~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/kernel/fork.c

Version: ~ [ linux-5.14-rc1 ] ~ [ linux-5.13.1 ] ~ [ linux-5.12.16 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.49 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.131 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.197 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.239 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.275 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.275 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  *  linux/kernel/fork.c
  3  *
  4  *  Copyright (C) 1991, 1992  Linus Torvalds
  5  */
  6 
  7 /*
  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()'
 12  */
 13 
 14 #include <linux/slab.h>
 15 #include <linux/init.h>
 16 #include <linux/unistd.h>
 17 #include <linux/module.h>
 18 #include <linux/vmalloc.h>
 19 #include <linux/completion.h>
 20 #include <linux/personality.h>
 21 #include <linux/mempolicy.h>
 22 #include <linux/sem.h>
 23 #include <linux/file.h>
 24 #include <linux/fdtable.h>
 25 #include <linux/iocontext.h>
 26 #include <linux/key.h>
 27 #include <linux/binfmts.h>
 28 #include <linux/mman.h>
 29 #include <linux/mmu_notifier.h>
 30 #include <linux/fs.h>
 31 #include <linux/mm.h>
 32 #include <linux/vmacache.h>
 33 #include <linux/nsproxy.h>
 34 #include <linux/capability.h>
 35 #include <linux/cpu.h>
 36 #include <linux/cgroup.h>
 37 #include <linux/security.h>
 38 #include <linux/hugetlb.h>
 39 #include <linux/seccomp.h>
 40 #include <linux/swap.h>
 41 #include <linux/syscalls.h>
 42 #include <linux/jiffies.h>
 43 #include <linux/futex.h>
 44 #include <linux/compat.h>
 45 #include <linux/kthread.h>
 46 #include <linux/task_io_accounting_ops.h>
 47 #include <linux/rcupdate.h>
 48 #include <linux/ptrace.h>
 49 #include <linux/mount.h>
 50 #include <linux/audit.h>
 51 #include <linux/memcontrol.h>
 52 #include <linux/ftrace.h>
 53 #include <linux/proc_fs.h>
 54 #include <linux/profile.h>
 55 #include <linux/rmap.h>
 56 #include <linux/ksm.h>
 57 #include <linux/acct.h>
 58 #include <linux/tsacct_kern.h>
 59 #include <linux/cn_proc.h>
 60 #include <linux/freezer.h>
 61 #include <linux/delayacct.h>
 62 #include <linux/taskstats_kern.h>
 63 #include <linux/random.h>
 64 #include <linux/tty.h>
 65 #include <linux/blkdev.h>
 66 #include <linux/fs_struct.h>
 67 #include <linux/magic.h>
 68 #include <linux/perf_event.h>
 69 #include <linux/posix-timers.h>
 70 #include <linux/user-return-notifier.h>
 71 #include <linux/oom.h>
 72 #include <linux/khugepaged.h>
 73 #include <linux/signalfd.h>
 74 #include <linux/uprobes.h>
 75 #include <linux/aio.h>
 76 #include <linux/compiler.h>
 77 #include <linux/sysctl.h>
 78 #include <linux/kcov.h>
 79 
 80 #include <asm/pgtable.h>
 81 #include <asm/pgalloc.h>
 82 #include <asm/uaccess.h>
 83 #include <asm/mmu_context.h>
 84 #include <asm/cacheflush.h>
 85 #include <asm/tlbflush.h>
 86 
 87 #include <trace/events/sched.h>
 88 
 89 #define CREATE_TRACE_POINTS
 90 #include <trace/events/task.h>
 91 
 92 /*
 93  * Minimum number of threads to boot the kernel
 94  */
 95 #define MIN_THREADS 20
 96 
 97 /*
 98  * Maximum number of threads
 99  */
100 #define MAX_THREADS FUTEX_TID_MASK
101 
102 /*
103  * Protected counters by write_lock_irq(&tasklist_lock)
104  */
105 unsigned long total_forks;      /* Handle normal Linux uptimes. */
106 int nr_threads;                 /* The idle threads do not count.. */
107 
108 int max_threads;                /* tunable limit on nr_threads */
109 
110 DEFINE_PER_CPU(unsigned long, process_counts) = 0;
111 
112 __cacheline_aligned DEFINE_RWLOCK(tasklist_lock);  /* outer */
113 
114 #ifdef CONFIG_PROVE_RCU
115 int lockdep_tasklist_lock_is_held(void)
116 {
117         return lockdep_is_held(&tasklist_lock);
118 }
119 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
120 #endif /* #ifdef CONFIG_PROVE_RCU */
121 
122 int nr_processes(void)
123 {
124         int cpu;
125         int total = 0;
126 
127         for_each_possible_cpu(cpu)
128                 total += per_cpu(process_counts, cpu);
129 
130         return total;
131 }
132 
133 void __weak arch_release_task_struct(struct task_struct *tsk)
134 {
135 }
136 
137 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
138 static struct kmem_cache *task_struct_cachep;
139 
140 static inline struct task_struct *alloc_task_struct_node(int node)
141 {
142         return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node);
143 }
144 
145 static inline void free_task_struct(struct task_struct *tsk)
146 {
147         kmem_cache_free(task_struct_cachep, tsk);
148 }
149 #endif
150 
151 void __weak arch_release_thread_stack(unsigned long *stack)
152 {
153 }
154 
155 #ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR
156 
157 /*
158  * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a
159  * kmemcache based allocator.
