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>
42 #include <linux/vmacache.h>
43 #include <linux/nsproxy.h>
44 #include <linux/capability.h>
45 #include <linux/cpu.h>
46 #include <linux/cgroup.h>
47 #include <linux/security.h>
48 #include <linux/hugetlb.h>
49 #include <linux/seccomp.h>
50 #include <linux/swap.h>
51 #include <linux/syscalls.h>
52 #include <linux/jiffies.h>
53 #include <linux/futex.h>
54 #include <linux/compat.h>
55 #include <linux/kthread.h>
56 #include <linux/task_io_accounting_ops.h>
57 #include <linux/rcupdate.h>
58 #include <linux/ptrace.h>
59 #include <linux/mount.h>
60 #include <linux/audit.h>
61 #include <linux/memcontrol.h>
62 #include <linux/ftrace.h>
63 #include <linux/proc_fs.h>
64 #include <linux/profile.h>
65 #include <linux/rmap.h>
66 #include <linux/ksm.h>
67 #include <linux/acct.h>
68 #include <linux/userfaultfd_k.h>
69 #include <linux/tsacct_kern.h>
70 #include <linux/cn_proc.h>
71 #include <linux/freezer.h>
72 #include <linux/delayacct.h>
73 #include <linux/taskstats_kern.h>
74 #include <linux/random.h>
75 #include <linux/tty.h>
76 #include <linux/blkdev.h>
77 #include <linux/fs_struct.h>
78 #include <linux/magic.h>
79 #include <linux/perf_event.h>
80 #include <linux/posix-timers.h>
81 #include <linux/user-return-notifier.h>
82 #include <linux/oom.h>
83 #include <linux/khugepaged.h>
84 #include <linux/signalfd.h>
85 #include <linux/uprobes.h>
86 #include <linux/aio.h>
87 #include <linux/compiler.h>
88 #include <linux/sysctl.h>
89 #include <linux/kcov.h>
90 #include <linux/livepatch.h>
92 #include <asm/pgtable.h>
93 #include <asm/pgalloc.h>
94 #include <linux/uaccess.h>
95 #include <asm/mmu_context.h>
96 #include <asm/cacheflush.h>
97 #include <asm/tlbflush.h>
99 #include <trace/events/sched.h>
101 #define CREATE_TRACE_POINTS
102 #include <trace/events/task.h>
103 #ifdef CONFIG_USER_NS
104 extern int unprivileged_userns_clone
;
106 #define unprivileged_userns_clone 0
110 * Minimum number of threads to boot the kernel
112 #define MIN_THREADS 20
115 * Maximum number of threads
117 #define MAX_THREADS FUTEX_TID_MASK
120 * Protected counters by write_lock_irq(&tasklist_lock)
122 unsigned long total_forks
; /* Handle normal Linux uptimes. */
123 int nr_threads
; /* The idle threads do not count.. */
125 int max_threads
; /* tunable limit on nr_threads */
127 DEFINE_PER_CPU(unsigned long, process_counts
) = 0;
129 __cacheline_aligned
DEFINE_RWLOCK(tasklist_lock
); /* outer */
131 #ifdef CONFIG_PROVE_RCU
132 int lockdep_tasklist_lock_is_held(void)
134 return lockdep_is_held(&tasklist_lock
);
136 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held
);
137 #endif /* #ifdef CONFIG_PROVE_RCU */
139 int nr_processes(void)
144 for_each_possible_cpu(cpu
)
145 total
+= per_cpu(process_counts
, cpu
);
150 void __weak
arch_release_task_struct(struct task_struct
*tsk
)
154 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
155 static struct kmem_cache
*task_struct_cachep
;
157 static inline struct task_struct
*alloc_task_struct_node(int node
)
159 return kmem_cache_alloc_node(task_struct_cachep
, GFP_KERNEL
, node
);
162 static inline void free_task_struct(struct task_struct
*tsk
)
164 kmem_cache_free(task_struct_cachep
, tsk
);
168 void __weak
arch_release_thread_stack(unsigned long *stack
)
172 #ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR
175 * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a
176 * kmemcache based allocator.
178 # if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK)
180 #ifdef CONFIG_VMAP_STACK
182 * vmalloc() is a bit slow, and calling vfree() enough times will force a TLB
183 * flush. Try to minimize the number of calls by caching stacks.
185 #define NR_CACHED_STACKS 2
186 static DEFINE_PER_CPU(struct vm_struct
*, cached_stacks
[NR_CACHED_STACKS
]);
188 static int free_vm_stack_cache(unsigned int cpu
)
190 struct vm_struct
**cached_vm_stacks
= per_cpu_ptr(cached_stacks
, cpu
);
193 for (i
= 0; i
< NR_CACHED_STACKS
; i
++) {
194 struct vm_struct
*vm_stack
= cached_vm_stacks
[i
];
199 vfree(vm_stack
->addr
);
200 cached_vm_stacks
[i
] = NULL
;
207 static unsigned long *alloc_thread_stack_node(struct task_struct
*tsk
, int node
)
209 #ifdef CONFIG_VMAP_STACK
213 for (i
= 0; i
< NR_CACHED_STACKS
; i
++) {
216 s
= this_cpu_xchg(cached_stacks
[i
], NULL
);
221 tsk
->stack_vm_area
= s
;
225 stack
= __vmalloc_node_range(THREAD_SIZE
, THREAD_SIZE
,
226 VMALLOC_START
, VMALLOC_END
,
229 0, node
, __builtin_return_address(0));
232 * We can't call find_vm_area() in interrupt context, and
233 * free_thread_stack() can be called in interrupt context,
234 * so cache the vm_struct.
237 tsk
->stack_vm_area
= find_vm_area(stack
);
240 struct page
*page
= alloc_pages_node(node
, THREADINFO_GFP
,
243 return page
? page_address(page
) : NULL
;
247 static inline void free_thread_stack(struct task_struct
*tsk
)
249 #ifdef CONFIG_VMAP_STACK
250 if (task_stack_vm_area(tsk
)) {
253 for (i
= 0; i
< NR_CACHED_STACKS
; i
++) {
254 if (this_cpu_cmpxchg(cached_stacks
[i
],
255 NULL
, tsk
->stack_vm_area
) != NULL
)
261 vfree_atomic(tsk
->stack
);
266 __free_pages(virt_to_page(tsk
->stack
), THREAD_SIZE_ORDER
);
269 static struct kmem_cache
*thread_stack_cache
;
271 static unsigned long *alloc_thread_stack_node(struct task_struct
*tsk
,
274 return kmem_cache_alloc_node(thread_stack_cache
, THREADINFO_GFP
, node
);
277 static void free_thread_stack(struct task_struct
*tsk
)
279 kmem_cache_free(thread_stack_cache
, tsk
->stack
);
282 void thread_stack_cache_init(void)
284 thread_stack_cache
= kmem_cache_create("thread_stack", THREAD_SIZE
,
285 THREAD_SIZE
, 0, NULL
);
286 BUG_ON(thread_stack_cache
== NULL
);
291 /* SLAB cache for signal_struct structures (tsk->signal) */
292 static struct kmem_cache
*signal_cachep
;
294 /* SLAB cache for sighand_struct structures (tsk->sighand) */
295 struct kmem_cache
*sighand_cachep
;
297 /* SLAB cache for files_struct structures (tsk->files) */
298 struct kmem_cache
*files_cachep
;
300 /* SLAB cache for fs_struct structures (tsk->fs) */
301 struct kmem_cache
*fs_cachep
;
303 /* SLAB cache for vm_area_struct structures */
304 struct kmem_cache
*vm_area_cachep
;
306 /* SLAB cache for mm_struct structures (tsk->mm) */
307 static struct kmem_cache
*mm_cachep
;
309 static void account_kernel_stack(struct task_struct
*tsk
, int account
)
311 void *stack
= task_stack_page(tsk
);
312 struct vm_struct
*vm
= task_stack_vm_area(tsk
);
314 BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK
) && PAGE_SIZE
% 1024 != 0);
319 BUG_ON(vm
->nr_pages
!= THREAD_SIZE
/ PAGE_SIZE
);
321 for (i
= 0; i
< THREAD_SIZE
/ PAGE_SIZE
; i
++) {
322 mod_zone_page_state(page_zone(vm
->pages
[i
]),
324 PAGE_SIZE
/ 1024 * account
);
327 /* All stack pages belong to the same memcg. */
328 mod_memcg_page_state(vm
->pages
[0], MEMCG_KERNEL_STACK_KB
,
329 account
* (THREAD_SIZE
/ 1024));
332 * All stack pages are in the same zone and belong to the
335 struct page
*first_page
= virt_to_page(stack
);
337 mod_zone_page_state(page_zone(first_page
), NR_KERNEL_STACK_KB
,
338 THREAD_SIZE
/ 1024 * account
);
340 mod_memcg_page_state(first_page
, MEMCG_KERNEL_STACK_KB
,
341 account
* (THREAD_SIZE
/ 1024));
345 static void release_task_stack(struct task_struct
*tsk
)
347 if (WARN_ON(tsk
->state
!= TASK_DEAD
))
348 return; /* Better to leak the stack than to free prematurely */
350 account_kernel_stack(tsk
, -1);
351 arch_release_thread_stack(tsk
->stack
);
352 free_thread_stack(tsk
);
354 #ifdef CONFIG_VMAP_STACK
355 tsk
->stack_vm_area
= NULL
;
359 #ifdef CONFIG_THREAD_INFO_IN_TASK
360 void put_task_stack(struct task_struct
*tsk
)
362 if (atomic_dec_and_test(&tsk
->stack_refcount
))
363 release_task_stack(tsk
);
367 void free_task(struct task_struct
*tsk
)
369 #ifndef CONFIG_THREAD_INFO_IN_TASK
371 * The task is finally done with both the stack and thread_info,
374 release_task_stack(tsk
);
377 * If the task had a separate stack allocation, it should be gone
380 WARN_ON_ONCE(atomic_read(&tsk
->stack_refcount
) != 0);
382 rt_mutex_debug_task_free(tsk
);
383 ftrace_graph_exit_task(tsk
);
384 put_seccomp_filter(tsk
);
385 arch_release_task_struct(tsk
);
386 if (tsk
->flags
& PF_KTHREAD
)
387 free_kthread_struct(tsk
);
388 free_task_struct(tsk
);
390 EXPORT_SYMBOL(free_task
);
392 static inline void free_signal_struct(struct signal_struct
*sig
)
394 taskstats_tgid_free(sig
);
395 sched_autogroup_exit(sig
);
397 * __mmdrop is not safe to call from softirq context on x86 due to
398 * pgd_dtor so postpone it to the async context
401 mmdrop_async(sig
->oom_mm
);
402 kmem_cache_free(signal_cachep
, sig
);
405 static inline void put_signal_struct(struct signal_struct
*sig
)
407 if (atomic_dec_and_test(&sig
->sigcnt
))
408 free_signal_struct(sig
);
411 void __put_task_struct(struct task_struct
*tsk
)
413 WARN_ON(!tsk
->exit_state
);
414 WARN_ON(atomic_read(&tsk
->usage
));
415 WARN_ON(tsk
== current
);
419 security_task_free(tsk
);
421 delayacct_tsk_free(tsk
);
422 put_signal_struct(tsk
->signal
);
424 if (!profile_handoff_task(tsk
))
427 EXPORT_SYMBOL_GPL(__put_task_struct
);
429 void __init __weak
arch_task_cache_init(void) { }
434 static void set_max_threads(unsigned int max_threads_suggested
)
439 * The number of threads shall be limited such that the thread
440 * structures may only consume a small part of the available memory.
