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/perf_event.h>
81 #include <linux/posix-timers.h>
82 #include <linux/user-return-notifier.h>
83 #include <linux/oom.h>
84 #include <linux/khugepaged.h>
85 #include <linux/signalfd.h>
86 #include <linux/uprobes.h>
87 #include <linux/aio.h>
88 #include <linux/compiler.h>
89 #include <linux/sysctl.h>
90 #include <linux/kcov.h>
91 #include <linux/livepatch.h>
92 #include <linux/thread_info.h>
94 #include <asm/pgtable.h>
95 #include <asm/pgalloc.h>
96 #include <linux/uaccess.h>
97 #include <asm/mmu_context.h>
98 #include <asm/cacheflush.h>
99 #include <asm/tlbflush.h>
101 #include <trace/events/sched.h>
103 #define CREATE_TRACE_POINTS
104 #include <trace/events/task.h>
105 #ifdef CONFIG_USER_NS
106 extern int unprivileged_userns_clone
;
108 #define unprivileged_userns_clone 0
112 * Minimum number of threads to boot the kernel
114 #define MIN_THREADS 20
117 * Maximum number of threads
119 #define MAX_THREADS FUTEX_TID_MASK
122 * Protected counters by write_lock_irq(&tasklist_lock)
124 unsigned long total_forks
; /* Handle normal Linux uptimes. */
125 int nr_threads
; /* The idle threads do not count.. */
127 int max_threads
; /* tunable limit on nr_threads */
129 DEFINE_PER_CPU(unsigned long, process_counts
) = 0;
131 __cacheline_aligned
DEFINE_RWLOCK(tasklist_lock
); /* outer */
133 #ifdef CONFIG_PROVE_RCU
134 int lockdep_tasklist_lock_is_held(void)
136 return lockdep_is_held(&tasklist_lock
);
138 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held
);
139 #endif /* #ifdef CONFIG_PROVE_RCU */
141 int nr_processes(void)
146 for_each_possible_cpu(cpu
)
147 total
+= per_cpu(process_counts
, cpu
);
152 void __weak
arch_release_task_struct(struct task_struct
*tsk
)
156 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
157 static struct kmem_cache
*task_struct_cachep
;
159 static inline struct task_struct
*alloc_task_struct_node(int node
)
161 return kmem_cache_alloc_node(task_struct_cachep
, GFP_KERNEL
, node
);
164 static inline void free_task_struct(struct task_struct
*tsk
)
166 kmem_cache_free(task_struct_cachep
, tsk
);
170 void __weak
arch_release_thread_stack(unsigned long *stack
)
174 #ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR
177 * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a
178 * kmemcache based allocator.
180 # if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK)
182 #ifdef CONFIG_VMAP_STACK
184 * vmalloc() is a bit slow, and calling vfree() enough times will force a TLB
185 * flush. Try to minimize the number of calls by caching stacks.
187 #define NR_CACHED_STACKS 2
188 static DEFINE_PER_CPU(struct vm_struct
*, cached_stacks
[NR_CACHED_STACKS
]);
190 static int free_vm_stack_cache(unsigned int cpu
)
192 struct vm_struct
**cached_vm_stacks
= per_cpu_ptr(cached_stacks
, cpu
);
195 for (i
= 0; i
< NR_CACHED_STACKS
; i
++) {
196 struct vm_struct
*vm_stack
= cached_vm_stacks
[i
];
201 vfree(vm_stack
->addr
);
202 cached_vm_stacks
[i
] = NULL
;
209 static unsigned long *alloc_thread_stack_node(struct task_struct
*tsk
, int node
)
211 #ifdef CONFIG_VMAP_STACK
215 for (i
= 0; i
< NR_CACHED_STACKS
; i
++) {
218 s
= this_cpu_xchg(cached_stacks
[i
], NULL
);
223 /* Clear stale pointers from reused stack. */
224 memset(s
->addr
, 0, THREAD_SIZE
);
226 tsk
->stack_vm_area
= s
;
230 stack
= __vmalloc_node_range(THREAD_SIZE
, THREAD_ALIGN
,
231 VMALLOC_START
, VMALLOC_END
,
234 0, node
, __builtin_return_address(0));
237 * We can't call find_vm_area() in interrupt context, and
238 * free_thread_stack() can be called in interrupt context,
239 * so cache the vm_struct.
242 tsk
->stack_vm_area
= find_vm_area(stack
);
245 struct page
*page
= alloc_pages_node(node
, THREADINFO_GFP
,
248 return page
? page_address(page
) : NULL
;
252 static inline void free_thread_stack(struct task_struct
*tsk
)
254 #ifdef CONFIG_VMAP_STACK
255 if (task_stack_vm_area(tsk
)) {
258 for (i
= 0; i
< NR_CACHED_STACKS
; i
++) {
259 if (this_cpu_cmpxchg(cached_stacks
[i
],
260 NULL
, tsk
->stack_vm_area
) != NULL
)
266 vfree_atomic(tsk
->stack
);
271 __free_pages(virt_to_page(tsk
->stack
), THREAD_SIZE_ORDER
);
274 static struct kmem_cache
*thread_stack_cache
;
276 static unsigned long *alloc_thread_stack_node(struct task_struct
*tsk
,
279 return kmem_cache_alloc_node(thread_stack_cache
, THREADINFO_GFP
, node
);
282 static void free_thread_stack(struct task_struct
*tsk
)
284 kmem_cache_free(thread_stack_cache
, tsk
->stack
);
287 void thread_stack_cache_init(void)
289 thread_stack_cache
= kmem_cache_create("thread_stack", THREAD_SIZE
,
290 THREAD_SIZE
, 0, NULL
);
291 BUG_ON(thread_stack_cache
== NULL
);
296 /* SLAB cache for signal_struct structures (tsk->signal) */
297 static struct kmem_cache
*signal_cachep
;
299 /* SLAB cache for sighand_struct structures (tsk->sighand) */
300 struct kmem_cache
*sighand_cachep
;
302 /* SLAB cache for files_struct structures (tsk->files) */
303 struct kmem_cache
*files_cachep
;
305 /* SLAB cache for fs_struct structures (tsk->fs) */
306 struct kmem_cache
*fs_cachep
;
308 /* SLAB cache for vm_area_struct structures */
309 static struct kmem_cache
*vm_area_cachep
;
311 /* SLAB cache for mm_struct structures (tsk->mm) */
312 static struct kmem_cache
*mm_cachep
;
314 struct vm_area_struct
*vm_area_alloc(struct mm_struct
*mm
)
316 struct vm_area_struct
*vma
= kmem_cache_zalloc(vm_area_cachep
, GFP_KERNEL
);
320 INIT_LIST_HEAD(&vma
->anon_vma_chain
);
325 struct vm_area_struct
*vm_area_dup(struct vm_area_struct
*orig
)
327 struct vm_area_struct
*new = kmem_cache_alloc(vm_area_cachep
, GFP_KERNEL
);
331 INIT_LIST_HEAD(&new->anon_vma_chain
);
336 void vm_area_free(struct vm_area_struct
*vma
)
338 kmem_cache_free(vm_area_cachep
, vma
);
341 static void account_kernel_stack(struct task_struct
*tsk
, int account
)
343 void *stack
= task_stack_page(tsk
);
344 struct vm_struct
*vm
= task_stack_vm_area(tsk
);
346 BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK
) && PAGE_SIZE
% 1024 != 0);
351 BUG_ON(vm
->nr_pages
!= THREAD_SIZE
/ PAGE_SIZE
);
353 for (i
= 0; i
< THREAD_SIZE
/ PAGE_SIZE
; i
++) {
354 mod_zone_page_state(page_zone(vm
->pages
[i
]),
356 PAGE_SIZE
/ 1024 * account
);
359 /* All stack pages belong to the same memcg. */
360 mod_memcg_page_state(vm
->pages
[0], MEMCG_KERNEL_STACK_KB
,
361 account
* (THREAD_SIZE
/ 1024));
364 * All stack pages are in the same zone and belong to the
367 struct page
*first_page
= virt_to_page(stack
);
369 mod_zone_page_state(page_zone(first_page
), NR_KERNEL_STACK_KB
,
370 THREAD_SIZE
/ 1024 * account
);
372 mod_memcg_page_state(first_page
, MEMCG_KERNEL_STACK_KB
,
373 account
* (THREAD_SIZE
/ 1024));
377 static void release_task_stack(struct task_struct
*tsk
)
379 if (WARN_ON(tsk
->state
!= TASK_DEAD
))
380 return; /* Better to leak the stack than to free prematurely */
382 account_kernel_stack(tsk
, -1);
383 arch_release_thread_stack(tsk
->stack
);
384 free_thread_stack(tsk
);
386 #ifdef CONFIG_VMAP_STACK
387 tsk
->stack_vm_area
= NULL
;
391 #ifdef CONFIG_THREAD_INFO_IN_TASK
392 void put_task_stack(struct task_struct
*tsk
)
394 if (atomic_dec_and_test(&tsk
->stack_refcount
))
395 release_task_stack(tsk
);
399 void free_task(struct task_struct
*tsk
)
401 #ifndef CONFIG_THREAD_INFO_IN_TASK
403 * The task is finally done with both the stack and thread_info,
406 release_task_stack(tsk
);
409 * If the task had a separate stack allocation, it should be gone
412 WARN_ON_ONCE(atomic_read(&tsk
->stack_refcount
) != 0);
414 rt_mutex_debug_task_free(tsk
);
415 ftrace_graph_exit_task(tsk
);
416 put_seccomp_filter(tsk
);
417 arch_release_task_struct(tsk
);
418 if (tsk
->flags
& PF_KTHREAD
)
419 free_kthread_struct(tsk
);
420 free_task_struct(tsk
);
422 EXPORT_SYMBOL(free_task
);
424 static inline void free_signal_struct(struct signal_struct
*sig
)
426 taskstats_tgid_free(sig
);
427 sched_autogroup_exit(sig
);
429 * __mmdrop is not safe to call from softirq context on x86 due to
430 * pgd_dtor so postpone it to the async context
433 mmdrop_async(sig
->oom_mm
);
434 kmem_cache_free(signal_cachep
, sig
);
437 static inline void put_signal_struct(struct signal_struct
*sig
)
439 if (atomic_dec_and_test(&sig
->sigcnt
))
440 free_signal_struct(sig
);
443 void __put_task_struct(struct task_struct
*tsk
)
445 WARN_ON(!tsk
->exit_state
);
446 WARN_ON(atomic_read(&tsk
->usage
));
447 WARN_ON(tsk
== current
);
451 security_task_free(tsk
);
453 delayacct_tsk_free(tsk
);
454 put_signal_struct(tsk
->signal
);
456 if (!profile_handoff_task(tsk
))
459 EXPORT_SYMBOL_GPL(__put_task_struct
);
461 void __init __weak
arch_task_cache_init(void) { }
466 static void set_max_threads(unsigned int max_threads_suggested
)
471 * The number of threads shall be limited such that the thread
472 * structures may only consume a small part of the available memory.