160  */
161 # if THREAD_SIZE >= PAGE_SIZE
162 static unsigned long *alloc_thread_stack_node(struct task_struct *tsk,
163                                                   int node)
164 {
165         struct page *page = alloc_kmem_pages_node(node, THREADINFO_GFP,
166                                                   THREAD_SIZE_ORDER);
167 
168         if (page)
169                 memcg_kmem_update_page_stat(page, MEMCG_KERNEL_STACK,
170                                             1 << THREAD_SIZE_ORDER);
171 
172         return page ? page_address(page) : NULL;
173 }
174 
175 static inline void free_thread_stack(unsigned long *stack)
176 {
177         struct page *page = virt_to_page(stack);
178 
179         memcg_kmem_update_page_stat(page, MEMCG_KERNEL_STACK,
180                                     -(1 << THREAD_SIZE_ORDER));
181         __free_kmem_pages(page, THREAD_SIZE_ORDER);
182 }
183 # else
184 static struct kmem_cache *thread_stack_cache;
185 
186 static unsigned long *alloc_thread_stack_node(struct task_struct *tsk,
187                                                   int node)
188 {
189         return kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node);
190 }
191 
192 static void free_thread_stack(unsigned long *stack)
193 {
194         kmem_cache_free(thread_stack_cache, stack);
195 }
196 
197 void thread_stack_cache_init(void)
198 {
199         thread_stack_cache = kmem_cache_create("thread_stack", THREAD_SIZE,
200                                               THREAD_SIZE, 0, NULL);
201         BUG_ON(thread_stack_cache == NULL);
202 }
203 # endif
204 #endif
205 
206 /* SLAB cache for signal_struct structures (tsk->signal) */
207 static struct kmem_cache *signal_cachep;
208 
209 /* SLAB cache for sighand_struct structures (tsk->sighand) */
210 struct kmem_cache *sighand_cachep;
211 
212 /* SLAB cache for files_struct structures (tsk->files) */
213 struct kmem_cache *files_cachep;
214 
215 /* SLAB cache for fs_struct structures (tsk->fs) */
216 struct kmem_cache *fs_cachep;
217 
218 /* SLAB cache for vm_area_struct structures */
219 struct kmem_cache *vm_area_cachep;
220 
221 /* SLAB cache for mm_struct structures (tsk->mm) */
222 static struct kmem_cache *mm_cachep;
223 
224 static void account_kernel_stack(unsigned long *stack, int account)
225 {
226         struct zone *zone = page_zone(virt_to_page(stack));
227 
228         mod_zone_page_state(zone, NR_KERNEL_STACK, account);
229 }
230 
231 void free_task(struct task_struct *tsk)
232 {
233         account_kernel_stack(tsk->stack, -1);
234         arch_release_thread_stack(tsk->stack);
235         free_thread_stack(tsk->stack);
236         rt_mutex_debug_task_free(tsk);
237         ftrace_graph_exit_task(tsk);
238         put_seccomp_filter(tsk);
239         arch_release_task_struct(tsk);
240         free_task_struct(tsk);
241 }
242 EXPORT_SYMBOL(free_task);
243 
244 static inline void free_signal_struct(struct signal_struct *sig)
245 {
246         taskstats_tgid_free(sig);
247         sched_autogroup_exit(sig);
248         kmem_cache_free(signal_cachep, sig);
249 }
250 
251 static inline void put_signal_struct(struct signal_struct *sig)
252 {
253         if (atomic_dec_and_test(&sig->sigcnt))
254                 free_signal_struct(sig);
255 }
256 
257 void __put_task_struct(struct task_struct *tsk)
258 {
259         WARN_ON(!tsk->exit_state);
260         WARN_ON(atomic_read(&tsk->usage));
261         WARN_ON(tsk == current);
262 
263         cgroup_free(tsk);
264         task_numa_free(tsk);
265         security_task_free(tsk);
266         exit_creds(tsk);
267         delayacct_tsk_free(tsk);
268         put_signal_struct(tsk->signal);
269 
270         ccs_free_task_security(tsk);
271         if (!profile_handoff_task(tsk))
272                 free_task(tsk);
273 }
274 EXPORT_SYMBOL_GPL(__put_task_struct);
275 
276 void __init __weak arch_task_cache_init(void) { }
277 
278 /*
279  * set_max_threads
280  */
281 static void set_max_threads(unsigned int max_threads_suggested)
282 {
283         u64 threads;
284 
285         /*
286          * The number of threads shall be limited such that the thread
287          * structures may only consume a small part of the available memory.
288          */
289         if (fls64(totalram_pages) + fls64(PAGE_SIZE) > 64)
290                 threads = MAX_THREADS;
291         else
292                 threads = div64_u64((u64) totalram_pages * (u64) PAGE_SIZE,
293                                     (u64) THREAD_SIZE * 8UL);
294 
295         if (threads > max_threads_suggested)
296                 threads = max_threads_suggested;
297 
298         max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS);
299 }
300 
301 #ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
302 /* Initialized by the architecture: */
303 int arch_task_struct_size __read_mostly;
304 #endif
305 
306 void __init fork_init(void)
307 {
308 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
309 #ifndef ARCH_MIN_TASKALIGN
310 #define ARCH_MIN_TASKALIGN      L1_CACHE_BYTES
311 #endif
312         /* create a slab on which task_structs can be allocated */
313         task_struct_cachep = kmem_cache_create("task_struct",
314                         arch_task_struct_size, ARCH_MIN_TASKALIGN,
315                         SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, NULL);
316 #endif
317 
318         /* do the arch specific task caches init */
319         arch_task_cache_init();
320 
321         set_max_threads(MAX_THREADS);
322 
323         init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
324         init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
325         init_task.signal->rlim[RLIMIT_SIGPENDING] =
326                 init_task.signal->rlim[RLIMIT_NPROC];
327 }
328 
329 int __weak arch_dup_task_struct(struct task_struct *dst,
330                                                struct task_struct *src)
331 {
332         *dst = *src;
333         return 0;
334 }
335 
336 void set_task_stack_end_magic(struct task_struct *tsk)
337 {
338         unsigned long *stackend;
339 
340         stackend = end_of_stack(tsk);
341         *stackend = STACK_END_MAGIC;    /* for overflow detection */
342 }
343 
344 static struct task_struct *dup_task_struct(struct task_struct *orig, int node)
345 {
346         struct task_struct *tsk;
347         unsigned long *stack;
348         int err;
349 
350         if (node == NUMA_NO_NODE)
351                 node = tsk_fork_get_node(orig);
352         tsk = alloc_task_struct_node(node);
353         if (!tsk)
354                 return NULL;
355 
356         stack = alloc_thread_stack_node(tsk, node);
357         if (!stack)
358                 goto free_tsk;
359 
360         err = arch_dup_task_struct(tsk, orig);
361         if (err)
362                 goto free_stack;
363 
364         tsk->stack = stack;
365 #ifdef CONFIG_SECCOMP
366         /*
367          * We must handle setting up seccomp filters once we're under
368          * the sighand lock in case orig has changed between now and
369          * then. Until then, filter must be NULL to avoid messing up
370          * the usage counts on the error path calling free_task.