442 if (fls64(totalram_pages
) + fls64(PAGE_SIZE
) > 64)
443 threads
= MAX_THREADS
;
445 threads
= div64_u64((u64
) totalram_pages
* (u64
) PAGE_SIZE
,
446 (u64
) THREAD_SIZE
* 8UL);
448 if (threads
> max_threads_suggested
)
449 threads
= max_threads_suggested
;
451 max_threads
= clamp_t(u64
, threads
, MIN_THREADS
, MAX_THREADS
);
454 #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
455 /* Initialized by the architecture: */
456 int arch_task_struct_size __read_mostly
;
459 void __init
fork_init(void)
462 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
463 #ifndef ARCH_MIN_TASKALIGN
464 #define ARCH_MIN_TASKALIGN 0
466 int align
= max_t(int, L1_CACHE_BYTES
, ARCH_MIN_TASKALIGN
);
468 /* create a slab on which task_structs can be allocated */
469 task_struct_cachep
= kmem_cache_create("task_struct",
470 arch_task_struct_size
, align
,
471 SLAB_PANIC
|SLAB_NOTRACK
|SLAB_ACCOUNT
, NULL
);
474 /* do the arch specific task caches init */
475 arch_task_cache_init();
477 set_max_threads(MAX_THREADS
);
479 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_cur
= max_threads
/2;
480 init_task
.signal
->rlim
[RLIMIT_NPROC
].rlim_max
= max_threads
/2;
481 init_task
.signal
->rlim
[RLIMIT_SIGPENDING
] =
482 init_task
.signal
->rlim
[RLIMIT_NPROC
];
484 for (i
= 0; i
< UCOUNT_COUNTS
; i
++) {
485 init_user_ns
.ucount_max
[i
] = max_threads
/2;
488 #ifdef CONFIG_VMAP_STACK
489 cpuhp_setup_state(CPUHP_BP_PREPARE_DYN
, "fork:vm_stack_cache",
490 NULL
, free_vm_stack_cache
);
494 int __weak
arch_dup_task_struct(struct task_struct
*dst
,
495 struct task_struct
*src
)
501 void set_task_stack_end_magic(struct task_struct
*tsk
)
503 unsigned long *stackend
;
505 stackend
= end_of_stack(tsk
);
506 *stackend
= STACK_END_MAGIC
; /* for overflow detection */
509 static struct task_struct
*dup_task_struct(struct task_struct
*orig
, int node
)
511 struct task_struct
*tsk
;
512 unsigned long *stack
;
513 struct vm_struct
*stack_vm_area
;
516 if (node
== NUMA_NO_NODE
)
517 node
= tsk_fork_get_node(orig
);
518 tsk
= alloc_task_struct_node(node
);
522 stack
= alloc_thread_stack_node(tsk
, node
);
526 stack_vm_area
= task_stack_vm_area(tsk
);
528 err
= arch_dup_task_struct(tsk
, orig
);
531 * arch_dup_task_struct() clobbers the stack-related fields. Make
532 * sure they're properly initialized before using any stack-related
536 #ifdef CONFIG_VMAP_STACK
537 tsk
->stack_vm_area
= stack_vm_area
;
539 #ifdef CONFIG_THREAD_INFO_IN_TASK
540 atomic_set(&tsk
->stack_refcount
, 1);
546 #ifdef CONFIG_SECCOMP
548 * We must handle setting up seccomp filters once we're under
549 * the sighand lock in case orig has changed between now and
550 * then. Until then, filter must be NULL to avoid messing up
551 * the usage counts on the error path calling free_task.
553 tsk
->seccomp
.filter
= NULL
;
556 setup_thread_stack(tsk
, orig
);
557 clear_user_return_notifier(tsk
);
558 clear_tsk_need_resched(tsk
);
559 set_task_stack_end_magic(tsk
);
561 #ifdef CONFIG_CC_STACKPROTECTOR
562 tsk
->stack_canary
= get_random_canary();
566 * One for us, one for whoever does the "release_task()" (usually
569 atomic_set(&tsk
->usage
, 2);
570 #ifdef CONFIG_BLK_DEV_IO_TRACE
573 tsk
->splice_pipe
= NULL
;
574 tsk
->task_frag
.page
= NULL
;
575 tsk
->wake_q
.next
= NULL
;
577 account_kernel_stack(tsk
, 1);
581 #ifdef CONFIG_FAULT_INJECTION
588 free_thread_stack(tsk
);
590 free_task_struct(tsk
);
595 static __latent_entropy
int dup_mmap(struct mm_struct
*mm
,
596 struct mm_struct
*oldmm
)
598 struct vm_area_struct
*mpnt
, *tmp
, *prev
, **pprev
;
599 struct rb_node
**rb_link
, *rb_parent
;
601 unsigned long charge
;
604 uprobe_start_dup_mmap();
605 if (down_write_killable(&oldmm
->mmap_sem
)) {
607 goto fail_uprobe_end
;
609 flush_cache_dup_mm(oldmm
);
610 uprobe_dup_mmap(oldmm
, mm
);
612 * Not linked in yet - no deadlock potential:
614 down_write_nested(&mm
->mmap_sem
, SINGLE_DEPTH_NESTING
);
616 /* No ordering required: file already has been exposed. */
617 RCU_INIT_POINTER(mm
->exe_file
, get_mm_exe_file(oldmm
));
619 mm
->total_vm
= oldmm
->total_vm
;
620 mm
->data_vm
= oldmm
->data_vm
;
621 mm
->exec_vm
= oldmm
->exec_vm
;
622 mm
->stack_vm
= oldmm
->stack_vm
;
624 rb_link
= &mm
->mm_rb
.rb_node
;
627 retval
= ksm_fork(mm
, oldmm
);
630 retval
= khugepaged_fork(mm
, oldmm
);
635 for (mpnt
= oldmm
->mmap
; mpnt
; mpnt
= mpnt
->vm_next
) {
638 if (mpnt
->vm_flags
& VM_DONTCOPY
) {
639 vm_stat_account(mm
, mpnt
->vm_flags
, -vma_pages(mpnt
));
643 if (mpnt
->vm_flags
& VM_ACCOUNT
) {
644 unsigned long len
= vma_pages(mpnt
);
646 if (security_vm_enough_memory_mm(oldmm
, len
)) /* sic */
650 tmp
= kmem_cache_alloc(vm_area_cachep
, GFP_KERNEL
);
654 INIT_LIST_HEAD(&tmp
->anon_vma_chain
);
655 retval
= vma_dup_policy(mpnt
, tmp
);
657 goto fail_nomem_policy
;
659 retval
= dup_userfaultfd(tmp
, &uf
);
661 goto fail_nomem_anon_vma_fork
;
662 if (anon_vma_fork(tmp
, mpnt
))
663 goto fail_nomem_anon_vma_fork
;
664 tmp
->vm_flags
&= ~(VM_LOCKED
| VM_LOCKONFAULT
);
665 tmp
->vm_next
= tmp
->vm_prev
= NULL
;
668 struct inode
*inode
= file_inode(file
);
669 struct address_space
*mapping
= file
->f_mapping
;
672 if (tmp
->vm_flags
& VM_DENYWRITE
)
673 atomic_dec(&inode
->i_writecount
);
674 i_mmap_lock_write(mapping
);
675 if (tmp
->vm_flags
& VM_SHARED
)
676 atomic_inc(&mapping
->i_mmap_writable
);
677 flush_dcache_mmap_lock(mapping
);
678 /* insert tmp into the share list, just after mpnt */
679 vma_interval_tree_insert_after(tmp
, mpnt
,
681 flush_dcache_mmap_unlock(mapping
);
682 i_mmap_unlock_write(mapping
);
686 * Clear hugetlb-related page reserves for children. This only
687 * affects MAP_PRIVATE mappings. Faults generated by the child
688 * are not guaranteed to succeed, even if read-only
690 if (is_vm_hugetlb_page(tmp
))
691 reset_vma_resv_huge_pages(tmp
);
694 * Link in the new vma and copy the page table entries.