474 if (fls64(totalram_pages
) + fls64(PAGE_SIZE
) > 64)
475 threads
= MAX_THREADS
;
477 threads
= div64_u64((u64
) totalram_pages
* (u64
) PAGE_SIZE
,
478 (u64
) THREAD_SIZE
* 8UL);
480 if (threads
> max_threads_suggested
)
481 threads
= max_threads_suggested
;
483 max_threads
= clamp_t(u64
, threads
, MIN_THREADS
, MAX_THREADS
);
486 #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
487 /* Initialized by the architecture: */
488 int arch_task_struct_size __read_mostly
;
491 void __init
fork_init(void)
494 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
495 #ifndef ARCH_MIN_TASKALIGN
496 #define ARCH_MIN_TASKALIGN 0
498 int align
= max_t(int, L1_CACHE_BYTES
, ARCH_MIN_TASKALIGN
);
500 /* create a slab on which task_structs can be allocated */
501 task_struct_cachep
= kmem_cache_create("task_struct",
502 arch_task_struct_size
, align
,
503 SLAB_PANIC
|SLAB_ACCOUNT
, NULL
);
506 /* do the arch specific task caches init */
507 arch_task_cache_init();
509 set_max_threads(MAX_THREADS
);
511 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_cur
= max_threads
/2;
512 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_max
= max_threads
/2;
513 init_task
.signal
->rlim
[RLIMIT_SIGPENDING
] =
514 init_task
.signal
->rlim
[RLIMIT_NPROC
];
516 for (i
= 0; i
< UCOUNT_COUNTS
; i
++) {
517 init_user_ns
.ucount_max
[i
] = max_threads
/2;
520 #ifdef CONFIG_VMAP_STACK
521 cpuhp_setup_state(CPUHP_BP_PREPARE_DYN
, "fork:vm_stack_cache",
522 NULL
, free_vm_stack_cache
);
525 lockdep_init_task(&init_task
);
528 int __weak
arch_dup_task_struct(struct task_struct
*dst
,
529 struct task_struct
*src
)
535 void set_task_stack_end_magic(struct task_struct
*tsk
)
537 unsigned long *stackend
;
539 stackend
= end_of_stack(tsk
);
540 *stackend
= STACK_END_MAGIC
; /* for overflow detection */
543 static struct task_struct
*dup_task_struct(struct task_struct
*orig
, int node
)
545 struct task_struct
*tsk
;
546 unsigned long *stack
;
547 struct vm_struct
*stack_vm_area
;
550 if (node
== NUMA_NO_NODE
)
551 node
= tsk_fork_get_node(orig
);
552 tsk
= alloc_task_struct_node(node
);
556 stack
= alloc_thread_stack_node(tsk
, node
);
560 stack_vm_area
= task_stack_vm_area(tsk
);
562 err
= arch_dup_task_struct(tsk
, orig
);
565 * arch_dup_task_struct() clobbers the stack-related fields. Make
566 * sure they're properly initialized before using any stack-related
570 #ifdef CONFIG_VMAP_STACK
571 tsk
->stack_vm_area
= stack_vm_area
;
573 #ifdef CONFIG_THREAD_INFO_IN_TASK
574 atomic_set(&tsk
->stack_refcount
, 1);
580 #ifdef CONFIG_SECCOMP
582 * We must handle setting up seccomp filters once we're under
583 * the sighand lock in case orig has changed between now and
584 * then. Until then, filter must be NULL to avoid messing up
585 * the usage counts on the error path calling free_task.
587 tsk
->seccomp
.filter
= NULL
;
590 setup_thread_stack(tsk
, orig
);
591 clear_user_return_notifier(tsk
);
592 clear_tsk_need_resched(tsk
);
593 set_task_stack_end_magic(tsk
);
595 #ifdef CONFIG_CC_STACKPROTECTOR
596 tsk
->stack_canary
= get_random_canary();
600 * One for us, one for whoever does the "release_task()" (usually
603 atomic_set(&tsk
->usage
, 2);
604 #ifdef CONFIG_BLK_DEV_IO_TRACE
607 tsk
->splice_pipe
= NULL
;
608 tsk
->task_frag
.page
= NULL
;
609 tsk
->wake_q
.next
= NULL
;
611 account_kernel_stack(tsk
, 1);
615 #ifdef CONFIG_FAULT_INJECTION
622 free_thread_stack(tsk
);
624 free_task_struct(tsk
);
629 static __latent_entropy
int dup_mmap(struct mm_struct
*mm
,
630 struct mm_struct
*oldmm
)
632 struct vm_area_struct
*mpnt
, *tmp
, *prev
, **pprev
;
633 struct rb_node
**rb_link
, *rb_parent
;
635 unsigned long charge
;
638 uprobe_start_dup_mmap();
639 if (down_write_killable(&oldmm
->mmap_sem
)) {
641 goto fail_uprobe_end
;
643 flush_cache_dup_mm(oldmm
);
644 uprobe_dup_mmap(oldmm
, mm
);
646 * Not linked in yet - no deadlock potential:
648 down_write_nested(&mm
->mmap_sem
, SINGLE_DEPTH_NESTING
);
650 /* No ordering required: file already has been exposed. */
651 RCU_INIT_POINTER(mm
->exe_file
, get_mm_exe_file(oldmm
));
653 mm
->total_vm
= oldmm
->total_vm
;
654 mm
->data_vm
= oldmm
->data_vm
;
655 mm
->exec_vm
= oldmm
->exec_vm
;
656 mm
->stack_vm
= oldmm
->stack_vm
;
658 rb_link
= &mm
->mm_rb
.rb_node
;
661 retval
= ksm_fork(mm
, oldmm
);
664 retval
= khugepaged_fork(mm
, oldmm
);
669 for (mpnt
= oldmm
->mmap
; mpnt
; mpnt
= mpnt
->vm_next
) {
672 if (mpnt
->vm_flags
& VM_DONTCOPY
) {
673 vm_stat_account(mm
, mpnt
->vm_flags
, -vma_pages(mpnt
));
678 * Don't duplicate many vmas if we've been oom-killed (for
681 if (fatal_signal_pending(current
)) {
685 if (mpnt
->vm_flags
& VM_ACCOUNT
) {
686 unsigned long len
= vma_pages(mpnt
);
688 if (security_vm_enough_memory_mm(oldmm
, len
)) /* sic */
692 tmp
= vm_area_dup(mpnt
);
695 retval
= vma_dup_policy(mpnt
, tmp
);
697 goto fail_nomem_policy
;
699 retval
= dup_userfaultfd(tmp
, &uf
);
701 goto fail_nomem_anon_vma_fork
;
702 if (tmp
->vm_flags
& VM_WIPEONFORK
) {
703 /* VM_WIPEONFORK gets a clean slate in the child. */
704 tmp
->anon_vma
= NULL
;
705 if (anon_vma_prepare(tmp
))
706 goto fail_nomem_anon_vma_fork
;
707 } else if (anon_vma_fork(tmp
, mpnt
))
708 goto fail_nomem_anon_vma_fork
;
709 tmp
->vm_flags
&= ~(VM_LOCKED
| VM_LOCKONFAULT
);
710 tmp
->vm_next
= tmp
->vm_prev
= NULL
;
713 struct inode
*inode
= file_inode(file
);
714 struct address_space
*mapping
= file
->f_mapping
;
717 if (tmp
->vm_flags
& VM_DENYWRITE
)
718 atomic_dec(&inode
->i_writecount
);
719 i_mmap_lock_write(mapping
);
720 if (tmp
->vm_flags
& VM_SHARED
)
721 atomic_inc(&mapping
->i_mmap_writable
);
722 flush_dcache_mmap_lock(mapping
);
723 /* insert tmp into the share list, just after mpnt */
724 vma_interval_tree_insert_after(tmp
, mpnt
,
726 flush_dcache_mmap_unlock(mapping
);
727 i_mmap_unlock_write(mapping
);
731 * Clear hugetlb-related page reserves for children. This only
732 * affects MAP_PRIVATE mappings. Faults generated by the child
733 * are not guaranteed to succeed, even if read-only
735 if (is_vm_hugetlb_page(tmp
))
736 reset_vma_resv_huge_pages(tmp
);
739 * Link in the new vma and copy the page table entries.