371          */
372         tsk->seccomp.filter = NULL;
373 #endif
374 
375         setup_thread_stack(tsk, orig);
376         clear_user_return_notifier(tsk);
377         clear_tsk_need_resched(tsk);
378         set_task_stack_end_magic(tsk);
379 
380 #ifdef CONFIG_CC_STACKPROTECTOR
381         tsk->stack_canary = get_random_int();
382 #endif
383 
384         /*
385          * One for us, one for whoever does the "release_task()" (usually
386          * parent)
387          */
388         atomic_set(&tsk->usage, 2);
389 #ifdef CONFIG_BLK_DEV_IO_TRACE
390         tsk->btrace_seq = 0;
391 #endif
392         tsk->splice_pipe = NULL;
393         tsk->task_frag.page = NULL;
394         tsk->wake_q.next = NULL;
395 
396         account_kernel_stack(stack, 1);
397 
398         kcov_task_init(tsk);
399 
400         return tsk;
401 
402 free_stack:
403         free_thread_stack(stack);
404 free_tsk:
405         free_task_struct(tsk);
406         return NULL;
407 }
408 
409 #ifdef CONFIG_MMU
410 static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
411 {
412         struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
413         struct rb_node **rb_link, *rb_parent;
414         int retval;
415         unsigned long charge;
416 
417         uprobe_start_dup_mmap();
418         if (down_write_killable(&oldmm->mmap_sem)) {
419                 retval = -EINTR;
420                 goto fail_uprobe_end;
421         }
422         flush_cache_dup_mm(oldmm);
423         uprobe_dup_mmap(oldmm, mm);
424         /*
425          * Not linked in yet - no deadlock potential:
426          */
427         down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
428 
429         /* No ordering required: file already has been exposed. */
430         RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
431 
432         mm->total_vm = oldmm->total_vm;
433         mm->data_vm = oldmm->data_vm;
434         mm->exec_vm = oldmm->exec_vm;
435         mm->stack_vm = oldmm->stack_vm;
436 
437         rb_link = &mm->mm_rb.rb_node;
438         rb_parent = NULL;
439         pprev = &mm->mmap;
440         retval = ksm_fork(mm, oldmm);
441         if (retval)
442                 goto out;
443         retval = khugepaged_fork(mm, oldmm);
444         if (retval)
445                 goto out;
446 
447         prev = NULL;
448         for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
449                 struct file *file;
450 
451                 if (mpnt->vm_flags & VM_DONTCOPY) {
452                         vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt));
453                         continue;
454                 }
455                 charge = 0;
456                 if (mpnt->vm_flags & VM_ACCOUNT) {
457                         unsigned long len = vma_pages(mpnt);
458 
459                         if (security_vm_enough_memory_mm(oldmm, len)) /* sic */
460                                 goto fail_nomem;
461                         charge = len;
462                 }
463                 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
464                 if (!tmp)
465                         goto fail_nomem;
466                 *tmp = *mpnt;
467                 INIT_LIST_HEAD(&tmp->anon_vma_chain);
468                 retval = vma_dup_policy(mpnt, tmp);
469                 if (retval)
470                         goto fail_nomem_policy;
471                 tmp->vm_mm = mm;
472                 if (anon_vma_fork(tmp, mpnt))
473                         goto fail_nomem_anon_vma_fork;
474                 tmp->vm_flags &=
475                         ~(VM_LOCKED|VM_LOCKONFAULT|VM_UFFD_MISSING|VM_UFFD_WP);
476                 tmp->vm_next = tmp->vm_prev = NULL;
477                 tmp->vm_userfaultfd_ctx = NULL_VM_UFFD_CTX;
478                 file = tmp->vm_file;
479                 if (file) {
480                         struct inode *inode = file_inode(file);
481                         struct address_space *mapping = file->f_mapping;
482 
483                         get_file(file);
484                         if (tmp->vm_flags & VM_DENYWRITE)
485                                 atomic_dec(&inode->i_writecount);
486                         i_mmap_lock_write(mapping);
487                         if (tmp->vm_flags & VM_SHARED)
488                                 atomic_inc(&mapping->i_mmap_writable);
489                         flush_dcache_mmap_lock(mapping);
490                         /* insert tmp into the share list, just after mpnt */
491                         vma_interval_tree_insert_after(tmp, mpnt,
492                                         &mapping->i_mmap);
493                         flush_dcache_mmap_unlock(mapping);
494                         i_mmap_unlock_write(mapping);
495                 }
496 
497                 /*
498                  * Clear hugetlb-related page reserves for children. This only
499                  * affects MAP_PRIVATE mappings. Faults generated by the child
500                  * are not guaranteed to succeed, even if read-only
501                  */
502                 if (is_vm_hugetlb_page(tmp))
503                         reset_vma_resv_huge_pages(tmp);
504 
505                 /*
506                  * Link in the new vma and copy the page table entries.
507                  */
508                 *pprev = tmp;
509                 pprev = &tmp->vm_next;
510                 tmp->vm_prev = prev;
511                 prev = tmp;
512 
513                 __vma_link_rb(mm, tmp, rb_link, rb_parent);
514                 rb_link = &tmp->vm_rb.rb_right;
515                 rb_parent = &tmp->vm_rb;
516 
517                 mm->map_count++;
518                 retval = copy_page_range(mm, oldmm, mpnt);
519 
520                 if (tmp->vm_ops && tmp->vm_ops->open)
521                         tmp->vm_ops->open(tmp);
522 
523                 if (retval)
524                         goto out;
525         }
526         /* a new mm has just been created */
527         arch_dup_mmap(oldmm, mm);
528         retval = 0;
529 out:
530         up_write(&mm->mmap_sem);
531         flush_tlb_mm(oldmm);
532         up_write(&oldmm->mmap_sem);
533 fail_uprobe_end:
534         uprobe_end_dup_mmap();
535         return retval;
536 fail_nomem_anon_vma_fork:
537         mpol_put(vma_policy(tmp));
538 fail_nomem_policy:
539         kmem_cache_free(vm_area_cachep, tmp);
540 fail_nomem:
541         retval = -ENOMEM;
542         vm_unacct_memory(charge);
543         goto out;
544 }
545 
546 static inline int mm_alloc_pgd(struct mm_struct *mm)
547 {
548         mm->pgd = pgd_alloc(mm);
549         if (unlikely(!mm->pgd))
550                 return -ENOMEM;
551         return 0;
552 }
553 
554 static inline void mm_free_pgd(struct mm_struct *mm)
555 {
556         pgd_free(mm, mm->pgd);