697 pprev
= &tmp
->vm_next
;
701 __vma_link_rb(mm
, tmp
, rb_link
, rb_parent
);
702 rb_link
= &tmp
->vm_rb
.rb_right
;
703 rb_parent
= &tmp
->vm_rb
;
706 retval
= copy_page_range(mm
, oldmm
, mpnt
);
708 if (tmp
->vm_ops
&& tmp
->vm_ops
->open
)
709 tmp
->vm_ops
->open(tmp
);
714 /* a new mm has just been created */
715 arch_dup_mmap(oldmm
, mm
);
718 up_write(&mm
->mmap_sem
);
720 up_write(&oldmm
->mmap_sem
);
721 dup_userfaultfd_complete(&uf
);
723 uprobe_end_dup_mmap();
725 fail_nomem_anon_vma_fork
:
726 mpol_put(vma_policy(tmp
));
728 kmem_cache_free(vm_area_cachep
, tmp
);
731 vm_unacct_memory(charge
);
735 static inline int mm_alloc_pgd(struct mm_struct
*mm
)
737 mm
->pgd
= pgd_alloc(mm
);
738 if (unlikely(!mm
->pgd
))
743 static inline void mm_free_pgd(struct mm_struct
*mm
)
745 pgd_free(mm
, mm
->pgd
);
748 static int dup_mmap(struct mm_struct
*mm
, struct mm_struct
*oldmm
)
750 down_write(&oldmm
->mmap_sem
);
751 RCU_INIT_POINTER(mm
->exe_file
, get_mm_exe_file(oldmm
));
752 up_write(&oldmm
->mmap_sem
);
755 #define mm_alloc_pgd(mm) (0)
756 #define mm_free_pgd(mm)
757 #endif /* CONFIG_MMU */
759 __cacheline_aligned_in_smp
DEFINE_SPINLOCK(mmlist_lock
);
761 #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
762 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
764 static unsigned long default_dump_filter
= MMF_DUMP_FILTER_DEFAULT
;
766 static int __init
coredump_filter_setup(char *s
)
768 default_dump_filter
=
769 (simple_strtoul(s
, NULL
, 0) << MMF_DUMP_FILTER_SHIFT
) &
770 MMF_DUMP_FILTER_MASK
;
774 __setup("coredump_filter=", coredump_filter_setup
);
776 #include <linux/init_task.h>
778 static void mm_init_aio(struct mm_struct
*mm
)
781 spin_lock_init(&mm
->ioctx_lock
);
782 mm
->ioctx_table
= NULL
;
786 static void mm_init_owner(struct mm_struct
*mm
, struct task_struct
*p
)
793 static void mm_init_uprobes_state(struct mm_struct
*mm
)
795 #ifdef CONFIG_UPROBES
796 mm
->uprobes_state
.xol_area
= NULL
;
800 static struct mm_struct
*mm_init(struct mm_struct
*mm
, struct task_struct
*p
,
801 struct user_namespace
*user_ns
)
805 mm
->vmacache_seqnum
= 0;
806 atomic_set(&mm
->mm_users
, 1);
807 atomic_set(&mm
->mm_count
, 1);
808 init_rwsem(&mm
->mmap_sem
);
809 INIT_LIST_HEAD(&mm
->mmlist
);
810 mm
->core_state
= NULL
;
811 atomic_long_set(&mm
->nr_ptes
, 0);
816 memset(&mm
->rss_stat
, 0, sizeof(mm
->rss_stat
));
817 spin_lock_init(&mm
->page_table_lock
);
820 mm_init_owner(mm
, p
);
821 RCU_INIT_POINTER(mm
->exe_file
, NULL
);
822 mmu_notifier_mm_init(mm
);
823 init_tlb_flush_pending(mm
);
824 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
825 mm
->pmd_huge_pte
= NULL
;
827 mm_init_uprobes_state(mm
);
830 mm
->flags
= current
->mm
->flags
& MMF_INIT_MASK
;
831 mm
->def_flags
= current
->mm
->def_flags
& VM_INIT_DEF_MASK
;
833 mm
->flags
= default_dump_filter
;
837 if (mm_alloc_pgd(mm
))
840 if (init_new_context(p
, mm
))
843 mm
->user_ns
= get_user_ns(user_ns
);
853 static void check_mm(struct mm_struct
*mm
)
857 for (i
= 0; i
< NR_MM_COUNTERS
; i
++) {
858 long x
= atomic_long_read(&mm
->rss_stat
.count
[i
]);
861 printk(KERN_ALERT
"BUG: Bad rss-counter state "
862 "mm:%p idx:%d val:%ld\n", mm
, i
, x
);
865 if (atomic_long_read(&mm
->nr_ptes
))
866 pr_alert("BUG: non-zero nr_ptes on freeing mm: %ld\n",
867 atomic_long_read(&mm
->nr_ptes
));
869 pr_alert("BUG: non-zero nr_pmds on freeing mm: %ld\n",
872 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
873 VM_BUG_ON_MM(mm
->pmd_huge_pte
, mm
);
878 * Allocate and initialize an mm_struct.
880 struct mm_struct
*mm_alloc(void)
882 struct mm_struct
*mm
;
888 memset(mm
, 0, sizeof(*mm
));
889 return mm_init(mm
, current
, current_user_ns());
893 * Called when the last reference to the mm
894 * is dropped: either by a lazy thread or by
895 * mmput. Free the page directory and the mm.
897 void __mmdrop(struct mm_struct
*mm
)
899 BUG_ON(mm
== &init_mm
);
902 mmu_notifier_mm_destroy(mm
);
904 put_user_ns(mm
->user_ns
);
907 EXPORT_SYMBOL_GPL(__mmdrop
);
909 static inline void __mmput(struct mm_struct
*mm
)
911 VM_BUG_ON(atomic_read(&mm
->mm_users
));
913 uprobe_clear_state(mm
);
916 khugepaged_exit(mm
); /* must run before exit_mmap */
918 mm_put_huge_zero_page(mm
);
919 set_mm_exe_file(mm
, NULL
);
920 if (!list_empty(&mm
->mmlist
)) {
921 spin_lock(&mmlist_lock
);
922 list_del(&mm
->mmlist
);
923 spin_unlock(&mmlist_lock
);
926 module_put(mm
->binfmt
->module
);
927 set_bit(MMF_OOM_SKIP
, &mm
->flags
);
932 * Decrement the use count and release all resources for an mm.
934 void mmput(struct mm_struct
*mm
)
938 if (atomic_dec_and_test(&mm
->mm_users
))
941 EXPORT_SYMBOL_GPL(mmput
);
944 static void mmput_async_fn(struct work_struct
*work
)
946 struct mm_struct
*mm
= container_of(work
, struct mm_struct
, async_put_work
);
950 void mmput_async(struct mm_struct
*mm
)
952 if (atomic_dec_and_test(&mm
->mm_users
)) {
953 INIT_WORK(&mm
->async_put_work
, mmput_async_fn
);
954 schedule_work(&mm
->async_put_work
);
960 * set_mm_exe_file - change a reference to the mm's executable file
962 * This changes mm's executable file (shown as symlink /proc/[pid]/exe).
964 * Main users are mmput() and sys_execve(). Callers prevent concurrent
965 * invocations: in mmput() nobody alive left, in execve task is single
966 * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the
967 * mm->exe_file, but does so without using set_mm_exe_file() in order
968 * to do avoid the need for any locks.
970 void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
)
972 struct file
*old_exe_file
;
975 * It is safe to dereference the exe_file without RCU as
976 * this function is only called if nobody else can access
977 * this mm -- see comment above for justification.
979 old_exe_file
= rcu_dereference_raw(mm
->exe_file
);
982 get_file(new_exe_file
);
983 rcu_assign_pointer(mm
->exe_file
, new_exe_file
);
989 * get_mm_exe_file - acquire a reference to the mm's executable file
991 * Returns %NULL if mm has no associated executable file.
992 * User must release file via fput().