742 pprev
= &tmp
->vm_next
;
746 __vma_link_rb(mm
, tmp
, rb_link
, rb_parent
);
747 rb_link
= &tmp
->vm_rb
.rb_right
;
748 rb_parent
= &tmp
->vm_rb
;
751 if (!(tmp
->vm_flags
& VM_WIPEONFORK
))
752 retval
= copy_page_range(mm
, oldmm
, mpnt
);
754 if (tmp
->vm_ops
&& tmp
->vm_ops
->open
)
755 tmp
->vm_ops
->open(tmp
);
760 /* a new mm has just been created */
761 retval
= arch_dup_mmap(oldmm
, mm
);
763 up_write(&mm
->mmap_sem
);
765 up_write(&oldmm
->mmap_sem
);
766 dup_userfaultfd_complete(&uf
);
768 uprobe_end_dup_mmap();
770 fail_nomem_anon_vma_fork
:
771 mpol_put(vma_policy(tmp
));
776 vm_unacct_memory(charge
);
780 static inline int mm_alloc_pgd(struct mm_struct
*mm
)
782 mm
->pgd
= pgd_alloc(mm
);
783 if (unlikely(!mm
->pgd
))
788 static inline void mm_free_pgd(struct mm_struct
*mm
)
790 pgd_free(mm
, mm
->pgd
);
793 static int dup_mmap(struct mm_struct
*mm
, struct mm_struct
*oldmm
)
795 down_write(&oldmm
->mmap_sem
);
796 RCU_INIT_POINTER(mm
->exe_file
, get_mm_exe_file(oldmm
));
797 up_write(&oldmm
->mmap_sem
);
800 #define mm_alloc_pgd(mm) (0)
801 #define mm_free_pgd(mm)
802 #endif /* CONFIG_MMU */
804 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(mmlist_lock
);
806 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
807 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
809 static unsigned long default_dump_filter
= MMF_DUMP_FILTER_DEFAULT
;
811 static int __init
coredump_filter_setup(char *s
)
813 default_dump_filter
=
814 (simple_strtoul(s
, NULL
, 0) << MMF_DUMP_FILTER_SHIFT
) &
815 MMF_DUMP_FILTER_MASK
;
819 __setup("coredump_filter=", coredump_filter_setup
);
821 #include <linux/init_task.h>
823 static void mm_init_aio(struct mm_struct
*mm
)
826 spin_lock_init(&mm
->ioctx_lock
);
827 mm
->ioctx_table
= NULL
;
831 static void mm_init_owner(struct mm_struct
*mm
, struct task_struct
*p
)
838 static void mm_init_uprobes_state(struct mm_struct
*mm
)
840 #ifdef CONFIG_UPROBES
841 mm
->uprobes_state
.xol_area
= NULL
;
845 static struct mm_struct
*mm_init(struct mm_struct
*mm
, struct task_struct
*p
,
846 struct user_namespace
*user_ns
)
850 mm
->vmacache_seqnum
= 0;
851 atomic_set(&mm
->mm_users
, 1);
852 atomic_set(&mm
->mm_count
, 1);
853 init_rwsem(&mm
->mmap_sem
);
854 INIT_LIST_HEAD(&mm
->mmlist
);
855 mm
->core_state
= NULL
;
856 mm_pgtables_bytes_init(mm
);
860 memset(&mm
->rss_stat
, 0, sizeof(mm
->rss_stat
));
861 spin_lock_init(&mm
->page_table_lock
);
864 mm_init_owner(mm
, p
);
865 RCU_INIT_POINTER(mm
->exe_file
, NULL
);
866 mmu_notifier_mm_init(mm
);
868 init_tlb_flush_pending(mm
);
869 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
870 mm
->pmd_huge_pte
= NULL
;
872 mm_init_uprobes_state(mm
);
875 mm
->flags
= current
->mm
->flags
& MMF_INIT_MASK
;
876 mm
->def_flags
= current
->mm
->def_flags
& VM_INIT_DEF_MASK
;
878 mm
->flags
= default_dump_filter
;
882 if (mm_alloc_pgd(mm
))
885 if (init_new_context(p
, mm
))
888 mm
->user_ns
= get_user_ns(user_ns
);
898 static void check_mm(struct mm_struct
*mm
)
902 for (i
= 0; i
< NR_MM_COUNTERS
; i
++) {
903 long x
= atomic_long_read(&mm
->rss_stat
.count
[i
]);
906 printk(KERN_ALERT
"BUG: Bad rss-counter state "
907 "mm:%p idx:%d val:%ld\n", mm
, i
, x
);
910 if (mm_pgtables_bytes(mm
))
911 pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n",
912 mm_pgtables_bytes(mm
));
914 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
915 VM_BUG_ON_MM(mm
->pmd_huge_pte
, mm
);
920 * Allocate and initialize an mm_struct.
922 struct mm_struct
*mm_alloc(void)
924 struct mm_struct
*mm
;
930 memset(mm
, 0, sizeof(*mm
));
931 return mm_init(mm
, current
, current_user_ns());
935 * Called when the last reference to the mm
936 * is dropped: either by a lazy thread or by
937 * mmput. Free the page directory and the mm.
939 void __mmdrop(struct mm_struct
*mm
)
941 BUG_ON(mm
== &init_mm
);
945 mmu_notifier_mm_destroy(mm
);
947 put_user_ns(mm
->user_ns
);
950 EXPORT_SYMBOL_GPL(__mmdrop
);
952 static inline void __mmput(struct mm_struct
*mm
)
954 VM_BUG_ON(atomic_read(&mm
->mm_users
));
956 uprobe_clear_state(mm
);
959 khugepaged_exit(mm
); /* must run before exit_mmap */
961 mm_put_huge_zero_page(mm
);
962 set_mm_exe_file(mm
, NULL
);
963 if (!list_empty(&mm
->mmlist
)) {
964 spin_lock(&mmlist_lock
);
965 list_del(&mm
->mmlist
);
966 spin_unlock(&mmlist_lock
);
969 module_put(mm
->binfmt
->module
);
974 * Decrement the use count and release all resources for an mm.
976 void mmput(struct mm_struct
*mm
)
980 if (atomic_dec_and_test(&mm
->mm_users
))
983 EXPORT_SYMBOL_GPL(mmput
);
986 static void mmput_async_fn(struct work_struct
*work
)
988 struct mm_struct
*mm
= container_of(work
, struct mm_struct
,
994 void mmput_async(struct mm_struct
*mm
)
996 if (atomic_dec_and_test(&mm
->mm_users
)) {
997 INIT_WORK(&mm
->async_put_work
, mmput_async_fn
);
998 schedule_work(&mm
->async_put_work
);
1004 * set_mm_exe_file - change a reference to the mm's executable file
1006 * This changes mm's executable file (shown as symlink /proc/[pid]/exe).
1008 * Main users are mmput() and sys_execve(). Callers prevent concurrent
1009 * invocations: in mmput() nobody alive left, in execve task is single
1010 * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the
1011 * mm->exe_file, but does so without using set_mm_exe_file() in order
1012 * to do avoid the need for any locks.
1014 void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
)
1016 struct file
*old_exe_file
;
1019 * It is safe to dereference the exe_file without RCU as
1020 * this function is only called if nobody else can access
1021 * this mm -- see comment above for justification.
1023 old_exe_file
= rcu_dereference_raw(mm
->exe_file
);
1026 get_file(new_exe_file
);
1027 rcu_assign_pointer(mm
->exe_file
, new_exe_file
);
1033 * get_mm_exe_file - acquire a reference to the mm's executable file
1035 * Returns %NULL if mm has no associated executable file.
1036 * User must release file via fput().
1038 struct file
*get_mm_exe_file(struct mm_struct
*mm
)
1040 struct file
*exe_file
;
1043 exe_file
= rcu_dereference(mm
->exe_file
);
1044 if (exe_file
&& !get_file_rcu(exe_file
))
1049 EXPORT_SYMBOL(get_mm_exe_file
);
1052 * get_task_exe_file - acquire a reference to the task's executable file
1054 * Returns %NULL if task's mm (if any) has no associated executable file or
1055 * this is a kernel thread with borrowed mm (see the comment above get_task_mm).
1056 * User must release file via fput().