557 }
558 #else
559 static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
560 {
561         down_write(&oldmm->mmap_sem);
562         RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
563         up_write(&oldmm->mmap_sem);
564         return 0;
565 }
566 #define mm_alloc_pgd(mm)        (0)
567 #define mm_free_pgd(mm)
568 #endif /* CONFIG_MMU */
569 
570 __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
571 
572 #define allocate_mm()   (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
573 #define free_mm(mm)     (kmem_cache_free(mm_cachep, (mm)))
574 
575 static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
576 
577 static int __init coredump_filter_setup(char *s)
578 {
579         default_dump_filter =
580                 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
581                 MMF_DUMP_FILTER_MASK;
582         return 1;
583 }
584 
585 __setup("coredump_filter=", coredump_filter_setup);
586 
587 #include <linux/init_task.h>
588 
589 static void mm_init_aio(struct mm_struct *mm)
590 {
591 #ifdef CONFIG_AIO
592         spin_lock_init(&mm->ioctx_lock);
593         mm->ioctx_table = NULL;
594 #endif
595 }
596 
597 static void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
598 {
599 #ifdef CONFIG_MEMCG
600         mm->owner = p;
601 #endif
602 }
603 
604 static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p)
605 {
606         mm->mmap = NULL;
607         mm->mm_rb = RB_ROOT;
608         mm->vmacache_seqnum = 0;
609         atomic_set(&mm->mm_users, 1);
610         atomic_set(&mm->mm_count, 1);
611         init_rwsem(&mm->mmap_sem);
612         INIT_LIST_HEAD(&mm->mmlist);
613         mm->core_state = NULL;
614         atomic_long_set(&mm->nr_ptes, 0);
615         mm_nr_pmds_init(mm);
616         mm->map_count = 0;
617         mm->locked_vm = 0;
618         mm->pinned_vm = 0;
619         memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
620         spin_lock_init(&mm->page_table_lock);
621         mm_init_cpumask(mm);
622         mm_init_aio(mm);
623         mm_init_owner(mm, p);
624         mmu_notifier_mm_init(mm);
625         clear_tlb_flush_pending(mm);
626 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
627         mm->pmd_huge_pte = NULL;
628 #endif
629 
630         if (current->mm) {
631                 mm->flags = current->mm->flags & MMF_INIT_MASK;
632                 mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK;
633         } else {
634                 mm->flags = default_dump_filter;
635                 mm->def_flags = 0;
636         }
637 
638         if (mm_alloc_pgd(mm))
639                 goto fail_nopgd;
640 
641         if (init_new_context(p, mm))
642                 goto fail_nocontext;
643 
644         return mm;
645 
646 fail_nocontext:
647         mm_free_pgd(mm);
648 fail_nopgd:
649         free_mm(mm);
650         return NULL;
651 }
652 
653 static void check_mm(struct mm_struct *mm)
654 {
655         int i;
656 
657         for (i = 0; i < NR_MM_COUNTERS; i++) {
658                 long x = atomic_long_read(&mm->rss_stat.count[i]);
659 
660                 if (unlikely(x))
661                         printk(KERN_ALERT "BUG: Bad rss-counter state "
662                                           "mm:%p idx:%d val:%ld\n", mm, i, x);
663         }
664 
665         if (atomic_long_read(&mm->nr_ptes))
666                 pr_alert("BUG: non-zero nr_ptes on freeing mm: %ld\n",
667                                 atomic_long_read(&mm->nr_ptes));
668         if (mm_nr_pmds(mm))
669                 pr_alert("BUG: non-zero nr_pmds on freeing mm: %ld\n",
670                                 mm_nr_pmds(mm));
671 
672 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
673         VM_BUG_ON_MM(mm->pmd_huge_pte, mm);
674 #endif
675 }
676 
677 /*
678  * Allocate and initialize an mm_struct.
679  */
680 struct mm_struct *mm_alloc(void)
681 {
682         struct mm_struct *mm;
683 
684         mm = allocate_mm();
685         if (!mm)
686                 return NULL;
687 
688         memset(mm, 0, sizeof(*mm));
689         return mm_init(mm, current);
690 }
691 
692 /*
693  * Called when the last reference to the mm
694  * is dropped: either by a lazy thread or by
695  * mmput. Free the page directory and the mm.
696  */
697 void __mmdrop(struct mm_struct *mm)
698 {
699         BUG_ON(mm == &init_mm);
700         mm_free_pgd(mm);
701         destroy_context(mm);
702         mmu_notifier_mm_destroy(mm);
703         check_mm(mm);
704         free_mm(mm);
705 }
706 EXPORT_SYMBOL_GPL(__mmdrop);
707 
708 static inline void __mmput(struct mm_struct *mm)
709 {
710         VM_BUG_ON(atomic_read(&mm->mm_users));
711 
712         uprobe_clear_state(mm);
713         exit_aio(mm);
714         ksm_exit(mm);
715         khugepaged_exit(mm); /* must run before exit_mmap */
716         exit_mmap(mm);
717         set_mm_exe_file(mm, NULL);
718         if (!list_empty(&mm->mmlist)) {
719                 spin_lock(&mmlist_lock);
720                 list_del(&mm->mmlist);
721                 spin_unlock(&mmlist_lock);
722         }
723         if (mm->binfmt)
724                 module_put(mm->binfmt->module);
725         mmdrop(mm);
726 }
727 
728 /*
729  * Decrement the use count and release all resources for an mm.
730  */
731 void mmput(struct mm_struct *mm)
732 {
733         might_sleep();
734 
735         if (atomic_dec_and_test(&mm->mm_users))
736                 __mmput(mm);
737 }
738 EXPORT_SYMBOL_GPL(mmput);
739 
740 #ifdef CONFIG_MMU
741 static void mmput_async_fn(struct work_struct *work)
742 {
743         struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work);
744         __mmput(mm);
745 }
746 
747 void mmput_async(struct mm_struct *mm)
748 {
749         if (atomic_dec_and_test(&mm->mm_users)) {
750                 INIT_WORK(&mm->async_put_work, mmput_async_fn);
751                 schedule_work(&mm->async_put_work);
752         }
753 }
754 #endif
755 
756 /**
757  * set_mm_exe_file - change a reference to the mm's executable file
758  *
759  * This changes mm's executable file (shown as symlink /proc/[pid]/exe).
760  *
761  * Main users are mmput() and sys_execve(). Callers prevent concurrent
762  * invocations: in mmput() nobody alive left, in execve task is single
763  * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the
764  * mm->exe_file, but does so without using set_mm_exe_file() in order
765  * to do avoid the need for any locks.
766  */
767 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
768 {
769         struct file *old_exe_file;
770 
771         /*
772          * It is safe to dereference the exe_file without RCU as
773          * this function is only called if nobody else can access
774          * this mm -- see comment above for justification.