994 struct file
*get_mm_exe_file(struct mm_struct
*mm
)
996 struct file
*exe_file
;
999 exe_file
= rcu_dereference(mm
->exe_file
);
1000 if (exe_file
&& !get_file_rcu(exe_file
))
1005 EXPORT_SYMBOL(get_mm_exe_file
);
1008 * get_task_exe_file - acquire a reference to the task's executable file
1010 * Returns %NULL if task's mm (if any) has no associated executable file or
1011 * this is a kernel thread with borrowed mm (see the comment above get_task_mm).
1012 * User must release file via fput().
1014 struct file
*get_task_exe_file(struct task_struct
*task
)
1016 struct file
*exe_file
= NULL
;
1017 struct mm_struct
*mm
;
1022 if (!(task
->flags
& PF_KTHREAD
))
1023 exe_file
= get_mm_exe_file(mm
);
1028 EXPORT_SYMBOL(get_task_exe_file
);
1031 * get_task_mm - acquire a reference to the task's mm
1033 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
1034 * this kernel workthread has transiently adopted a user mm with use_mm,
1035 * to do its AIO) is not set and if so returns a reference to it, after
1036 * bumping up the use count. User must release the mm via mmput()
1037 * after use. Typically used by /proc and ptrace.
1039 struct mm_struct
*get_task_mm(struct task_struct
*task
)
1041 struct mm_struct
*mm
;
1046 if (task
->flags
& PF_KTHREAD
)
1054 EXPORT_SYMBOL_GPL(get_task_mm
);
1056 struct mm_struct
*mm_access(struct task_struct
*task
, unsigned int mode
)
1058 struct mm_struct
*mm
;
1061 err
= mutex_lock_killable(&task
->signal
->cred_guard_mutex
);
1063 return ERR_PTR(err
);
1065 mm
= get_task_mm(task
);
1066 if (mm
&& mm
!= current
->mm
&&
1067 !ptrace_may_access(task
, mode
)) {
1069 mm
= ERR_PTR(-EACCES
);
1071 mutex_unlock(&task
->signal
->cred_guard_mutex
);
1076 static void complete_vfork_done(struct task_struct
*tsk
)
1078 struct completion
*vfork
;
1081 vfork
= tsk
->vfork_done
;
1082 if (likely(vfork
)) {
1083 tsk
->vfork_done
= NULL
;
1089 static int wait_for_vfork_done(struct task_struct
*child
,
1090 struct completion
*vfork
)
1094 freezer_do_not_count();
1095 killed
= wait_for_completion_killable(vfork
);
1100 child
->vfork_done
= NULL
;
1104 put_task_struct(child
);
1108 /* Please note the differences between mmput and mm_release.
1109 * mmput is called whenever we stop holding onto a mm_struct,
1110 * error success whatever.
1112 * mm_release is called after a mm_struct has been removed
1113 * from the current process.
1115 * This difference is important for error handling, when we
1116 * only half set up a mm_struct for a new process and need to restore
1117 * the old one. Because we mmput the new mm_struct before
1118 * restoring the old one. . .
1119 * Eric Biederman 10 January 1998
1121 void mm_release(struct task_struct
*tsk
, struct mm_struct
*mm
)
1123 /* Get rid of any futexes when releasing the mm */
1125 if (unlikely(tsk
->robust_list
)) {
1126 exit_robust_list(tsk
);
1127 tsk
->robust_list
= NULL
;
1129 #ifdef CONFIG_COMPAT
1130 if (unlikely(tsk
->compat_robust_list
)) {
1131 compat_exit_robust_list(tsk
);
1132 tsk
->compat_robust_list
= NULL
;
1135 if (unlikely(!list_empty(&tsk
->pi_state_list
)))
1136 exit_pi_state_list(tsk
);
1139 uprobe_free_utask(tsk
);
1141 /* Get rid of any cached register state */
1142 deactivate_mm(tsk
, mm
);
1145 * Signal userspace if we're not exiting with a core dump
1146 * because we want to leave the value intact for debugging
1149 if (tsk
->clear_child_tid
) {
1150 if (!(tsk
->signal
->flags
& SIGNAL_GROUP_COREDUMP
) &&
1151 atomic_read(&mm
->mm_users
) > 1) {
1153 * We don't check the error code - if userspace has
1154 * not set up a proper pointer then tough luck.
1156 put_user(0, tsk
->clear_child_tid
);
1157 sys_futex(tsk
->clear_child_tid
, FUTEX_WAKE
,
1160 tsk
->clear_child_tid
= NULL
;
1164 * All done, finally we can wake up parent and return this mm to him.
1165 * Also kthread_stop() uses this completion for synchronization.
1167 if (tsk
->vfork_done
)
1168 complete_vfork_done(tsk
);
1172 * Allocate a new mm structure and copy contents from the
1173 * mm structure of the passed in task structure.
1175 static struct mm_struct
*dup_mm(struct task_struct
*tsk
)
1177 struct mm_struct
*mm
, *oldmm
= current
->mm
;
1184 memcpy(mm
, oldmm
, sizeof(*mm
));
1186 if (!mm_init(mm
, tsk
, mm
->user_ns
))
1189 err
= dup_mmap(mm
, oldmm
);
1193 mm
->hiwater_rss
= get_mm_rss(mm
);
1194 mm
->hiwater_vm
= mm
->total_vm
;
1196 if (mm
->binfmt
&& !try_module_get(mm
->binfmt
->module
))
1202 /* don't put binfmt in mmput, we haven't got module yet */
1210 static int copy_mm(unsigned long clone_flags
, struct task_struct
*tsk
)
1212 struct mm_struct
*mm
, *oldmm
;
1215 tsk
->min_flt
= tsk
->maj_flt
= 0;
1216 tsk
->nvcsw
= tsk
->nivcsw
= 0;
1217 #ifdef CONFIG_DETECT_HUNG_TASK
1218 tsk
->last_switch_count
= tsk
->nvcsw
+ tsk
->nivcsw
;
1222 tsk
->active_mm
= NULL
;
1225 * Are we cloning a kernel thread?
1227 * We need to steal a active VM for that..
1229 oldmm
= current
->mm
;
1233 /* initialize the new vmacache entries */
1234 vmacache_flush(tsk
);
1236 if (clone_flags
& CLONE_VM
) {
1249 tsk
->active_mm
= mm
;
1256 static int copy_fs(unsigned long clone_flags
, struct task_struct
*tsk
)
1258 struct fs_struct
*fs
= current
->fs
;
1259 if (clone_flags
& CLONE_FS
) {
1260 /* tsk->fs is already what we want */
1261 spin_lock(&fs
->lock
);
1263 spin_unlock(&fs
->lock
);
1267 spin_unlock(&fs
->lock
);
1270 tsk
->fs
= copy_fs_struct(fs
);
1276 static int copy_files(unsigned long clone_flags
, struct task_struct
*tsk
)
1278 struct files_struct
*oldf
, *newf
;
1282 * A background process may not have any files ...
1284 oldf
= current
->files
;
1288 if (clone_flags
& CLONE_FILES
) {
1289 atomic_inc(&oldf
->count
);
1293 newf
= dup_fd(oldf
, &error
);
1303 static int copy_io(unsigned long clone_flags
, struct task_struct
*tsk
)
1306 struct io_context
*ioc
= current
->io_context
;
1307 struct io_context
*new_ioc
;
1312 * Share io context with parent, if CLONE_IO is set
1314 if (clone_flags
& CLONE_IO
) {
1316 tsk
->io_context
= ioc
;
1317 } else if (ioprio_valid(ioc
->ioprio
)) {
1318 new_ioc
= get_task_io_context(tsk
, GFP_KERNEL
, NUMA_NO_NODE
);
1319 if (unlikely(!new_ioc
))
1322 new_ioc
->ioprio
= ioc
->ioprio
;
1323 put_io_context(new_ioc
);
1329 static int copy_sighand(unsigned long clone_flags
, struct task_struct
*tsk
)
1331 struct sighand_struct
*sig
;
1333 if (clone_flags
& CLONE_SIGHAND
) {
1334 atomic_inc(¤t
->sighand
->count
);
1337 sig
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
1338 rcu_assign_pointer(tsk
->sighand
, sig
);
1342 atomic_set(&sig
->count
, 1);
1343 memcpy(sig
->action
, current
->sighand
->action
, sizeof(sig
->action
));
1347 void __cleanup_sighand(struct sighand_struct
*sighand
)
1349 if (atomic_dec_and_test(&sighand
->count
)) {
1350 signalfd_cleanup(sighand
);
1352 * sighand_cachep is SLAB_TYPESAFE_BY_RCU so we can free it
1353 * without an RCU grace period, see __lock_task_sighand().
1355 kmem_cache_free(sighand_cachep
, sighand
);
1359 #ifdef CONFIG_POSIX_TIMERS
1361 * Initialize POSIX timer handling for a thread group.