1058 struct file
*get_task_exe_file(struct task_struct
*task
)
1060 struct file
*exe_file
= NULL
;
1061 struct mm_struct
*mm
;
1066 if (!(task
->flags
& PF_KTHREAD
))
1067 exe_file
= get_mm_exe_file(mm
);
1072 EXPORT_SYMBOL(get_task_exe_file
);
1075 * get_task_mm - acquire a reference to the task's mm
1077 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
1078 * this kernel workthread has transiently adopted a user mm with use_mm,
1079 * to do its AIO) is not set and if so returns a reference to it, after
1080 * bumping up the use count. User must release the mm via mmput()
1081 * after use. Typically used by /proc and ptrace.
1083 struct mm_struct
*get_task_mm(struct task_struct
*task
)
1085 struct mm_struct
*mm
;
1090 if (task
->flags
& PF_KTHREAD
)
1098 EXPORT_SYMBOL_GPL(get_task_mm
);
1100 struct mm_struct
*mm_access(struct task_struct
*task
, unsigned int mode
)
1102 struct mm_struct
*mm
;
1105 err
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
1107 return ERR_PTR(err
);
1109 mm
= get_task_mm(task
);
1110 if (mm
&& mm
!= current
->mm
&&
1111 !ptrace_may_access(task
, mode
)) {
1113 mm
= ERR_PTR(-EACCES
);
1115 mutex_unlock(&task
->signal
->cred_guard_mutex
);
1120 static void complete_vfork_done(struct task_struct
*tsk
)
1122 struct completion
*vfork
;
1125 vfork
= tsk
->vfork_done
;
1126 if (likely(vfork
)) {
1127 tsk
->vfork_done
= NULL
;
1133 static int wait_for_vfork_done(struct task_struct
*child
,
1134 struct completion
*vfork
)
1138 freezer_do_not_count();
1139 killed
= wait_for_completion_killable(vfork
);
1144 child
->vfork_done
= NULL
;
1148 put_task_struct(child
);
1152 /* Please note the differences between mmput and mm_release.
1153 * mmput is called whenever we stop holding onto a mm_struct,
1154 * error success whatever.
1156 * mm_release is called after a mm_struct has been removed
1157 * from the current process.
1159 * This difference is important for error handling, when we
1160 * only half set up a mm_struct for a new process and need to restore
1161 * the old one. Because we mmput the new mm_struct before
1162 * restoring the old one. . .
1163 * Eric Biederman 10 January 1998
1165 void mm_release(struct task_struct
*tsk
, struct mm_struct
*mm
)
1167 /* Get rid of any futexes when releasing the mm */
1169 if (unlikely(tsk
->robust_list
)) {
1170 exit_robust_list(tsk
);
1171 tsk
->robust_list
= NULL
;
1173 #ifdef CONFIG_COMPAT
1174 if (unlikely(tsk
->compat_robust_list
)) {
1175 compat_exit_robust_list(tsk
);
1176 tsk
->compat_robust_list
= NULL
;
1179 if (unlikely(!list_empty(&tsk
->pi_state_list
)))
1180 exit_pi_state_list(tsk
);
1183 uprobe_free_utask(tsk
);
1185 /* Get rid of any cached register state */
1186 deactivate_mm(tsk
, mm
);
1189 * Signal userspace if we're not exiting with a core dump
1190 * because we want to leave the value intact for debugging
1193 if (tsk
->clear_child_tid
) {
1194 if (!(tsk
->signal
->flags
& SIGNAL_GROUP_COREDUMP
) &&
1195 atomic_read(&mm
->mm_users
) > 1) {
1197 * We don't check the error code - if userspace has
1198 * not set up a proper pointer then tough luck.
1200 put_user(0, tsk
->clear_child_tid
);
1201 sys_futex(tsk
->clear_child_tid
, FUTEX_WAKE
,
1204 tsk
->clear_child_tid
= NULL
;
1208 * All done, finally we can wake up parent and return this mm to him.
1209 * Also kthread_stop() uses this completion for synchronization.
1211 if (tsk
->vfork_done
)
1212 complete_vfork_done(tsk
);
1216 * Allocate a new mm structure and copy contents from the
1217 * mm structure of the passed in task structure.
1219 static struct mm_struct
*dup_mm(struct task_struct
*tsk
)
1221 struct mm_struct
*mm
, *oldmm
= current
->mm
;
1228 memcpy(mm
, oldmm
, sizeof(*mm
));
1230 if (!mm_init(mm
, tsk
, mm
->user_ns
))
1233 err
= dup_mmap(mm
, oldmm
);
1237 mm
->hiwater_rss
= get_mm_rss(mm
);
1238 mm
->hiwater_vm
= mm
->total_vm
;
1240 if (mm
->binfmt
&& !try_module_get(mm
->binfmt
->module
))
1246 /* don't put binfmt in mmput, we haven't got module yet */
1254 static int copy_mm(unsigned long clone_flags
, struct task_struct
*tsk
)
1256 struct mm_struct
*mm
, *oldmm
;
1259 tsk
->min_flt
= tsk
->maj_flt
= 0;
1260 tsk
->nvcsw
= tsk
->nivcsw
= 0;
1261 #ifdef CONFIG_DETECT_HUNG_TASK
1262 tsk
->last_switch_count
= tsk
->nvcsw
+ tsk
->nivcsw
;
1266 tsk
->active_mm
= NULL
;
1269 * Are we cloning a kernel thread?
1271 * We need to steal a active VM for that..
1273 oldmm
= current
->mm
;
1277 /* initialize the new vmacache entries */
1278 vmacache_flush(tsk
);
1280 if (clone_flags
& CLONE_VM
) {
1293 tsk
->active_mm
= mm
;
1300 static int copy_fs(unsigned long clone_flags
, struct task_struct
*tsk
)
1302 struct fs_struct
*fs
= current
->fs
;
1303 if (clone_flags
& CLONE_FS
) {
1304 /* tsk->fs is already what we want */
1305 spin_lock(&fs
->lock
);
1307 spin_unlock(&fs
->lock
);
1311 spin_unlock(&fs
->lock
);
1314 tsk
->fs
= copy_fs_struct(fs
);
1320 static int copy_files(unsigned long clone_flags
, struct task_struct
*tsk
)
1322 struct files_struct
*oldf
, *newf
;
1326 * A background process may not have any files ...
1328 oldf
= current
->files
;
1332 if (clone_flags
& CLONE_FILES
) {
1333 atomic_inc(&oldf
->count
);
1337 newf
= dup_fd(oldf
, &error
);
1347 static int copy_io(unsigned long clone_flags
, struct task_struct
*tsk
)
1350 struct io_context
*ioc
= current
->io_context
;
1351 struct io_context
*new_ioc
;
1356 * Share io context with parent, if CLONE_IO is set
1358 if (clone_flags
& CLONE_IO
) {
1360 tsk
->io_context
= ioc
;
1361 } else if (ioprio_valid(ioc
->ioprio
)) {
1362 new_ioc
= get_task_io_context(tsk
, GFP_KERNEL
, NUMA_NO_NODE
);
1363 if (unlikely(!new_ioc
))
1366 new_ioc
->ioprio
= ioc
->ioprio
;
1367 put_io_context(new_ioc
);
1373 static int copy_sighand(unsigned long clone_flags
, struct task_struct
*tsk
)
1375 struct sighand_struct
*sig
;
1377 if (clone_flags
& CLONE_SIGHAND
) {
1378 atomic_inc(¤t
->sighand
->count
);
1381 sig
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
1382 rcu_assign_pointer(tsk
->sighand
, sig
);
1386 atomic_set(&sig
->count
, 1);
1387 spin_lock_irq(¤t
->sighand
->siglock
);
1388 memcpy(sig
->action
, current
->sighand
->action
, sizeof(sig
->action
));
1389 spin_unlock_irq(¤t
->sighand
->siglock
);
1393 void __cleanup_sighand(struct sighand_struct
*sighand
)
1395 if (atomic_dec_and_test(&sighand
->count
)) {
1396 signalfd_cleanup(sighand
);
1398 * sighand_cachep is SLAB_TYPESAFE_BY_RCU so we can free it
1399 * without an RCU grace period, see __lock_task_sighand().
1401 kmem_cache_free(sighand_cachep
, sighand
);
1405 #ifdef CONFIG_POSIX_TIMERS
1407 * Initialize POSIX timer handling for a thread group.