775          */
776         old_exe_file = rcu_dereference_raw(mm->exe_file);
777 
778         if (new_exe_file)
779                 get_file(new_exe_file);
780         rcu_assign_pointer(mm->exe_file, new_exe_file);
781         if (old_exe_file)
782                 fput(old_exe_file);
783 }
784 
785 /**
786  * get_mm_exe_file - acquire a reference to the mm's executable file
787  *
788  * Returns %NULL if mm has no associated executable file.
789  * User must release file via fput().
790  */
791 struct file *get_mm_exe_file(struct mm_struct *mm)
792 {
793         struct file *exe_file;
794 
795         rcu_read_lock();
796         exe_file = rcu_dereference(mm->exe_file);
797         if (exe_file && !get_file_rcu(exe_file))
798                 exe_file = NULL;
799         rcu_read_unlock();
800         return exe_file;
801 }
802 EXPORT_SYMBOL(get_mm_exe_file);
803 
804 /**
805  * get_task_exe_file - acquire a reference to the task's executable file
806  *
807  * Returns %NULL if task's mm (if any) has no associated executable file or
808  * this is a kernel thread with borrowed mm (see the comment above get_task_mm).
809  * User must release file via fput().
810  */
811 struct file *get_task_exe_file(struct task_struct *task)
812 {
813         struct file *exe_file = NULL;
814         struct mm_struct *mm;
815 
816         task_lock(task);
817         mm = task->mm;
818         if (mm) {
819                 if (!(task->flags & PF_KTHREAD))
820                         exe_file = get_mm_exe_file(mm);
821         }
822         task_unlock(task);
823         return exe_file;
824 }
825 EXPORT_SYMBOL(get_task_exe_file);
826 
827 /**
828  * get_task_mm - acquire a reference to the task's mm
829  *
830  * Returns %NULL if the task has no mm.  Checks PF_KTHREAD (meaning
831  * this kernel workthread has transiently adopted a user mm with use_mm,
832  * to do its AIO) is not set and if so returns a reference to it, after
833  * bumping up the use count.  User must release the mm via mmput()
834  * after use.  Typically used by /proc and ptrace.
835  */
836 struct mm_struct *get_task_mm(struct task_struct *task)
837 {
838         struct mm_struct *mm;
839 
840         task_lock(task);
841         mm = task->mm;
842         if (mm) {
843                 if (task->flags & PF_KTHREAD)
844                         mm = NULL;
845                 else
846                         atomic_inc(&mm->mm_users);
847         }
848         task_unlock(task);
849         return mm;
850 }
851 EXPORT_SYMBOL_GPL(get_task_mm);
852 
853 struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
854 {
855         struct mm_struct *mm;
856         int err;
857 
858         err =  mutex_lock_killable(&task->signal->cred_guard_mutex);
859         if (err)
860                 return ERR_PTR(err);
861 
862         mm = get_task_mm(task);
863         if (mm && mm != current->mm &&
864                         !ptrace_may_access(task, mode)) {
865                 mmput(mm);
866                 mm = ERR_PTR(-EACCES);
867         }
868         mutex_unlock(&task->signal->cred_guard_mutex);
869 
870         return mm;
871 }
872 
873 static void complete_vfork_done(struct task_struct *tsk)
874 {
875         struct completion *vfork;
876 
877         task_lock(tsk);
878         vfork = tsk->vfork_done;
879         if (likely(vfork)) {
880                 tsk->vfork_done = NULL;
881                 complete(vfork);
882         }
883         task_unlock(tsk);
884 }
885 
886 static int wait_for_vfork_done(struct task_struct *child,
887                                 struct completion *vfork)
888 {
889         int killed;
890 
891         freezer_do_not_count();
892         killed = wait_for_completion_killable(vfork);
893         freezer_count();
894 
895         if (killed) {
896                 task_lock(child);
897                 child->vfork_done = NULL;
898                 task_unlock(child);
899         }
900 
901         put_task_struct(child);
902         return killed;
903 }
904 
905 /* Please note the differences between mmput and mm_release.
906  * mmput is called whenever we stop holding onto a mm_struct,
907  * error success whatever.
908  *
909  * mm_release is called after a mm_struct has been removed
910  * from the current process.
911  *
912  * This difference is important for error handling, when we
913  * only half set up a mm_struct for a new process and need to restore
914  * the old one.  Because we mmput the new mm_struct before
915  * restoring the old one. . .
916  * Eric Biederman 10 January 1998
917  */
918 void mm_release(struct task_struct *tsk, struct mm_struct *mm)
919 {
920         /* Get rid of any futexes when releasing the mm */
921 #ifdef CONFIG_FUTEX
922         if (unlikely(tsk->robust_list)) {
923                 exit_robust_list(tsk);
924                 tsk->robust_list = NULL;
925         }
926 #ifdef CONFIG_COMPAT
927         if (unlikely(tsk->compat_robust_list)) {
928                 compat_exit_robust_list(tsk);
929                 tsk->compat_robust_list = NULL;
930         }
931 #endif
932         if (unlikely(!list_empty(&tsk->pi_state_list)))
933                 exit_pi_state_list(tsk);
934 #endif
935 
936         uprobe_free_utask(tsk);
937 
938         /* Get rid of any cached register state */
939         deactivate_mm(tsk, mm);
940 
941         /*
942          * Signal userspace if we're not exiting with a core dump
943          * because we want to leave the value intact for debugging
944          * purposes.
945          */
946         if (tsk->clear_child_tid) {
947                 if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) &&
948                     atomic_read(&mm->mm_users) > 1) {
949                         /*
950                          * We don't check the error code - if userspace has
951                          * not set up a proper pointer then tough luck.
952                          */
953                         put_user(0, tsk->clear_child_tid);
954                         sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
955                                         1, NULL, NULL, 0);
956                 }
957                 tsk->clear_child_tid = NULL;
958         }
959 
960         /*
961          * All done, finally we can wake up parent and return this mm to him.
962          * Also kthread_stop() uses this completion for synchronization.
963          */
964         if (tsk->vfork_done)
965                 complete_vfork_done(tsk);
966 }
967 
968 /*
969  * Allocate a new mm structure and copy contents from the
970  * mm structure of the passed in task structure.