1363 static void posix_cpu_timers_init_group(struct signal_struct
*sig
)
1365 unsigned long cpu_limit
;
1367 cpu_limit
= READ_ONCE(sig
->rlim
[RLIMIT_CPU
].rlim_cur
);
1368 if (cpu_limit
!= RLIM_INFINITY
) {
1369 sig
->cputime_expires
.prof_exp
= cpu_limit
* NSEC_PER_SEC
;
1370 sig
->cputimer
.running
= true;
1373 /* The timer lists. */
1374 INIT_LIST_HEAD(&sig
->cpu_timers
[0]);
1375 INIT_LIST_HEAD(&sig
->cpu_timers
[1]);
1376 INIT_LIST_HEAD(&sig
->cpu_timers
[2]);
1379 static inline void posix_cpu_timers_init_group(struct signal_struct
*sig
) { }
1382 static int copy_signal(unsigned long clone_flags
, struct task_struct
*tsk
)
1384 struct signal_struct
*sig
;
1386 if (clone_flags
& CLONE_THREAD
)
1389 sig
= kmem_cache_zalloc(signal_cachep
, GFP_KERNEL
);
1394 sig
->nr_threads
= 1;
1395 atomic_set(&sig
->live
, 1);
1396 atomic_set(&sig
->sigcnt
, 1);
1398 /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */
1399 sig
->thread_head
= (struct list_head
)LIST_HEAD_INIT(tsk
->thread_node
);
1400 tsk
->thread_node
= (struct list_head
)LIST_HEAD_INIT(sig
->thread_head
);
1402 init_waitqueue_head(&sig
->wait_chldexit
);
1403 sig
->curr_target
= tsk
;
1404 init_sigpending(&sig
->shared_pending
);
1405 seqlock_init(&sig
->stats_lock
);
1406 prev_cputime_init(&sig
->prev_cputime
);
1408 #ifdef CONFIG_POSIX_TIMERS
1409 INIT_LIST_HEAD(&sig
->posix_timers
);
1410 hrtimer_init(&sig
->real_timer
, CLOCK_MONOTONIC
, HRTIMER_MODE_REL
);
1411 sig
->real_timer
.function
= it_real_fn
;
1414 task_lock(current
->group_leader
);
1415 memcpy(sig
->rlim
, current
->signal
->rlim
, sizeof sig
->rlim
);
1416 task_unlock(current
->group_leader
);
1418 posix_cpu_timers_init_group(sig
);
1420 tty_audit_fork(sig
);
1421 sched_autogroup_fork(sig
);
1423 sig
->oom_score_adj
= current
->signal
->oom_score_adj
;
1424 sig
->oom_score_adj_min
= current
->signal
->oom_score_adj_min
;
1426 mutex_init(&sig
->cred_guard_mutex
);
1431 static void copy_seccomp(struct task_struct
*p
)
1433 #ifdef CONFIG_SECCOMP
1435 * Must be called with sighand->lock held, which is common to
1436 * all threads in the group. Holding cred_guard_mutex is not
1437 * needed because this new task is not yet running and cannot
1440 assert_spin_locked(¤t
->sighand
->siglock
);
1442 /* Ref-count the new filter user, and assign it. */
1443 get_seccomp_filter(current
);
1444 p
->seccomp
= current
->seccomp
;
1447 * Explicitly enable no_new_privs here in case it got set
1448 * between the task_struct being duplicated and holding the
1449 * sighand lock. The seccomp state and nnp must be in sync.
1451 if (task_no_new_privs(current
))
1452 task_set_no_new_privs(p
);
1455 * If the parent gained a seccomp mode after copying thread
1456 * flags and between before we held the sighand lock, we have
1457 * to manually enable the seccomp thread flag here.
1459 if (p
->seccomp
.mode
!= SECCOMP_MODE_DISABLED
)
1460 set_tsk_thread_flag(p
, TIF_SECCOMP
);
1464 SYSCALL_DEFINE1(set_tid_address
, int __user
*, tidptr
)
1466 current
->clear_child_tid
= tidptr
;
1468 return task_pid_vnr(current
);
1471 static void rt_mutex_init_task(struct task_struct
*p
)
1473 raw_spin_lock_init(&p
->pi_lock
);
1474 #ifdef CONFIG_RT_MUTEXES
1475 p
->pi_waiters
= RB_ROOT
;
1476 p
->pi_waiters_leftmost
= NULL
;
1477 p
->pi_top_task
= NULL
;
1478 p
->pi_blocked_on
= NULL
;
1482 #ifdef CONFIG_POSIX_TIMERS
1484 * Initialize POSIX timer handling for a single task.
1486 static void posix_cpu_timers_init(struct task_struct
*tsk
)
1488 tsk
->cputime_expires
.prof_exp
= 0;
1489 tsk
->cputime_expires
.virt_exp
= 0;
1490 tsk
->cputime_expires
.sched_exp
= 0;
1491 INIT_LIST_HEAD(&tsk
->cpu_timers
[0]);
1492 INIT_LIST_HEAD(&tsk
->cpu_timers
[1]);
1493 INIT_LIST_HEAD(&tsk
->cpu_timers
[2]);
1496 static inline void posix_cpu_timers_init(struct task_struct
*tsk
) { }
1500 init_task_pid(struct task_struct
*task
, enum pid_type type
, struct pid
*pid
)
1502 task
->pids
[type
].pid
= pid
;
1505 static inline void rcu_copy_process(struct task_struct
*p
)
1507 #ifdef CONFIG_PREEMPT_RCU
1508 p
->rcu_read_lock_nesting
= 0;
1509 p
->rcu_read_unlock_special
.s
= 0;
1510 p
->rcu_blocked_node
= NULL
;
1511 INIT_LIST_HEAD(&p
->rcu_node_entry
);
1512 #endif /* #ifdef CONFIG_PREEMPT_RCU */
1513 #ifdef CONFIG_TASKS_RCU
1514 p
->rcu_tasks_holdout
= false;
1515 INIT_LIST_HEAD(&p
->rcu_tasks_holdout_list
);
1516 p
->rcu_tasks_idle_cpu
= -1;
1517 #endif /* #ifdef CONFIG_TASKS_RCU */
1521 * This creates a new process as a copy of the old one,
1522 * but does not actually start it yet.
1524 * It copies the registers, and all the appropriate
1525 * parts of the process environment (as per the clone
1526 * flags). The actual kick-off is left to the caller.
1528 static __latent_entropy
struct task_struct
*copy_process(
1529 unsigned long clone_flags
,
1530 unsigned long stack_start
,
1531 unsigned long stack_size
,
1532 int __user
*child_tidptr
,
1539 struct task_struct
*p
;
1541 if ((clone_flags
& (CLONE_NEWNS
|CLONE_FS
)) == (CLONE_NEWNS
|CLONE_FS
))
1542 return ERR_PTR(-EINVAL
);
1544 if ((clone_flags
& (CLONE_NEWUSER
|CLONE_FS
)) == (CLONE_NEWUSER
|CLONE_FS
))
1545 return ERR_PTR(-EINVAL
);
1547 if ((clone_flags
& CLONE_NEWUSER
) && !unprivileged_userns_clone
)
1548 if (!capable(CAP_SYS_ADMIN
))
1549 return ERR_PTR(-EPERM
);
1552 * Thread groups must share signals as well, and detached threads
1553 * can only be started up within the thread group.
1555 if ((clone_flags
& CLONE_THREAD
) && !(clone_flags
& CLONE_SIGHAND
))
1556 return ERR_PTR(-EINVAL
);
1559 * Shared signal handlers imply shared VM. By way of the above,
1560 * thread groups also imply shared VM. Blocking this case allows
1561 * for various simplifications in other code.
1563 if ((clone_flags
& CLONE_SIGHAND
) && !(clone_flags
& CLONE_VM
))
1564 return ERR_PTR(-EINVAL
);
1567 * Siblings of global init remain as zombies on exit since they are
1568 * not reaped by their parent (swapper). To solve this and to avoid
1569 * multi-rooted process trees, prevent global and container-inits
1570 * from creating siblings.
1572 if ((clone_flags
& CLONE_PARENT
) &&
1573 current
->signal
->flags
& SIGNAL_UNKILLABLE
)
1574 return ERR_PTR(-EINVAL
);
1577 * If the new process will be in a different pid or user namespace
1578 * do not allow it to share a thread group with the forking task.
1580 if (clone_flags
& CLONE_THREAD
) {
1581 if ((clone_flags
& (CLONE_NEWUSER
| CLONE_NEWPID
)) ||
1582 (task_active_pid_ns(current
) !=
1583 current
->nsproxy
->pid_ns_for_children
))
1584 return ERR_PTR(-EINVAL
);
1587 retval
= security_task_create(clone_flags
);
1592 p
= dup_task_struct(current
, node
);
1597 * This _must_ happen before we call free_task(), i.e. before we jump
1598 * to any of the bad_fork_* labels. This is to avoid freeing
1599 * p->set_child_tid which is (ab)used as a kthread's data pointer for
1600 * kernel threads (PF_KTHREAD).
1602 p
->set_child_tid
= (clone_flags
& CLONE_CHILD_SETTID
) ? child_tidptr
: NULL
;
1604 * Clear TID on mm_release()?
1606 p
->clear_child_tid
= (clone_flags
& CLONE_CHILD_CLEARTID
) ? child_tidptr
: NULL
;
1608 ftrace_graph_init_task(p
);
1610 rt_mutex_init_task(p
);
1612 #ifdef CONFIG_PROVE_LOCKING
1613 DEBUG_LOCKS_WARN_ON(!p
->hardirqs_enabled
);
1614 DEBUG_LOCKS_WARN_ON(!p
->softirqs_enabled
);
1617 if (atomic_read(&p
->real_cred
->user
->processes
) >=
1618 task_rlimit(p
, RLIMIT_NPROC
)) {
1619 if (p
->real_cred
->user
!= INIT_USER
&&
1620 !capable(CAP_SYS_RESOURCE
) && !capable(CAP_SYS_ADMIN
))
1623 current
->flags
&= ~PF_NPROC_EXCEEDED
;
1625 retval
= copy_creds(p
, clone_flags
);
1630 * If multiple threads are within copy_process(), then this check
1631 * triggers too late. This doesn't hurt, the check is only there
1632 * to stop root fork bombs.