1409 static void posix_cpu_timers_init_group(struct signal_struct
*sig
)
1411 unsigned long cpu_limit
;
1413 cpu_limit
= READ_ONCE(sig
->rlim
[RLIMIT_CPU
].rlim_cur
);
1414 if (cpu_limit
!= RLIM_INFINITY
) {
1415 sig
->cputime_expires
.prof_exp
= cpu_limit
* NSEC_PER_SEC
;
1416 sig
->cputimer
.running
= true;
1419 /* The timer lists. */
1420 INIT_LIST_HEAD(&sig
->cpu_timers
[0]);
1421 INIT_LIST_HEAD(&sig
->cpu_timers
[1]);
1422 INIT_LIST_HEAD(&sig
->cpu_timers
[2]);
1425 static inline void posix_cpu_timers_init_group(struct signal_struct
*sig
) { }
1428 static int copy_signal(unsigned long clone_flags
, struct task_struct
*tsk
)
1430 struct signal_struct
*sig
;
1432 if (clone_flags
& CLONE_THREAD
)
1435 sig
= kmem_cache_zalloc(signal_cachep
, GFP_KERNEL
);
1440 sig
->nr_threads
= 1;
1441 atomic_set(&sig
->live
, 1);
1442 atomic_set(&sig
->sigcnt
, 1);
1444 /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */
1445 sig
->thread_head
= (struct list_head
)LIST_HEAD_INIT(tsk
->thread_node
);
1446 tsk
->thread_node
= (struct list_head
)LIST_HEAD_INIT(sig
->thread_head
);
1448 init_waitqueue_head(&sig
->wait_chldexit
);
1449 sig
->curr_target
= tsk
;
1450 init_sigpending(&sig
->shared_pending
);
1451 seqlock_init(&sig
->stats_lock
);
1452 prev_cputime_init(&sig
->prev_cputime
);
1454 #ifdef CONFIG_POSIX_TIMERS
1455 INIT_LIST_HEAD(&sig
->posix_timers
);
1456 hrtimer_init(&sig
->real_timer
, CLOCK_MONOTONIC
, HRTIMER_MODE_REL
);
1457 sig
->real_timer
.function
= it_real_fn
;
1460 task_lock(current
->group_leader
);
1461 memcpy(sig
->rlim
, current
->signal
->rlim
, sizeof sig
->rlim
);
1462 task_unlock(current
->group_leader
);
1464 posix_cpu_timers_init_group(sig
);
1466 tty_audit_fork(sig
);
1467 sched_autogroup_fork(sig
);
1469 sig
->oom_score_adj
= current
->signal
->oom_score_adj
;
1470 sig
->oom_score_adj_min
= current
->signal
->oom_score_adj_min
;
1472 mutex_init(&sig
->cred_guard_mutex
);
1477 static void copy_seccomp(struct task_struct
*p
)
1479 #ifdef CONFIG_SECCOMP
1481 * Must be called with sighand->lock held, which is common to
1482 * all threads in the group. Holding cred_guard_mutex is not
1483 * needed because this new task is not yet running and cannot
1486 assert_spin_locked(¤t
->sighand
->siglock
);
1488 /* Ref-count the new filter user, and assign it. */
1489 get_seccomp_filter(current
);
1490 p
->seccomp
= current
->seccomp
;
1493 * Explicitly enable no_new_privs here in case it got set
1494 * between the task_struct being duplicated and holding the
1495 * sighand lock. The seccomp state and nnp must be in sync.
1497 if (task_no_new_privs(current
))
1498 task_set_no_new_privs(p
);
1501 * If the parent gained a seccomp mode after copying thread
1502 * flags and between before we held the sighand lock, we have
1503 * to manually enable the seccomp thread flag here.
1505 if (p
->seccomp
.mode
!= SECCOMP_MODE_DISABLED
)
1506 set_tsk_thread_flag(p
, TIF_SECCOMP
);
1510 SYSCALL_DEFINE1(set_tid_address
, int __user
*, tidptr
)
1512 current
->clear_child_tid
= tidptr
;
1514 return task_pid_vnr(current
);
1517 static void rt_mutex_init_task(struct task_struct
*p
)
1519 raw_spin_lock_init(&p
->pi_lock
);
1520 #ifdef CONFIG_RT_MUTEXES
1521 p
->pi_waiters
= RB_ROOT_CACHED
;
1522 p
->pi_top_task
= NULL
;
1523 p
->pi_blocked_on
= NULL
;
1527 #ifdef CONFIG_POSIX_TIMERS
1529 * Initialize POSIX timer handling for a single task.
1531 static void posix_cpu_timers_init(struct task_struct
*tsk
)
1533 tsk
->cputime_expires
.prof_exp
= 0;
1534 tsk
->cputime_expires
.virt_exp
= 0;
1535 tsk
->cputime_expires
.sched_exp
= 0;
1536 INIT_LIST_HEAD(&tsk
->cpu_timers
[0]);
1537 INIT_LIST_HEAD(&tsk
->cpu_timers
[1]);
1538 INIT_LIST_HEAD(&tsk
->cpu_timers
[2]);
1541 static inline void posix_cpu_timers_init(struct task_struct
*tsk
) { }
1545 init_task_pid(struct task_struct
*task
, enum pid_type type
, struct pid
*pid
)
1547 task
->pids
[type
].pid
= pid
;
1550 static inline void rcu_copy_process(struct task_struct
*p
)
1552 #ifdef CONFIG_PREEMPT_RCU
1553 p
->rcu_read_lock_nesting
= 0;
1554 p
->rcu_read_unlock_special
.s
= 0;
1555 p
->rcu_blocked_node
= NULL
;
1556 INIT_LIST_HEAD(&p
->rcu_node_entry
);
1557 #endif /* #ifdef CONFIG_PREEMPT_RCU */
1558 #ifdef CONFIG_TASKS_RCU
1559 p
->rcu_tasks_holdout
= false;
1560 INIT_LIST_HEAD(&p
->rcu_tasks_holdout_list
);
1561 p
->rcu_tasks_idle_cpu
= -1;
1562 #endif /* #ifdef CONFIG_TASKS_RCU */
1566 * This creates a new process as a copy of the old one,
1567 * but does not actually start it yet.
1569 * It copies the registers, and all the appropriate
1570 * parts of the process environment (as per the clone
1571 * flags). The actual kick-off is left to the caller.
1573 static __latent_entropy
struct task_struct
*copy_process(
1574 unsigned long clone_flags
,
1575 unsigned long stack_start
,
1576 unsigned long stack_size
,
1577 int __user
*child_tidptr
,
1584 struct task_struct
*p
;
1586 if ((clone_flags
& (CLONE_NEWNS
|CLONE_FS
)) == (CLONE_NEWNS
|CLONE_FS
))
1587 return ERR_PTR(-EINVAL
);
1589 if ((clone_flags
& (CLONE_NEWUSER
|CLONE_FS
)) == (CLONE_NEWUSER
|CLONE_FS
))
1590 return ERR_PTR(-EINVAL
);
1592 if ((clone_flags
& CLONE_NEWUSER
) && !unprivileged_userns_clone
)
1593 if (!capable(CAP_SYS_ADMIN
))
1594 return ERR_PTR(-EPERM
);
1597 * Thread groups must share signals as well, and detached threads
1598 * can only be started up within the thread group.
1600 if ((clone_flags
& CLONE_THREAD
) && !(clone_flags
& CLONE_SIGHAND
))
1601 return ERR_PTR(-EINVAL
);
1604 * Shared signal handlers imply shared VM. By way of the above,
1605 * thread groups also imply shared VM. Blocking this case allows
1606 * for various simplifications in other code.
1608 if ((clone_flags
& CLONE_SIGHAND
) && !(clone_flags
& CLONE_VM
))
1609 return ERR_PTR(-EINVAL
);
1612 * Siblings of global init remain as zombies on exit since they are
1613 * not reaped by their parent (swapper). To solve this and to avoid
1614 * multi-rooted process trees, prevent global and container-inits
1615 * from creating siblings.
1617 if ((clone_flags
& CLONE_PARENT
) &&
1618 current
->signal
->flags
& SIGNAL_UNKILLABLE
)
1619 return ERR_PTR(-EINVAL
);
1622 * If the new process will be in a different pid or user namespace
1623 * do not allow it to share a thread group with the forking task.
1625 if (clone_flags
& CLONE_THREAD
) {
1626 if ((clone_flags
& (CLONE_NEWUSER
| CLONE_NEWPID
)) ||
1627 (task_active_pid_ns(current
) !=
1628 current
->nsproxy
->pid_ns_for_children
))
1629 return ERR_PTR(-EINVAL
);
1633 p
= dup_task_struct(current
, node
);
1638 * This _must_ happen before we call free_task(), i.e. before we jump
1639 * to any of the bad_fork_* labels. This is to avoid freeing
1640 * p->set_child_tid which is (ab)used as a kthread's data pointer for
1641 * kernel threads (PF_KTHREAD).
1643 p
->set_child_tid
= (clone_flags
& CLONE_CHILD_SETTID
) ? child_tidptr
: NULL
;
1645 * Clear TID on mm_release()?
1647 p
->clear_child_tid
= (clone_flags
& CLONE_CHILD_CLEARTID
) ? child_tidptr
: NULL
;
1649 ftrace_graph_init_task(p
);
1651 rt_mutex_init_task(p
);
1653 #ifdef CONFIG_PROVE_LOCKING
1654 DEBUG_LOCKS_WARN_ON(!p
->hardirqs_enabled
);
1655 DEBUG_LOCKS_WARN_ON(!p
->softirqs_enabled
);
1658 if (atomic_read(&p
->real_cred
->user
->processes
) >=
1659 task_rlimit(p
, RLIMIT_NPROC
)) {
1660 if (p
->real_cred
->user
!= INIT_USER
&&
1661 !capable(CAP_SYS_RESOURCE
) && !capable(CAP_SYS_ADMIN
))
1664 current
->flags
&= ~PF_NPROC_EXCEEDED
;
1666 retval
= copy_creds(p
, clone_flags
);
1671 * If multiple threads are within copy_process(), then this check
1672 * triggers too late. This doesn't hurt, the check is only there
1673 * to stop root fork bombs.