971  */
972 static struct mm_struct *dup_mm(struct task_struct *tsk)
973 {
974         struct mm_struct *mm, *oldmm = current->mm;
975         int err;
976 
977         mm = allocate_mm();
978         if (!mm)
979                 goto fail_nomem;
980 
981         memcpy(mm, oldmm, sizeof(*mm));
982 
983         if (!mm_init(mm, tsk))
984                 goto fail_nomem;
985 
986         err = dup_mmap(mm, oldmm);
987         if (err)
988                 goto free_pt;
989 
990         mm->hiwater_rss = get_mm_rss(mm);
991         mm->hiwater_vm = mm->total_vm;
992 
993         if (mm->binfmt && !try_module_get(mm->binfmt->module))
994                 goto free_pt;
995 
996         return mm;
997 
998 free_pt:
999         /* don't put binfmt in mmput, we haven't got module yet */
1000         mm->binfmt = NULL;
1001         mmput(mm);
1002 
1003 fail_nomem:
1004         return NULL;
1005 }
1006 
1007 static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
1008 {
1009         struct mm_struct *mm, *oldmm;
1010         int retval;
1011 
1012         tsk->min_flt = tsk->maj_flt = 0;
1013         tsk->nvcsw = tsk->nivcsw = 0;
1014 #ifdef CONFIG_DETECT_HUNG_TASK
1015         tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
1016 #endif
1017 
1018         tsk->mm = NULL;
1019         tsk->active_mm = NULL;
1020 
1021         /*
1022          * Are we cloning a kernel thread?
1023          *
1024          * We need to steal a active VM for that..
1025          */
1026         oldmm = current->mm;
1027         if (!oldmm)
1028                 return 0;
1029 
1030         /* initialize the new vmacache entries */
1031         vmacache_flush(tsk);
1032 
1033         if (clone_flags & CLONE_VM) {
1034                 atomic_inc(&oldmm->mm_users);
1035                 mm = oldmm;
1036                 goto good_mm;
1037         }
1038 
1039         retval = -ENOMEM;
1040         mm = dup_mm(tsk);
1041         if (!mm)
1042                 goto fail_nomem;
1043 
1044 good_mm:
1045         tsk->mm = mm;
1046         tsk->active_mm = mm;
1047         return 0;
1048 
1049 fail_nomem:
1050         return retval;
1051 }
1052 
1053 static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
1054 {
1055         struct fs_struct *fs = current->fs;
1056         if (clone_flags & CLONE_FS) {
1057                 /* tsk->fs is already what we want */
1058                 spin_lock(&fs->lock);
1059                 if (fs->in_exec) {
1060                         spin_unlock(&fs->lock);
1061                         return -EAGAIN;
1062                 }
1063                 fs->users++;
1064                 spin_unlock(&fs->lock);
1065                 return 0;
1066         }
1067         tsk->fs = copy_fs_struct(fs);
1068         if (!tsk->fs)
1069                 return -ENOMEM;
1070         return 0;
1071 }
1072 
1073 static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
1074 {
1075         struct files_struct *oldf, *newf;
1076         int error = 0;
1077 
1078         /*
1079          * A background process may not have any files ...
1080          */
1081         oldf = current->files;
1082         if (!oldf)
1083                 goto out;
1084 
1085         if (clone_flags & CLONE_FILES) {
1086                 atomic_inc(&oldf->count);
1087                 goto out;
1088         }
1089 
1090         newf = dup_fd(oldf, &error);
1091         if (!newf)
1092                 goto out;
1093 
1094         tsk->files = newf;
1095         error = 0;
1096 out:
1097         return error;
1098 }
1099 
1100 static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
1101 {
1102 #ifdef CONFIG_BLOCK
1103         struct io_context *ioc = current->io_context;
1104         struct io_context *new_ioc;
1105 
1106         if (!ioc)
1107                 return 0;
1108         /*
1109          * Share io context with parent, if CLONE_IO is set
1110          */
1111         if (clone_flags & CLONE_IO) {
1112                 ioc_task_link(ioc);
1113                 tsk->io_context = ioc;
1114         } else if (ioprio_valid(ioc->ioprio)) {
1115                 new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE);
1116                 if (unlikely(!new_ioc))
1117                         return -ENOMEM;
1118 
1119                 new_ioc->ioprio = ioc->ioprio;
1120                 put_io_context(new_ioc);
1121         }
1122 #endif
1123         return 0;
1124 }
1125 
1126 static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
1127 {
1128         struct sighand_struct *sig;
1129 
1130         if (clone_flags & CLONE_SIGHAND) {
1131                 atomic_inc(&current->sighand->count);
1132                 return 0;
1133         }
1134         sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
1135         rcu_assign_pointer(tsk->sighand, sig);
1136         if (!sig)
1137                 return -ENOMEM;
1138 
1139         atomic_set(&sig->count, 1);
1140         memcpy(sig->action, current->sighand->action, sizeof(sig->action));
1141         return 0;
1142 }
1143 
1144 void __cleanup_sighand(struct sighand_struct *sighand)
1145 {
1146         if (atomic_dec_and_test(&sighand->count)) {
1147                 signalfd_cleanup(sighand);
1148                 /*
1149                  * sighand_cachep is SLAB_DESTROY_BY_RCU so we can free it
1150                  * without an RCU grace period, see __lock_task_sighand().
1151                  */
1152                 kmem_cache_free(sighand_cachep, sighand);
1153         }
1154 }
1155 
1156 /*
1157  * Initialize POSIX timer handling for a thread group.
1158  */
1159 static void posix_cpu_timers_init_group(struct signal_struct *sig)
1160 {
1161         unsigned long cpu_limit;
1162 
1163         cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
1164         if (cpu_limit != RLIM_INFINITY) {
1165                 sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit);
1166                 sig->cputimer.running = true;
1167         }
1168 
1169         /* The timer lists. */
1170         INIT_LIST_HEAD(&sig->cpu_timers[0]);
1171         INIT_LIST_HEAD(&sig->cpu_timers[1]);
1172         INIT_LIST_HEAD(&sig->cpu_timers[2]);
1173 }
1174 
1175 static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
1176 {
1177         struct signal_struct *sig;
1178 
1179         if (clone_flags & CLONE_THREAD)
1180                 return 0;
1181 
1182         sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
1183         tsk->signal = sig;
1184         if (!sig)
1185                 return -ENOMEM;
1186 
1187         sig->nr_threads = 1;
1188         atomic_set(&sig->live, 1);
1189         atomic_set(&sig->sigcnt, 1);
1190 
1191         /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */
1192         sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node);
1193         tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head);
1194 
1195         init_waitqueue_head(&sig->wait_chldexit);
1196         sig->curr_target = tsk;
1197         init_sigpending(&sig->shared_pending);
1198         INIT_LIST_HEAD(&sig->posix_timers);
1199         seqlock_init(&sig->stats_lock);
1200         prev_cputime_init(&sig->prev_cputime);
1201 
1202         hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
1203         sig->real_timer.function = it_real_fn;
1204 
1205         task_lock(current->group_leader);
1206         memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
1207         task_unlock(current->group_leader);
1208 
1209         posix_cpu_timers_init_group(sig);
1210 
1211         tty_audit_fork(sig);
1212         sched_autogroup_fork(sig);
1213 
1214         sig->oom_score_adj = current->signal->oom_score_adj;
1215         sig->oom_score_adj_min = current->signal->oom_score_adj_min;
1216 
1217         sig->has_child_subreaper = current->signal->has_child_subreaper ||
1218                                    current->signal->is_child_subreaper;
1219 
1220         mutex_init(&sig->cred_guard_mutex);
1221 
1222         return 0;
1223 }
1224 
1225 static void copy_seccomp(struct task_struct *p)
1226 {
1227 #ifdef CONFIG_SECCOMP
1228         /*
1229          * Must be called with sighand->lock held, which is common to
1230          * all threads in the group. Holding cred_guard_mutex is not
1231          * needed because this new task is not yet running and cannot
1232          * be racing exec.