1635 if (nr_threads
>= max_threads
)
1636 goto bad_fork_cleanup_count
;
1638 delayacct_tsk_init(p
); /* Must remain after dup_task_struct() */
1639 p
->flags
&= ~(PF_SUPERPRIV
| PF_WQ_WORKER
| PF_IDLE
);
1640 p
->flags
|= PF_FORKNOEXEC
;
1641 INIT_LIST_HEAD(&p
->children
);
1642 INIT_LIST_HEAD(&p
->sibling
);
1643 rcu_copy_process(p
);
1644 p
->vfork_done
= NULL
;
1645 spin_lock_init(&p
->alloc_lock
);
1647 init_sigpending(&p
->pending
);
1649 p
->utime
= p
->stime
= p
->gtime
= 0;
1650 #ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
1651 p
->utimescaled
= p
->stimescaled
= 0;
1653 prev_cputime_init(&p
->prev_cputime
);
1655 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1656 seqcount_init(&p
->vtime
.seqcount
);
1657 p
->vtime
.starttime
= 0;
1658 p
->vtime
.state
= VTIME_INACTIVE
;
1661 #if defined(SPLIT_RSS_COUNTING)
1662 memset(&p
->rss_stat
, 0, sizeof(p
->rss_stat
));
1665 p
->default_timer_slack_ns
= current
->timer_slack_ns
;
1667 task_io_accounting_init(&p
->ioac
);
1668 acct_clear_integrals(p
);
1670 posix_cpu_timers_init(p
);
1672 p
->start_time
= ktime_get_ns();
1673 p
->real_start_time
= ktime_get_boot_ns();
1674 p
->io_context
= NULL
;
1675 p
->audit_context
= NULL
;
1678 p
->mempolicy
= mpol_dup(p
->mempolicy
);
1679 if (IS_ERR(p
->mempolicy
)) {
1680 retval
= PTR_ERR(p
->mempolicy
);
1681 p
->mempolicy
= NULL
;
1682 goto bad_fork_cleanup_threadgroup_lock
;
1685 #ifdef CONFIG_CPUSETS
1686 p
->cpuset_mem_spread_rotor
= NUMA_NO_NODE
;
1687 p
->cpuset_slab_spread_rotor
= NUMA_NO_NODE
;
1688 seqcount_init(&p
->mems_allowed_seq
);
1690 #ifdef CONFIG_TRACE_IRQFLAGS
1692 p
->hardirqs_enabled
= 0;
1693 p
->hardirq_enable_ip
= 0;
1694 p
->hardirq_enable_event
= 0;
1695 p
->hardirq_disable_ip
= _THIS_IP_
;
1696 p
->hardirq_disable_event
= 0;
1697 p
->softirqs_enabled
= 1;
1698 p
->softirq_enable_ip
= _THIS_IP_
;
1699 p
->softirq_enable_event
= 0;
1700 p
->softirq_disable_ip
= 0;
1701 p
->softirq_disable_event
= 0;
1702 p
->hardirq_context
= 0;
1703 p
->softirq_context
= 0;
1706 p
->pagefault_disabled
= 0;
1708 #ifdef CONFIG_LOCKDEP
1709 p
->lockdep_depth
= 0; /* no locks held yet */
1710 p
->curr_chain_key
= 0;
1711 p
->lockdep_recursion
= 0;
1714 #ifdef CONFIG_DEBUG_MUTEXES
1715 p
->blocked_on
= NULL
; /* not blocked yet */
1717 #ifdef CONFIG_BCACHE
1718 p
->sequential_io
= 0;
1719 p
->sequential_io_avg
= 0;
1722 /* Perform scheduler related setup. Assign this task to a CPU. */
1723 retval
= sched_fork(clone_flags
, p
);
1725 goto bad_fork_cleanup_policy
;
1727 retval
= perf_event_init_task(p
);
1729 goto bad_fork_cleanup_policy
;
1730 retval
= audit_alloc(p
);
1732 goto bad_fork_cleanup_perf
;
1733 /* copy all the process information */
1735 retval
= security_task_alloc(p
, clone_flags
);
1737 goto bad_fork_cleanup_audit
;
1738 retval
= copy_semundo(clone_flags
, p
);
1740 goto bad_fork_cleanup_security
;
1741 retval
= copy_files(clone_flags
, p
);
1743 goto bad_fork_cleanup_semundo
;
1744 retval
= copy_fs(clone_flags
, p
);
1746 goto bad_fork_cleanup_files
;
1747 retval
= copy_sighand(clone_flags
, p
);
1749 goto bad_fork_cleanup_fs
;
1750 retval
= copy_signal(clone_flags
, p
);
1752 goto bad_fork_cleanup_sighand
;
1753 retval
= copy_mm(clone_flags
, p
);
1755 goto bad_fork_cleanup_signal
;
1756 retval
= copy_namespaces(clone_flags
, p
);
1758 goto bad_fork_cleanup_mm
;
1759 retval
= copy_io(clone_flags
, p
);
1761 goto bad_fork_cleanup_namespaces
;
1762 retval
= copy_thread_tls(clone_flags
, stack_start
, stack_size
, p
, tls
);
1764 goto bad_fork_cleanup_io
;
1766 if (pid
!= &init_struct_pid
) {
1767 pid
= alloc_pid(p
->nsproxy
->pid_ns_for_children
);
1769 retval
= PTR_ERR(pid
);
1770 goto bad_fork_cleanup_thread
;
1778 p
->robust_list
= NULL
;
1779 #ifdef CONFIG_COMPAT
1780 p
->compat_robust_list
= NULL
;
1782 INIT_LIST_HEAD(&p
->pi_state_list
);
1783 p
->pi_state_cache
= NULL
;
1786 * sigaltstack should be cleared when sharing the same VM
1788 if ((clone_flags
& (CLONE_VM
|CLONE_VFORK
)) == CLONE_VM
)
1792 * Syscall tracing and stepping should be turned off in the
1793 * child regardless of CLONE_PTRACE.
1795 user_disable_single_step(p
);
1796 clear_tsk_thread_flag(p
, TIF_SYSCALL_TRACE
);
1797 #ifdef TIF_SYSCALL_EMU
1798 clear_tsk_thread_flag(p
, TIF_SYSCALL_EMU
);
1800 clear_all_latency_tracing(p
);
1802 /* ok, now we should be set up.. */
1803 p
->pid
= pid_nr(pid
);
1804 if (clone_flags
& CLONE_THREAD
) {
1805 p
->exit_signal
= -1;
1806 p
->group_leader
= current
->group_leader
;
1807 p
->tgid
= current
->tgid
;
1809 if (clone_flags
& CLONE_PARENT
)
1810 p
->exit_signal
= current
->group_leader
->exit_signal
;
1812 p
->exit_signal
= (clone_flags
& CSIGNAL
);
1813 p
->group_leader
= p
;
1818 p
->nr_dirtied_pause
= 128 >> (PAGE_SHIFT
- 10);
1819 p
->dirty_paused_when
= 0;
1821 p
->pdeath_signal
= 0;
1822 INIT_LIST_HEAD(&p
->thread_group
);
1823 p
->task_works
= NULL
;
1825 cgroup_threadgroup_change_begin(current
);
1827 * Ensure that the cgroup subsystem policies allow the new process to be
1828 * forked. It should be noted the the new process's css_set can be changed
1829 * between here and cgroup_post_fork() if an organisation operation is in
1832 retval
= cgroup_can_fork(p
);
1834 goto bad_fork_free_pid
;
1837 * Make it visible to the rest of the system, but dont wake it up yet.
1838 * Need tasklist lock for parent etc handling!
1840 write_lock_irq(&tasklist_lock
);
1842 /* CLONE_PARENT re-uses the old parent */
1843 if (clone_flags
& (CLONE_PARENT
|CLONE_THREAD
)) {
1844 p
->real_parent
= current
->real_parent
;
1845 p
->parent_exec_id
= current
->parent_exec_id
;
1847 p
->real_parent
= current
;
1848 p
->parent_exec_id
= current
->self_exec_id
;
1851 klp_copy_process(p
);
1853 spin_lock(¤t
->sighand
->siglock
);
1856 * Copy seccomp details explicitly here, in case they were changed
1857 * before holding sighand lock.
1862 * Process group and session signals need to be delivered to just the
1863 * parent before the fork or both the parent and the child after the
1864 * fork. Restart if a signal comes in before we add the new process to
1865 * it's process group.
1866 * A fatal signal pending means that current will exit, so the new
1867 * thread can't slip out of an OOM kill (or normal SIGKILL).