1676 if (nr_threads
>= max_threads
)
1677 goto bad_fork_cleanup_count
;
1679 delayacct_tsk_init(p
); /* Must remain after dup_task_struct() */
1680 p
->flags
&= ~(PF_SUPERPRIV
| PF_WQ_WORKER
| PF_IDLE
);
1681 p
->flags
|= PF_FORKNOEXEC
;
1682 INIT_LIST_HEAD(&p
->children
);
1683 INIT_LIST_HEAD(&p
->sibling
);
1684 rcu_copy_process(p
);
1685 p
->vfork_done
= NULL
;
1686 spin_lock_init(&p
->alloc_lock
);
1688 init_sigpending(&p
->pending
);
1690 p
->utime
= p
->stime
= p
->gtime
= 0;
1691 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1692 p
->utimescaled
= p
->stimescaled
= 0;
1694 prev_cputime_init(&p
->prev_cputime
);
1696 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1697 seqcount_init(&p
->vtime
.seqcount
);
1698 p
->vtime
.starttime
= 0;
1699 p
->vtime
.state
= VTIME_INACTIVE
;
1702 #if defined(SPLIT_RSS_COUNTING)
1703 memset(&p
->rss_stat
, 0, sizeof(p
->rss_stat
));
1706 p
->default_timer_slack_ns
= current
->timer_slack_ns
;
1708 task_io_accounting_init(&p
->ioac
);
1709 acct_clear_integrals(p
);
1711 posix_cpu_timers_init(p
);
1713 p
->io_context
= NULL
;
1714 p
->audit_context
= NULL
;
1717 p
->mempolicy
= mpol_dup(p
->mempolicy
);
1718 if (IS_ERR(p
->mempolicy
)) {
1719 retval
= PTR_ERR(p
->mempolicy
);
1720 p
->mempolicy
= NULL
;
1721 goto bad_fork_cleanup_threadgroup_lock
;
1724 #ifdef CONFIG_CPUSETS
1725 p
->cpuset_mem_spread_rotor
= NUMA_NO_NODE
;
1726 p
->cpuset_slab_spread_rotor
= NUMA_NO_NODE
;
1727 seqcount_init(&p
->mems_allowed_seq
);
1729 #ifdef CONFIG_TRACE_IRQFLAGS
1731 p
->hardirqs_enabled
= 0;
1732 p
->hardirq_enable_ip
= 0;
1733 p
->hardirq_enable_event
= 0;
1734 p
->hardirq_disable_ip
= _THIS_IP_
;
1735 p
->hardirq_disable_event
= 0;
1736 p
->softirqs_enabled
= 1;
1737 p
->softirq_enable_ip
= _THIS_IP_
;
1738 p
->softirq_enable_event
= 0;
1739 p
->softirq_disable_ip
= 0;
1740 p
->softirq_disable_event
= 0;
1741 p
->hardirq_context
= 0;
1742 p
->softirq_context
= 0;
1745 p
->pagefault_disabled
= 0;
1747 #ifdef CONFIG_LOCKDEP
1748 p
->lockdep_depth
= 0; /* no locks held yet */
1749 p
->curr_chain_key
= 0;
1750 p
->lockdep_recursion
= 0;
1751 lockdep_init_task(p
);
1754 #ifdef CONFIG_DEBUG_MUTEXES
1755 p
->blocked_on
= NULL
; /* not blocked yet */
1757 #ifdef CONFIG_BCACHE
1758 p
->sequential_io
= 0;
1759 p
->sequential_io_avg
= 0;
1761 #ifdef CONFIG_SECURITY
1765 /* Perform scheduler related setup. Assign this task to a CPU. */
1766 retval
= sched_fork(clone_flags
, p
);
1768 goto bad_fork_cleanup_policy
;
1770 retval
= perf_event_init_task(p
);
1772 goto bad_fork_cleanup_policy
;
1773 retval
= audit_alloc(p
);
1775 goto bad_fork_cleanup_perf
;
1776 /* copy all the process information */
1778 retval
= security_task_alloc(p
, clone_flags
);
1780 goto bad_fork_cleanup_audit
;
1781 retval
= copy_semundo(clone_flags
, p
);
1783 goto bad_fork_cleanup_security
;
1784 retval
= copy_files(clone_flags
, p
);
1786 goto bad_fork_cleanup_semundo
;
1787 retval
= copy_fs(clone_flags
, p
);
1789 goto bad_fork_cleanup_files
;
1790 retval
= copy_sighand(clone_flags
, p
);
1792 goto bad_fork_cleanup_fs
;
1793 retval
= copy_signal(clone_flags
, p
);
1795 goto bad_fork_cleanup_sighand
;
1796 retval
= copy_mm(clone_flags
, p
);
1798 goto bad_fork_cleanup_signal
;
1799 retval
= copy_namespaces(clone_flags
, p
);
1801 goto bad_fork_cleanup_mm
;
1802 retval
= copy_io(clone_flags
, p
);
1804 goto bad_fork_cleanup_namespaces
;
1805 retval
= copy_thread_tls(clone_flags
, stack_start
, stack_size
, p
, tls
);
1807 goto bad_fork_cleanup_io
;
1809 if (pid
!= &init_struct_pid
) {
1810 pid
= alloc_pid(p
->nsproxy
->pid_ns_for_children
);
1812 retval
= PTR_ERR(pid
);
1813 goto bad_fork_cleanup_thread
;
1821 p
->robust_list
= NULL
;
1822 #ifdef CONFIG_COMPAT
1823 p
->compat_robust_list
= NULL
;
1825 INIT_LIST_HEAD(&p
->pi_state_list
);
1826 p
->pi_state_cache
= NULL
;
1829 * sigaltstack should be cleared when sharing the same VM
1831 if ((clone_flags
& (CLONE_VM
|CLONE_VFORK
)) == CLONE_VM
)
1835 * Syscall tracing and stepping should be turned off in the
1836 * child regardless of CLONE_PTRACE.
1838 user_disable_single_step(p
);
1839 clear_tsk_thread_flag(p
, TIF_SYSCALL_TRACE
);
1840 #ifdef TIF_SYSCALL_EMU
1841 clear_tsk_thread_flag(p
, TIF_SYSCALL_EMU
);
1843 clear_all_latency_tracing(p
);
1845 /* ok, now we should be set up.. */
1846 p
->pid
= pid_nr(pid
);
1847 if (clone_flags
& CLONE_THREAD
) {
1848 p
->exit_signal
= -1;
1849 p
->group_leader
= current
->group_leader
;
1850 p
->tgid
= current
->tgid
;
1852 if (clone_flags
& CLONE_PARENT
)
1853 p
->exit_signal
= current
->group_leader
->exit_signal
;
1855 p
->exit_signal
= (clone_flags
& CSIGNAL
);
1856 p
->group_leader
= p
;
1861 p
->nr_dirtied_pause
= 128 >> (PAGE_SHIFT
- 10);
1862 p
->dirty_paused_when
= 0;
1864 p
->pdeath_signal
= 0;
1865 INIT_LIST_HEAD(&p
->thread_group
);
1866 p
->task_works
= NULL
;
1868 cgroup_threadgroup_change_begin(current
);
1870 * Ensure that the cgroup subsystem policies allow the new process to be
1871 * forked. It should be noted the the new process's css_set can be changed
1872 * between here and cgroup_post_fork() if an organisation operation is in
1875 retval
= cgroup_can_fork(p
);
1877 goto bad_fork_free_pid
;
1880 * From this point on we must avoid any synchronous user-space
1881 * communication until we take the tasklist-lock. In particular, we do
1882 * not want user-space to be able to predict the process start-time by
1883 * stalling fork(2) after we recorded the start_time but before it is
1884 * visible to the system.
1887 p
->start_time
= ktime_get_ns();
1888 p
->real_start_time
= ktime_get_boot_ns();
1891 * Make it visible to the rest of the system, but dont wake it up yet.
1892 * Need tasklist lock for parent etc handling!
1894 write_lock_irq(&tasklist_lock
);
1896 /* CLONE_PARENT re-uses the old parent */
1897 if (clone_flags
& (CLONE_PARENT
|CLONE_THREAD
)) {
1898 p
->real_parent
= current
->real_parent
;
1899 p
->parent_exec_id
= current
->parent_exec_id
;
1901 p
->real_parent
= current
;
1902 p
->parent_exec_id
= current
->self_exec_id
;
1905 klp_copy_process(p
);
1907 spin_lock(¤t
->sighand
->siglock
);
1910 * Copy seccomp details explicitly here, in case they were changed
1911 * before holding sighand lock.
1916 * Process group and session signals need to be delivered to just the
1917 * parent before the fork or both the parent and the child after the
1918 * fork. Restart if a signal comes in before we add the new process to
1919 * it's process group.
1920 * A fatal signal pending means that current will exit, so the new
1921 * thread can't slip out of an OOM kill (or normal SIGKILL).