1233          */
1234         assert_spin_locked(&current->sighand->siglock);
1235 
1236         /* Ref-count the new filter user, and assign it. */
1237         get_seccomp_filter(current);
1238         p->seccomp = current->seccomp;
1239 
1240         /*
1241          * Explicitly enable no_new_privs here in case it got set
1242          * between the task_struct being duplicated and holding the
1243          * sighand lock. The seccomp state and nnp must be in sync.
1244          */
1245         if (task_no_new_privs(current))
1246                 task_set_no_new_privs(p);
1247 
1248         /*
1249          * If the parent gained a seccomp mode after copying thread
1250          * flags and between before we held the sighand lock, we have
1251          * to manually enable the seccomp thread flag here.
1252          */
1253         if (p->seccomp.mode != SECCOMP_MODE_DISABLED)
1254                 set_tsk_thread_flag(p, TIF_SECCOMP);
1255 #endif
1256 }
1257 
1258 SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
1259 {
1260         current->clear_child_tid = tidptr;
1261 
1262         return task_pid_vnr(current);
1263 }
1264 
1265 static void rt_mutex_init_task(struct task_struct *p)
1266 {
1267         raw_spin_lock_init(&p->pi_lock);
1268 #ifdef CONFIG_RT_MUTEXES
1269         p->pi_waiters = RB_ROOT;
1270         p->pi_waiters_leftmost = NULL;
1271         p->pi_blocked_on = NULL;
1272 #endif
1273 }
1274 
1275 /*
1276  * Initialize POSIX timer handling for a single task.
1277  */
1278 static void posix_cpu_timers_init(struct task_struct *tsk)
1279 {
1280         tsk->cputime_expires.prof_exp = 0;
1281         tsk->cputime_expires.virt_exp = 0;
1282         tsk->cputime_expires.sched_exp = 0;
1283         INIT_LIST_HEAD(&tsk->cpu_timers[0]);
1284         INIT_LIST_HEAD(&tsk->cpu_timers[1]);
1285         INIT_LIST_HEAD(&tsk->cpu_timers[2]);
1286 }
1287 
1288 static inline void
1289 init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid)
1290 {
1291          task->pids[type].pid = pid;
1292 }
1293 
1294 /*
1295  * This creates a new process as a copy of the old one,
1296  * but does not actually start it yet.
1297  *
1298  * It copies the registers, and all the appropriate
1299  * parts of the process environment (as per the clone
1300  * flags). The actual kick-off is left to the caller.
1301  */
1302 static struct task_struct *copy_process(unsigned long clone_flags,
1303                                         unsigned long stack_start,
1304                                         unsigned long stack_size,
1305                                         int __user *child_tidptr,
1306                                         struct pid *pid,
1307                                         int trace,
1308                                         unsigned long tls,
1309                                         int node)
1310 {
1311         int retval;
1312         struct task_struct *p;
1313 
1314         if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
1315                 return ERR_PTR(-EINVAL);
1316 
1317         if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS))
1318                 return ERR_PTR(-EINVAL);
1319 
1320         /*
1321          * Thread groups must share signals as well, and detached threads
1322          * can only be started up within the thread group.
1323          */
1324         if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1325                 return ERR_PTR(-EINVAL);
1326 
1327         /*
1328          * Shared signal handlers imply shared VM. By way of the above,
1329          * thread groups also imply shared VM. Blocking this case allows
1330          * for various simplifications in other code.
1331          */
1332         if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1333                 return ERR_PTR(-EINVAL);
1334 
1335         /*
1336          * Siblings of global init remain as zombies on exit since they are
1337          * not reaped by their parent (swapper). To solve this and to avoid
1338          * multi-rooted process trees, prevent global and container-inits
1339          * from creating siblings.
1340          */
1341         if ((clone_flags & CLONE_PARENT) &&
1342                                 current->signal->flags & SIGNAL_UNKILLABLE)
1343                 return ERR_PTR(-EINVAL);
1344 
1345         /*
1346          * If the new process will be in a different pid or user namespace
1347          * do not allow it to share a thread group with the forking task.
1348          */
1349         if (clone_flags & CLONE_THREAD) {
1350                 if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) ||
1351                     (task_active_pid_ns(current) !=
1352                                 current->nsproxy->pid_ns_for_children))
1353                         return ERR_PTR(-EINVAL);
1354         }
1355 
1356         retval = security_task_create(clone_flags);
1357         if (retval)
1358                 goto fork_out;
1359 
1360         retval = -ENOMEM;
1361         p = dup_task_struct(current, node);
1362         if (!p)
1363                 goto fork_out;
1364 
1365         ftrace_graph_init_task(p);
1366 
1367         rt_mutex_init_task(p);
1368 
1369 #ifdef CONFIG_PROVE_LOCKING
1370         DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1371         DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1372 #endif
1373         retval = -EAGAIN;
1374         if (atomic_read(&p->real_cred->user->processes) >=
1375                         task_rlimit(p, RLIMIT_NPROC)) {
1376                 if (p->real_cred->user != INIT_USER &&
1377                     !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN))
1378                         goto bad_fork_free;
1379         }
1380         current->flags &= ~PF_NPROC_EXCEEDED;
1381 
1382         retval = copy_creds(p, clone_flags);
1383         if (retval < 0)
1384                 goto bad_fork_free;
1385 
1386         /*
1387          * If multiple threads are within copy_process(), then this check
1388          * triggers too late. This doesn't hurt, the check is only there
1389          * to stop root fork bombs.