1869 recalc_sigpending();
1870 if (signal_pending(current
)) {
1871 retval
= -ERESTARTNOINTR
;
1872 goto bad_fork_cancel_cgroup
;
1874 if (unlikely(!(ns_of_pid(pid
)->nr_hashed
& PIDNS_HASH_ADDING
))) {
1876 goto bad_fork_cancel_cgroup
;
1879 if (likely(p
->pid
)) {
1880 ptrace_init_task(p
, (clone_flags
& CLONE_PTRACE
) || trace
);
1882 init_task_pid(p
, PIDTYPE_PID
, pid
);
1883 if (thread_group_leader(p
)) {
1884 init_task_pid(p
, PIDTYPE_PGID
, task_pgrp(current
));
1885 init_task_pid(p
, PIDTYPE_SID
, task_session(current
));
1887 if (is_child_reaper(pid
)) {
1888 ns_of_pid(pid
)->child_reaper
= p
;
1889 p
->signal
->flags
|= SIGNAL_UNKILLABLE
;
1892 p
->signal
->leader_pid
= pid
;
1893 p
->signal
->tty
= tty_kref_get(current
->signal
->tty
);
1895 * Inherit has_child_subreaper flag under the same
1896 * tasklist_lock with adding child to the process tree
1897 * for propagate_has_child_subreaper optimization.
1899 p
->signal
->has_child_subreaper
= p
->real_parent
->signal
->has_child_subreaper
||
1900 p
->real_parent
->signal
->is_child_subreaper
;
1901 list_add_tail(&p
->sibling
, &p
->real_parent
->children
);
1902 list_add_tail_rcu(&p
->tasks
, &init_task
.tasks
);
1903 attach_pid(p
, PIDTYPE_PGID
);
1904 attach_pid(p
, PIDTYPE_SID
);
1905 __this_cpu_inc(process_counts
);
1907 current
->signal
->nr_threads
++;
1908 atomic_inc(¤t
->signal
->live
);
1909 atomic_inc(¤t
->signal
->sigcnt
);
1910 list_add_tail_rcu(&p
->thread_group
,
1911 &p
->group_leader
->thread_group
);
1912 list_add_tail_rcu(&p
->thread_node
,
1913 &p
->signal
->thread_head
);
1915 attach_pid(p
, PIDTYPE_PID
);
1920 spin_unlock(¤t
->sighand
->siglock
);
1921 syscall_tracepoint_update(p
);
1922 write_unlock_irq(&tasklist_lock
);
1924 proc_fork_connector(p
);
1925 cgroup_post_fork(p
);
1926 cgroup_threadgroup_change_end(current
);
1929 trace_task_newtask(p
, clone_flags
);
1930 uprobe_copy_process(p
, clone_flags
);
1934 bad_fork_cancel_cgroup
:
1935 spin_unlock(¤t
->sighand
->siglock
);
1936 write_unlock_irq(&tasklist_lock
);
1937 cgroup_cancel_fork(p
);
1939 cgroup_threadgroup_change_end(current
);
1940 if (pid
!= &init_struct_pid
)
1942 bad_fork_cleanup_thread
:
1944 bad_fork_cleanup_io
:
1947 bad_fork_cleanup_namespaces
:
1948 exit_task_namespaces(p
);
1949 bad_fork_cleanup_mm
:
1952 bad_fork_cleanup_signal
:
1953 if (!(clone_flags
& CLONE_THREAD
))
1954 free_signal_struct(p
->signal
);
1955 bad_fork_cleanup_sighand
:
1956 __cleanup_sighand(p
->sighand
);
1957 bad_fork_cleanup_fs
:
1958 exit_fs(p
); /* blocking */
1959 bad_fork_cleanup_files
:
1960 exit_files(p
); /* blocking */
1961 bad_fork_cleanup_semundo
:
1963 bad_fork_cleanup_security
:
1964 security_task_free(p
);
1965 bad_fork_cleanup_audit
:
1967 bad_fork_cleanup_perf
:
1968 perf_event_free_task(p
);
1969 bad_fork_cleanup_policy
:
1971 mpol_put(p
->mempolicy
);
1972 bad_fork_cleanup_threadgroup_lock
:
1974 delayacct_tsk_free(p
);
1975 bad_fork_cleanup_count
:
1976 atomic_dec(&p
->cred
->user
->processes
);
1979 p
->state
= TASK_DEAD
;
1983 return ERR_PTR(retval
);
1986 static inline void init_idle_pids(struct pid_link
*links
)
1990 for (type
= PIDTYPE_PID
; type
< PIDTYPE_MAX
; ++type
) {
1991 INIT_HLIST_NODE(&links
[type
].node
); /* not really needed */
1992 links
[type
].pid
= &init_struct_pid
;
1996 struct task_struct
*fork_idle(int cpu
)
1998 struct task_struct
*task
;
1999 task
= copy_process(CLONE_VM
, 0, 0, NULL
, &init_struct_pid
, 0, 0,
2001 if (!IS_ERR(task
)) {
2002 init_idle_pids(task
->pids
);
2003 init_idle(task
, cpu
);
2010 * Ok, this is the main fork-routine.
2012 * It copies the process, and if successful kick-starts
2013 * it and waits for it to finish using the VM if required.
2015 long _do_fork(unsigned long clone_flags
,
2016 unsigned long stack_start
,
2017 unsigned long stack_size
,
2018 int __user
*parent_tidptr
,
2019 int __user
*child_tidptr
,
2022 struct task_struct
*p
;
2027 * Determine whether and which event to report to ptracer. When
2028 * called from kernel_thread or CLONE_UNTRACED is explicitly
2029 * requested, no event is reported; otherwise, report if the event
2030 * for the type of forking is enabled.
2032 if (!(clone_flags
& CLONE_UNTRACED
)) {
2033 if (clone_flags
& CLONE_VFORK
)
2034 trace
= PTRACE_EVENT_VFORK
;
2035 else if ((clone_flags
& CSIGNAL
) != SIGCHLD
)
2036 trace
= PTRACE_EVENT_CLONE
;
2038 trace
= PTRACE_EVENT_FORK
;
2040 if (likely(!ptrace_event_enabled(current
, trace
)))
2044 p
= copy_process(clone_flags
, stack_start
, stack_size
,
2045 child_tidptr
, NULL
, trace
, tls
, NUMA_NO_NODE
);
2046 add_latent_entropy();
2048 * Do this prior waking up the new thread - the thread pointer
2049 * might get invalid after that point, if the thread exits quickly.
2052 struct completion vfork
;
2055 trace_sched_process_fork(current
, p
);
2057 pid
= get_task_pid(p
, PIDTYPE_PID
);
2060 if (clone_flags
& CLONE_PARENT_SETTID
)
2061 put_user(nr
, parent_tidptr
);
2063 if (clone_flags
& CLONE_VFORK
) {
2064 p
->vfork_done
= &vfork
;
2065 init_completion(&vfork
);
2069 wake_up_new_task(p
);
2071 /* forking complete and child started to run, tell ptracer */
2072 if (unlikely(trace
))
2073 ptrace_event_pid(trace
, pid
);
2075 if (clone_flags
& CLONE_VFORK
) {
2076 if (!wait_for_vfork_done(p
, &vfork
))
2077 ptrace_event_pid(PTRACE_EVENT_VFORK_DONE
, pid
);
2087 #ifndef CONFIG_HAVE_COPY_THREAD_TLS
2088 /* For compatibility with architectures that call do_fork directly rather than
2089 * using the syscall entry points below. */
2090 long do_fork(unsigned long clone_flags
,
2091 unsigned long stack_start
,
2092 unsigned long stack_size
,
2093 int __user
*parent_tidptr
,
2094 int __user
*child_tidptr
)
2096 return _do_fork(clone_flags
, stack_start
, stack_size
,
2097 parent_tidptr
, child_tidptr
, 0);
2102 * Create a kernel thread.