1923 recalc_sigpending();
1924 if (signal_pending(current
)) {
1925 retval
= -ERESTARTNOINTR
;
1926 goto bad_fork_cancel_cgroup
;
1928 if (unlikely(!(ns_of_pid(pid
)->pid_allocated
& PIDNS_ADDING
))) {
1930 goto bad_fork_cancel_cgroup
;
1933 if (likely(p
->pid
)) {
1934 ptrace_init_task(p
, (clone_flags
& CLONE_PTRACE
) || trace
);
1936 init_task_pid(p
, PIDTYPE_PID
, pid
);
1937 if (thread_group_leader(p
)) {
1938 init_task_pid(p
, PIDTYPE_PGID
, task_pgrp(current
));
1939 init_task_pid(p
, PIDTYPE_SID
, task_session(current
));
1941 if (is_child_reaper(pid
)) {
1942 ns_of_pid(pid
)->child_reaper
= p
;
1943 p
->signal
->flags
|= SIGNAL_UNKILLABLE
;
1946 p
->signal
->leader_pid
= pid
;
1947 p
->signal
->tty
= tty_kref_get(current
->signal
->tty
);
1949 * Inherit has_child_subreaper flag under the same
1950 * tasklist_lock with adding child to the process tree
1951 * for propagate_has_child_subreaper optimization.
1953 p
->signal
->has_child_subreaper
= p
->real_parent
->signal
->has_child_subreaper
||
1954 p
->real_parent
->signal
->is_child_subreaper
;
1955 list_add_tail(&p
->sibling
, &p
->real_parent
->children
);
1956 list_add_tail_rcu(&p
->tasks
, &init_task
.tasks
);
1957 attach_pid(p
, PIDTYPE_PGID
);
1958 attach_pid(p
, PIDTYPE_SID
);
1959 __this_cpu_inc(process_counts
);
1961 current
->signal
->nr_threads
++;
1962 atomic_inc(¤t
->signal
->live
);
1963 atomic_inc(¤t
->signal
->sigcnt
);
1964 list_add_tail_rcu(&p
->thread_group
,
1965 &p
->group_leader
->thread_group
);
1966 list_add_tail_rcu(&p
->thread_node
,
1967 &p
->signal
->thread_head
);
1969 attach_pid(p
, PIDTYPE_PID
);
1974 spin_unlock(¤t
->sighand
->siglock
);
1975 syscall_tracepoint_update(p
);
1976 write_unlock_irq(&tasklist_lock
);
1978 proc_fork_connector(p
);
1979 cgroup_post_fork(p
);
1980 cgroup_threadgroup_change_end(current
);
1983 trace_task_newtask(p
, clone_flags
);
1984 uprobe_copy_process(p
, clone_flags
);
1988 bad_fork_cancel_cgroup
:
1989 spin_unlock(¤t
->sighand
->siglock
);
1990 write_unlock_irq(&tasklist_lock
);
1991 cgroup_cancel_fork(p
);
1993 cgroup_threadgroup_change_end(current
);
1994 if (pid
!= &init_struct_pid
)
1996 bad_fork_cleanup_thread
:
1998 bad_fork_cleanup_io
:
2001 bad_fork_cleanup_namespaces
:
2002 exit_task_namespaces(p
);
2003 bad_fork_cleanup_mm
:
2006 bad_fork_cleanup_signal
:
2007 if (!(clone_flags
& CLONE_THREAD
))
2008 free_signal_struct(p
->signal
);
2009 bad_fork_cleanup_sighand
:
2010 __cleanup_sighand(p
->sighand
);
2011 bad_fork_cleanup_fs
:
2012 exit_fs(p
); /* blocking */
2013 bad_fork_cleanup_files
:
2014 exit_files(p
); /* blocking */
2015 bad_fork_cleanup_semundo
:
2017 bad_fork_cleanup_security
:
2018 security_task_free(p
);
2019 bad_fork_cleanup_audit
:
2021 bad_fork_cleanup_perf
:
2022 perf_event_free_task(p
);
2023 bad_fork_cleanup_policy
:
2024 lockdep_free_task(p
);
2026 mpol_put(p
->mempolicy
);
2027 bad_fork_cleanup_threadgroup_lock
:
2029 delayacct_tsk_free(p
);
2030 bad_fork_cleanup_count
:
2031 atomic_dec(&p
->cred
->user
->processes
);
2034 p
->state
= TASK_DEAD
;
2038 return ERR_PTR(retval
);
2041 static inline void init_idle_pids(struct pid_link
*links
)
2045 for (type
= PIDTYPE_PID
; type
< PIDTYPE_MAX
; ++type
) {
2046 INIT_HLIST_NODE(&links
[type
].node
); /* not really needed */
2047 links
[type
].pid
= &init_struct_pid
;
2051 struct task_struct
*fork_idle(int cpu
)
2053 struct task_struct
*task
;
2054 task
= copy_process(CLONE_VM
, 0, 0, NULL
, &init_struct_pid
, 0, 0,
2056 if (!IS_ERR(task
)) {
2057 init_idle_pids(task
->pids
);
2058 init_idle(task
, cpu
);
2065 * Ok, this is the main fork-routine.
2067 * It copies the process, and if successful kick-starts
2068 * it and waits for it to finish using the VM if required.
2070 long _do_fork(unsigned long clone_flags
,
2071 unsigned long stack_start
,
2072 unsigned long stack_size
,
2073 int __user
*parent_tidptr
,
2074 int __user
*child_tidptr
,
2077 struct task_struct
*p
;
2082 * Determine whether and which event to report to ptracer. When
2083 * called from kernel_thread or CLONE_UNTRACED is explicitly
2084 * requested, no event is reported; otherwise, report if the event
2085 * for the type of forking is enabled.
2087 if (!(clone_flags
& CLONE_UNTRACED
)) {
2088 if (clone_flags
& CLONE_VFORK
)
2089 trace
= PTRACE_EVENT_VFORK
;
2090 else if ((clone_flags
& CSIGNAL
) != SIGCHLD
)
2091 trace
= PTRACE_EVENT_CLONE
;
2093 trace
= PTRACE_EVENT_FORK
;
2095 if (likely(!ptrace_event_enabled(current
, trace
)))
2099 p
= copy_process(clone_flags
, stack_start
, stack_size
,
2100 child_tidptr
, NULL
, trace
, tls
, NUMA_NO_NODE
);
2101 add_latent_entropy();
2103 * Do this prior waking up the new thread - the thread pointer
2104 * might get invalid after that point, if the thread exits quickly.
2107 struct completion vfork
;
2110 trace_sched_process_fork(current
, p
);
2112 pid
= get_task_pid(p
, PIDTYPE_PID
);
2115 if (clone_flags
& CLONE_PARENT_SETTID
)
2116 put_user(nr
, parent_tidptr
);
2118 if (clone_flags
& CLONE_VFORK
) {
2119 p
->vfork_done
= &vfork
;
2120 init_completion(&vfork
);
2124 wake_up_new_task(p
);
2126 /* forking complete and child started to run, tell ptracer */
2127 if (unlikely(trace
))
2128 ptrace_event_pid(trace
, pid
);
2130 if (clone_flags
& CLONE_VFORK
) {
2131 if (!wait_for_vfork_done(p
, &vfork
))
2132 ptrace_event_pid(PTRACE_EVENT_VFORK_DONE
, pid
);
2142 #ifndef CONFIG_HAVE_COPY_THREAD_TLS
2143 /* For compatibility with architectures that call do_fork directly rather than
2144 * using the syscall entry points below. */
2145 long do_fork(unsigned long clone_flags
,
2146 unsigned long stack_start
,
2147 unsigned long stack_size
,
2148 int __user
*parent_tidptr
,
2149 int __user
*child_tidptr
)
2151 return _do_fork(clone_flags
, stack_start
, stack_size
,
2152 parent_tidptr
, child_tidptr
, 0);
2157 * Create a kernel thread.
