->count) > 1)) {
2485 *new_fdp = dup_fd(fd, &error);
2494 * unshare allows a process to 'unshare' part of the process
2495 * context which was originally shared using clone. copy_*
2496 * functions used by do_fork() cannot be used here directly
2497 * because they modify an inactive task_struct that is being
2498 * constructed. Here we are modifying the current, active,
2501 int ksys_unshare(unsigned long unshare_flags)
2503 struct fs_struct *fs, *new_fs = NULL;
2504 struct files_struct *fd, *new_fd = NULL;
2505 struct cred *new_cred = NULL;
2506 struct nsproxy *new_nsproxy = NULL;
2511 * If unsharing a user namespace must also unshare the thread group
2512 * and unshare the filesystem root and working directories.
2514 if (unshare_flags & CLONE_NEWUSER)
2515 unshare_flags |= CLONE_THREAD | CLONE_FS;
2517 * If unsharing vm, must also unshare signal handlers.
2519 if (unshare_flags & CLONE_VM)
2520 unshare_flags |= CLONE_SIGHAND;
2522 * If unsharing a signal handlers, must also unshare the signal queues.
2524 if (unshare_flags & CLONE_SIGHAND)
2525 unshare_flags |= CLONE_THREAD;
2527 * If unsharing namespace, must also unshare filesystem information.
2529 if (unshare_flags & CLONE_NEWNS)
2530 unshare_flags |= CLONE_FS;
2532 err = check_unshare_flags(unshare_flags);
2534 goto bad_unshare_out;
2536 * CLONE_NEWIPC must also detach from the undolist: after switching
2537 * to a new ipc namespace, the semaphore arrays from the old
2538 * namespace are unreachable.
2540 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
2542 err = unshare_fs(unshare_flags, &new_fs);
2544 goto bad_unshare_out;
2545 err = unshare_fd(unshare_flags, &new_fd);
2547 goto bad_unshare_cleanup_fs;
2548 err = unshare_userns(unshare_flags, &new_cred);
2550 goto bad_unshare_cleanup_fd;
2551 err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
2554 goto bad_unshare_cleanup_cred;
2556 if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) {
2559 * CLONE_SYSVSEM is equivalent to sys_exit().
2563 if (unshare_flags & CLONE_NEWIPC) {
2564 /* Orphan segments in old ns (see sem above). */
2566 shm_init_task(current);
2570 switch_task_namespaces(current, new_nsproxy);
2576 spin_lock(&fs->lock);
2577 current->fs = new_fs;
2582 spin_unlock(&fs->lock);
2586 fd = current->files;
2587 current->files = new_fd;
2591 task_unlock(current);
2594 /* Install the new user namespace */
2595 commit_creds(new_cred);
2600 perf_event_namespaces(current);
2602 bad_unshare_cleanup_cred:
2605 bad_unshare_cleanup_fd:
2607 put_files_struct(new_fd);
2609 bad_unshare_cleanup_fs:
2611 free_fs_struct(new_fs);
2617 SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
2619 return ksys_unshare(unshare_flags);
2623 * Helper to unshare the files of the current task.
2624 * We don't want to expose copy_files internals to
2625 * the exec layer of the kernel.
2628 int unshare_files(struct files_struct **displaced)
2630 struct task_struct *task = current;
2631 struct files_struct *copy = NULL;
2634 error = unshare_fd(CLONE_FILES, ©);
2635 if (error || !copy) {
2639 *displaced = task->files;
2646 int sysctl_max_threads(struct ctl_table *table, int write,
2647 void __user *buffer, size_t *lenp, loff_t *ppos)
2651 int threads = max_threads;
2652 int min = MIN_THREADS;
2653 int max = MAX_THREADS;
2660 ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos);
2664 set_max_threads(threads);