2104 pid_t
kernel_thread(int (*fn
)(void *), void *arg
, unsigned long flags
)
2106 return _do_fork(flags
|CLONE_VM
|CLONE_UNTRACED
, (unsigned long)fn
,
2107 (unsigned long)arg
, NULL
, NULL
, 0);
2110 #ifdef __ARCH_WANT_SYS_FORK
2111 SYSCALL_DEFINE0(fork
)
2114 return _do_fork(SIGCHLD
, 0, 0, NULL
, NULL
, 0);
2116 /* can not support in nommu mode */
2122 #ifdef __ARCH_WANT_SYS_VFORK
2123 SYSCALL_DEFINE0(vfork
)
2125 return _do_fork(CLONE_VFORK
| CLONE_VM
| SIGCHLD
, 0,
2130 #ifdef __ARCH_WANT_SYS_CLONE
2131 #ifdef CONFIG_CLONE_BACKWARDS
2132 SYSCALL_DEFINE5(clone
, unsigned long, clone_flags
, unsigned long, newsp
,
2133 int __user
*, parent_tidptr
,
2135 int __user
*, child_tidptr
)
2136 #elif defined(CONFIG_CLONE_BACKWARDS2)
2137 SYSCALL_DEFINE5(clone
, unsigned long, newsp
, unsigned long, clone_flags
,
2138 int __user
*, parent_tidptr
,
2139 int __user
*, child_tidptr
,
2141 #elif defined(CONFIG_CLONE_BACKWARDS3)
2142 SYSCALL_DEFINE6(clone
, unsigned long, clone_flags
, unsigned long, newsp
,
2144 int __user
*, parent_tidptr
,
2145 int __user
*, child_tidptr
,
2148 SYSCALL_DEFINE5(clone
, unsigned long, clone_flags
, unsigned long, newsp
,
2149 int __user
*, parent_tidptr
,
2150 int __user
*, child_tidptr
,
2154 return _do_fork(clone_flags
, newsp
, 0, parent_tidptr
, child_tidptr
, tls
);
2158 void walk_process_tree(struct task_struct
*top
, proc_visitor visitor
, void *data
)
2160 struct task_struct
*leader
, *parent
, *child
;
2163 read_lock(&tasklist_lock
);
2164 leader
= top
= top
->group_leader
;
2166 for_each_thread(leader
, parent
) {
2167 list_for_each_entry(child
, &parent
->children
, sibling
) {
2168 res
= visitor(child
, data
);
2180 if (leader
!= top
) {
2182 parent
= child
->real_parent
;
2183 leader
= parent
->group_leader
;
2187 read_unlock(&tasklist_lock
);
2190 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
2191 #define ARCH_MIN_MMSTRUCT_ALIGN 0
2194 static void sighand_ctor(void *data
)
2196 struct sighand_struct
*sighand
= data
;
2198 spin_lock_init(&sighand
->siglock
);
2199 init_waitqueue_head(&sighand
->signalfd_wqh
);
2202 void __init
proc_caches_init(void)
2204 sighand_cachep
= kmem_cache_create("sighand_cache",
2205 sizeof(struct sighand_struct
), 0,
2206 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_TYPESAFE_BY_RCU
|
2207 SLAB_NOTRACK
|SLAB_ACCOUNT
, sighand_ctor
);
2208 signal_cachep
= kmem_cache_create("signal_cache",
2209 sizeof(struct signal_struct
), 0,
2210 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
|SLAB_ACCOUNT
,
2212 files_cachep
= kmem_cache_create("files_cache",
2213 sizeof(struct files_struct
), 0,
2214 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
|SLAB_ACCOUNT
,
2216 fs_cachep
= kmem_cache_create("fs_cache",
2217 sizeof(struct fs_struct
), 0,
2218 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
|SLAB_ACCOUNT
,
2221 * FIXME! The "sizeof(struct mm_struct)" currently includes the
2222 * whole struct cpumask for the OFFSTACK case. We could change
2223 * this to *only* allocate as much of it as required by the
2224 * maximum number of CPU's we can ever have. The cpumask_allocation
2225 * is at the end of the structure, exactly for that reason.
2227 mm_cachep
= kmem_cache_create("mm_struct",
2228 sizeof(struct mm_struct
), ARCH_MIN_MMSTRUCT_ALIGN
,
2229 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
|SLAB_NOTRACK
|SLAB_ACCOUNT
,
2231 vm_area_cachep
= KMEM_CACHE(vm_area_struct
, SLAB_PANIC
|SLAB_ACCOUNT
);
2233 nsproxy_cache_init();
2237 * Check constraints on flags passed to the unshare system call.
2239 static int check_unshare_flags(unsigned long unshare_flags
)
2241 if (unshare_flags
& ~(CLONE_THREAD
|CLONE_FS
|CLONE_NEWNS
|CLONE_SIGHAND
|
2242 CLONE_VM
|CLONE_FILES
|CLONE_SYSVSEM
|
2243 CLONE_NEWUTS
|CLONE_NEWIPC
|CLONE_NEWNET
|
2244 CLONE_NEWUSER
|CLONE_NEWPID
|CLONE_NEWCGROUP
))
2247 * Not implemented, but pretend it works if there is nothing
2248 * to unshare. Note that unsharing the address space or the
2249 * signal handlers also need to unshare the signal queues (aka
2252 if (unshare_flags
& (CLONE_THREAD
| CLONE_SIGHAND
| CLONE_VM
)) {
2253 if (!thread_group_empty(current
))
2256 if (unshare_flags
& (CLONE_SIGHAND
| CLONE_VM
)) {
2257 if (atomic_read(¤t
->sighand
->count
) > 1)
2260 if (unshare_flags
& CLONE_VM
) {
2261 if (!current_is_single_threaded())
2269 * Unshare the filesystem structure if it is being shared
2271 static int unshare_fs(unsigned long unshare_flags
, struct fs_struct
**new_fsp
)
2273 struct fs_struct
*fs
= current
->fs
;
2275 if (!(unshare_flags
& CLONE_FS
) || !fs
)
2278 /* don't need lock here; in the worst case we'll do useless copy */
2282 *new_fsp
= copy_fs_struct(fs
);
2290 * Unshare file descriptor table if it is being shared
2292 static int unshare_fd(unsigned long unshare_flags
, struct files_struct
**new_fdp
)
2294 struct files_struct
*fd
= current
->files
;
2297 if ((unshare_flags
& CLONE_FILES
) &&
2298 (fd
&& atomic_read(&fd
->count
) > 1)) {
2299 *new_fdp
= dup_fd(fd
, &error
);
2308 * unshare allows a process to 'unshare' part of the process
2309 * context which was originally shared using clone. copy_*
2310 * functions used by do_fork() cannot be used here directly
2311 * because they modify an inactive task_struct that is being
2312 * constructed. Here we are modifying the current, active,
2315 SYSCALL_DEFINE1(unshare
, unsigned long, unshare_flags
)
2317 struct fs_struct
*fs
, *new_fs
= NULL
;
2318 struct files_struct
*fd
, *new_fd
= NULL
;
2319 struct cred
*new_cred
= NULL
;
2320 struct nsproxy
*new_nsproxy
= NULL
;
2325 * If unsharing a user namespace must also unshare the thread group
2326 * and unshare the filesystem root and working directories.
2328 if (unshare_flags
& CLONE_NEWUSER
)
2329 unshare_flags
|= CLONE_THREAD
| CLONE_FS
;
2331 * If unsharing vm, must also unshare signal handlers.
2333 if (unshare_flags
& CLONE_VM
)
2334 unshare_flags
|= CLONE_SIGHAND
;
2336 * If unsharing a signal handlers, must also unshare the signal queues.
2338 if (unshare_flags
& CLONE_SIGHAND
)
2339 unshare_flags
|= CLONE_THREAD
;
2341 * If unsharing namespace, must also unshare filesystem information.
2343 if (unshare_flags
& CLONE_NEWNS
)
2344 unshare_flags
|= CLONE_FS
;
2346 if ((unshare_flags
& CLONE_NEWUSER
) && !unprivileged_userns_clone
) {
2348 if (!capable(CAP_SYS_ADMIN
))
2349 goto bad_unshare_out
;
2352 err
= check_unshare_flags(unshare_flags
);
2354 goto bad_unshare_out
;
2356 * CLONE_NEWIPC must also detach from the undolist: after switching
2357 * to a new ipc namespace, the semaphore arrays from the old
2358 * namespace are unreachable.
2360 if (unshare_flags
& (CLONE_NEWIPC
|CLONE_SYSVSEM
))
2362 err
= unshare_fs(unshare_flags
, &new_fs
);
2364 goto bad_unshare_out
;
2365 err
= unshare_fd(unshare_flags
, &new_fd
);
2367 goto bad_unshare_cleanup_fs
;
2368 err
= unshare_userns(unshare_flags
, &new_cred
);
2370 goto bad_unshare_cleanup_fd
;
2371 err
= unshare_nsproxy_namespaces(unshare_flags
, &new_nsproxy
,
2374 goto bad_unshare_cleanup_cred
;
2376 if (new_fs
|| new_fd
|| do_sysvsem
|| new_cred
|| new_nsproxy
) {
2379 * CLONE_SYSVSEM is equivalent to sys_exit().
2383 if (unshare_flags
& CLONE_NEWIPC
) {
2384 /* Orphan segments in old ns (see sem above). */
2386 shm_init_task(current
);
2390 switch_task_namespaces(current
, new_nsproxy
);
2396 spin_lock(&fs
->lock
);
2397 current
->fs
= new_fs
;
2402 spin_unlock(&fs
->lock
);
2406 fd
= current
->files
;
2407 current
->files
= new_fd
;
2411 task_unlock(current
);
2414 /* Install the new user namespace */
2415 commit_creds(new_cred
);
2420 perf_event_namespaces(current
);
2422 bad_unshare_cleanup_cred
:
2425 bad_unshare_cleanup_fd
:
2427 put_files_struct(new_fd
);
2429 bad_unshare_cleanup_fs
:
2431 free_fs_struct(new_fs
);
2438 * Helper to unshare the files of the current task.
2439 * We don't want to expose copy_files internals to
2440 * the exec layer of the kernel.
2443 int unshare_files(struct files_struct
**displaced
)
2445 struct task_struct
*task
= current
;
2446 struct files_struct
*copy
= NULL
;
2449 error
= unshare_fd(CLONE_FILES
, ©
);
2450 if (error
|| !copy
) {
2454 *displaced
= task
->files
;
2461 int sysctl_max_threads(struct ctl_table
*table
, int write
,
2462 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
2466 int threads
= max_threads
;
2467 int min
= MIN_THREADS
;
2468 int max
= MAX_THREADS
;
2475 ret
= proc_dointvec_minmax(&t
, write
, buffer
, lenp
, ppos
);
2479 set_max_threads(threads
);