2159 pid_t
kernel_thread(int (*fn
)(void *), void *arg
, unsigned long flags
)
2161 return _do_fork(flags
|CLONE_VM
|CLONE_UNTRACED
, (unsigned long)fn
,
2162 (unsigned long)arg
, NULL
, NULL
, 0);
2165 #ifdef __ARCH_WANT_SYS_FORK
2166 SYSCALL_DEFINE0(fork
)
2169 return _do_fork(SIGCHLD
, 0, 0, NULL
, NULL
, 0);
2171 /* can not support in nommu mode */
2177 #ifdef __ARCH_WANT_SYS_VFORK
2178 SYSCALL_DEFINE0(vfork
)
2180 return _do_fork(CLONE_VFORK
| CLONE_VM
| SIGCHLD
, 0,
2185 #ifdef __ARCH_WANT_SYS_CLONE
2186 #ifdef CONFIG_CLONE_BACKWARDS
2187 SYSCALL_DEFINE5(clone
, unsigned long, clone_flags
, unsigned long, newsp
,
2188 int __user
*, parent_tidptr
,
2190 int __user
*, child_tidptr
)
2191 #elif defined(CONFIG_CLONE_BACKWARDS2)
2192 SYSCALL_DEFINE5(clone
, unsigned long, newsp
, unsigned long, clone_flags
,
2193 int __user
*, parent_tidptr
,
2194 int __user
*, child_tidptr
,
2196 #elif defined(CONFIG_CLONE_BACKWARDS3)
2197 SYSCALL_DEFINE6(clone
, unsigned long, clone_flags
, unsigned long, newsp
,
2199 int __user
*, parent_tidptr
,
2200 int __user
*, child_tidptr
,
2203 SYSCALL_DEFINE5(clone
, unsigned long, clone_flags
, unsigned long, newsp
,
2204 int __user
*, parent_tidptr
,
2205 int __user
*, child_tidptr
,
2209 return _do_fork(clone_flags
, newsp
, 0, parent_tidptr
, child_tidptr
, tls
);
2213 void walk_process_tree(struct task_struct
*top
, proc_visitor visitor
, void *data
)
2215 struct task_struct
*leader
, *parent
, *child
;
2218 read_lock(&tasklist_lock
);
2219 leader
= top
= top
->group_leader
;
2221 for_each_thread(leader
, parent
) {
2222 list_for_each_entry(child
, &parent
->children
, sibling
) {
2223 res
= visitor(child
, data
);
2235 if (leader
!= top
) {
2237 parent
= child
->real_parent
;
2238 leader
= parent
->group_leader
;
2242 read_unlock(&tasklist_lock
);
2245 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
2246 #define ARCH_MIN_MMSTRUCT_ALIGN 0
2249 static void sighand_ctor(void *data
)
2251 struct sighand_struct
*sighand
= data
;
2253 spin_lock_init(&sighand
->siglock
);
2254 init_waitqueue_head(&sighand
->signalfd_wqh
);
2257 void __init
proc_caches_init(void)
2259 sighand_cachep
= kmem_cache_create("sighand_cache",
2260 sizeof(struct sighand_struct
), 0,
2261 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_TYPESAFE_BY_RCU
|
2262 SLAB_ACCOUNT
, sighand_ctor
);
2263 signal_cachep
= kmem_cache_create("signal_cache",
2264 sizeof(struct signal_struct
), 0,
2265 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_ACCOUNT
,
2267 files_cachep
= kmem_cache_create("files_cache",
2268 sizeof(struct files_struct
), 0,
2269 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_ACCOUNT
,
2271 fs_cachep
= kmem_cache_create("fs_cache",
2272 sizeof(struct fs_struct
), 0,
2273 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_ACCOUNT
,
2276 * FIXME! The "sizeof(struct mm_struct)" currently includes the
2277 * whole struct cpumask for the OFFSTACK case. We could change
2278 * this to *only* allocate as much of it as required by the
2279 * maximum number of CPU's we can ever have. The cpumask_allocation
2280 * is at the end of the structure, exactly for that reason.
2282 mm_cachep
= kmem_cache_create("mm_struct",
2283 sizeof(struct mm_struct
), ARCH_MIN_MMSTRUCT_ALIGN
,
2284 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_ACCOUNT
,
2286 vm_area_cachep
= KMEM_CACHE(vm_area_struct
, SLAB_PANIC
|SLAB_ACCOUNT
);
2288 nsproxy_cache_init();
2292 * Check constraints on flags passed to the unshare system call.
2294 static int check_unshare_flags(unsigned long unshare_flags
)
2296 if (unshare_flags
& ~(CLONE_THREAD
|CLONE_FS
|CLONE_NEWNS
|CLONE_SIGHAND
|
2297 CLONE_VM
|CLONE_FILES
|CLONE_SYSVSEM
|
2298 CLONE_NEWUTS
|CLONE_NEWIPC
|CLONE_NEWNET
|
2299 CLONE_NEWUSER
|CLONE_NEWPID
|CLONE_NEWCGROUP
))
2302 * Not implemented, but pretend it works if there is nothing
2303 * to unshare. Note that unsharing the address space or the
2304 * signal handlers also need to unshare the signal queues (aka
2307 if (unshare_flags
& (CLONE_THREAD
| CLONE_SIGHAND
| CLONE_VM
)) {
2308 if (!thread_group_empty(current
))
2311 if (unshare_flags
& (CLONE_SIGHAND
| CLONE_VM
)) {
2312 if (atomic_read(¤t
->sighand
->count
) > 1)
2315 if (unshare_flags
& CLONE_VM
) {
2316 if (!current_is_single_threaded())
2324 * Unshare the filesystem structure if it is being shared
2326 static int unshare_fs(unsigned long unshare_flags
, struct fs_struct
**new_fsp
)
2328 struct fs_struct
*fs
= current
->fs
;
2330 if (!(unshare_flags
& CLONE_FS
) || !fs
)
2333 /* don't need lock here; in the worst case we'll do useless copy */
2337 *new_fsp
= copy_fs_struct(fs
);
2345 * Unshare file descriptor table if it is being shared
2347 static int unshare_fd(unsigned long unshare_flags
, struct files_struct
**new_fdp
)
2349 struct files_struct
*fd
= current
->files
;
2352 if ((unshare_flags
& CLONE_FILES
) &&
2353 (fd
&& atomic_read(&fd
->count
) > 1)) {
2354 *new_fdp
= dup_fd(fd
, &error
);
2363 * unshare allows a process to 'unshare' part of the process
2364 * context which was originally shared using clone. copy_*
2365 * functions used by do_fork() cannot be used here directly
2366 * because they modify an inactive task_struct that is being
2367 * constructed. Here we are modifying the current, active,
2370 SYSCALL_DEFINE1(unshare
, unsigned long, unshare_flags
)
2372 struct fs_struct
*fs
, *new_fs
= NULL
;
2373 struct files_struct
*fd
, *new_fd
= NULL
;
2374 struct cred
*new_cred
= NULL
;
2375 struct nsproxy
*new_nsproxy
= NULL
;
2380 * If unsharing a user namespace must also unshare the thread group
2381 * and unshare the filesystem root and working directories.
2383 if (unshare_flags
& CLONE_NEWUSER
)
2384 unshare_flags
|= CLONE_THREAD
| CLONE_FS
;
2386 * If unsharing vm, must also unshare signal handlers.
2388 if (unshare_flags
& CLONE_VM
)
2389 unshare_flags
|= CLONE_SIGHAND
;
2391 * If unsharing a signal handlers, must also unshare the signal queues.
2393 if (unshare_flags
& CLONE_SIGHAND
)
2394 unshare_flags
|= CLONE_THREAD
;
2396 * If unsharing namespace, must also unshare filesystem information.
2398 if (unshare_flags
& CLONE_NEWNS
)
2399 unshare_flags
|= CLONE_FS
;
2401 if ((unshare_flags
& CLONE_NEWUSER
) && !unprivileged_userns_clone
) {
2403 if (!capable(CAP_SYS_ADMIN
))
2404 goto bad_unshare_out
;
2407 err
= check_unshare_flags(unshare_flags
);
2409 goto bad_unshare_out
;
2411 * CLONE_NEWIPC must also detach from the undolist: after switching
2412 * to a new ipc namespace, the semaphore arrays from the old
2413 * namespace are unreachable.
2415 if (unshare_flags
& (CLONE_NEWIPC
|CLONE_SYSVSEM
))
2417 err
= unshare_fs(unshare_flags
, &new_fs
);
2419 goto bad_unshare_out
;
2420 err
= unshare_fd(unshare_flags
, &new_fd
);
2422 goto bad_unshare_cleanup_fs
;
2423 err
= unshare_userns(unshare_flags
, &new_cred
);
2425 goto bad_unshare_cleanup_fd
;
2426 err
= unshare_nsproxy_namespaces(unshare_flags
, &new_nsproxy
,
2429 goto bad_unshare_cleanup_cred
;
2431 if (new_fs
|| new_fd
|| do_sysvsem
|| new_cred
|| new_nsproxy
) {
2434 * CLONE_SYSVSEM is equivalent to sys_exit().
2438 if (unshare_flags
& CLONE_NEWIPC
) {
2439 /* Orphan segments in old ns (see sem above). */
2441 shm_init_task(current
);
2445 switch_task_namespaces(current
, new_nsproxy
);
2451 spin_lock(&fs
->lock
);
2452 current
->fs
= new_fs
;
2457 spin_unlock(&fs
->lock
);
2461 fd
= current
->files
;
2462 current
->files
= new_fd
;
2466 task_unlock(current
);
2469 /* Install the new user namespace */
2470 commit_creds(new_cred
);
2475 perf_event_namespaces(current
);
2477 bad_unshare_cleanup_cred
:
2480 bad_unshare_cleanup_fd
:
2482 put_files_struct(new_fd
);
2484 bad_unshare_cleanup_fs
:
2486 free_fs_struct(new_fs
);
2493 * Helper to unshare the files of the current task.
2494 * We don't want to expose copy_files internals to
2495 * the exec layer of the kernel.
2498 int unshare_files(struct files_struct
**displaced
)
2500 struct task_struct
*task
= current
;
2501 struct files_struct
*copy
= NULL
;
2504 error
= unshare_fd(CLONE_FILES
, ©
);
2505 if (error
|| !copy
) {
2509 *displaced
= task
->files
;
2516 int sysctl_max_threads(struct ctl_table
*table
, int write
,
2517 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
2521 int threads
= max_threads
;
2522 int min
= MIN_THREADS
;
2523 int max
= MAX_THREADS
;
2530 ret
= proc_dointvec_minmax(&t
, write
, buffer
, lenp
, ppos
);
2534 set_max_threads(threads
);