1390          */
1391         retval = -EAGAIN;
1392         if (nr_threads >= max_threads)
1393                 goto bad_fork_cleanup_count;
1394 
1395         delayacct_tsk_init(p);  /* Must remain after dup_task_struct() */
1396         p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER);
1397         p->flags |= PF_FORKNOEXEC;
1398         INIT_LIST_HEAD(&p->children);
1399         INIT_LIST_HEAD(&p->sibling);
1400         rcu_copy_process(p);
1401         p->vfork_done = NULL;
1402         spin_lock_init(&p->alloc_lock);
1403 
1404         init_sigpending(&p->pending);
1405 
1406         p->utime = p->stime = p->gtime = 0;
1407         p->utimescaled = p->stimescaled = 0;
1408         prev_cputime_init(&p->prev_cputime);
1409 
1410 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1411         seqcount_init(&p->vtime_seqcount);
1412         p->vtime_snap = 0;
1413         p->vtime_snap_whence = VTIME_INACTIVE;
1414 #endif
1415 
1416 #if defined(SPLIT_RSS_COUNTING)
1417         memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1418 #endif
1419 
1420         p->default_timer_slack_ns = current->timer_slack_ns;
1421 
1422         task_io_accounting_init(&p->ioac);
1423         acct_clear_integrals(p);
1424 
1425         posix_cpu_timers_init(p);
1426 
1427         p->start_time = ktime_get_ns();
1428         p->real_start_time = ktime_get_boot_ns();
1429         p->io_context = NULL;
1430         p->audit_context = NULL;
1431         cgroup_fork(p);
1432 #ifdef CONFIG_NUMA
1433         p->mempolicy = mpol_dup(p->mempolicy);
1434         if (IS_ERR(p->mempolicy)) {
1435                 retval = PTR_ERR(p->mempolicy);
1436                 p->mempolicy = NULL;
1437                 goto bad_fork_cleanup_threadgroup_lock;
1438         }
1439 #endif
1440 #ifdef CONFIG_CPUSETS
1441         p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
1442         p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
1443         seqcount_init(&p->mems_allowed_seq);
1444 #endif
1445 #ifdef CONFIG_TRACE_IRQFLAGS
1446         p->irq_events = 0;
1447         p->hardirqs_enabled = 0;
1448         p->hardirq_enable_ip = 0;
1449         p->hardirq_enable_event = 0;
1450         p->hardirq_disable_ip = _THIS_IP_;
1451         p->hardirq_disable_event = 0;
1452         p->softirqs_enabled = 1;
1453         p->softirq_enable_ip = _THIS_IP_;
1454         p->softirq_enable_event = 0;
1455         p->softirq_disable_ip = 0;
1456         p->softirq_disable_event = 0;
1457         p->hardirq_context = 0;
1458         p->softirq_context = 0;
1459 #endif
1460 
1461         p->pagefault_disabled = 0;
1462 
1463 #ifdef CONFIG_LOCKDEP
1464         p->lockdep_depth = 0; /* no locks held yet */
1465         p->curr_chain_key = 0;
1466         p->lockdep_recursion = 0;
1467 #endif
1468 
1469 #ifdef CONFIG_DEBUG_MUTEXES
1470         p->blocked_on = NULL; /* not blocked yet */
1471 #endif
1472 #ifdef CONFIG_BCACHE
1473         p->sequential_io        = 0;
1474         p->sequential_io_avg    = 0;
1475 #endif
1476 
1477         /* Perform scheduler related setup. Assign this task to a CPU. */
1478         retval = sched_fork(clone_flags, p);
1479         if (retval)
1480                 goto bad_fork_cleanup_policy;
1481 
1482         retval = perf_event_init_task(p);
1483         if (retval)
1484                 goto bad_fork_cleanup_policy;
1485         retval = audit_alloc(p);
1486         if (retval)
1487                 goto bad_fork_cleanup_perf;
1488         /* copy all the process information */
1489         shm_init_task(p);
1490         retval = ccs_alloc_task_security(p);
1491         if (retval)
1492                 goto bad_fork_cleanup_audit;
1493         retval = copy_semundo(clone_flags, p);
1494         if (retval)
1495                 goto bad_fork_cleanup_audit;
1496         retval = copy_files(clone_flags, p);
1497         if (retval)
1498                 goto bad_fork_cleanup_semundo;
1499         retval = copy_fs(clone_flags, p);
1500         if (retval)
1501                 goto bad_fork_cleanup_files;
1502         retval = copy_sighand(clone_flags, p);
1503         if (retval)
1504                 goto bad_fork_cleanup_fs;
1505         retval = copy_signal(clone_flags, p);
1506         if (retval)
1507                 goto bad_fork_cleanup_sighand;
1508         retval = copy_mm(clone_flags, p);
1509         if (retval)
1510                 goto bad_fork_cleanup_signal;
1511         retval = copy_namespaces(clone_flags, p);
1512         if (retval)
1513                 goto bad_fork_cleanup_mm;
1514         retval = copy_io(clone_flags, p);
1515         if (retval)
1516                 goto bad_fork_cleanup_namespaces;
1517         retval = copy_thread_tls(clone_flags, stack_start, stack_size, p, tls);
1518         if (retval)
1519                 goto bad_fork_cleanup_io;
1520 
1521         if (pid != &init_struct_pid) {
1522                 pid = alloc_pid(p->nsproxy->pid_ns_for_children);
1523                 if (IS_ERR(pid)) {
1524                         retval = PTR_ERR(pid);
1525                         goto bad_fork_cleanup_thread;
1526                 }
1527         }
1528 
1529         p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1530         /*
1531          * Clear TID on mm_release()?
1532          */
1533         p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL;
1534 #ifdef CONFIG_BLOCK
1535         p->plug = NULL;
1536 #endif
1537 #ifdef CONFIG_FUTEX
1538         p->robust_list = NULL;
1539 #ifdef CONFIG_COMPAT
1540         p->compat_robust_list = NULL;
1541 #endif
1542         INIT_LIST_HEAD(&p->pi_state_list);
1543         p->pi_state_cache = NULL;
1544 #endif
1545         /*
1546          * sigaltstack should be cleared when sharing the same VM
1547          */
1548         if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1549                 sas_ss_reset(p);
1550 
1551         /*
1552          * Syscall tracing and stepping should be turned off in the
1553          * child regardless of CLONE_PTRACE.
1554          */
1555         user_disable_single_step(p);
1556         clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
1557 #ifdef TIF_SYSCALL_EMU
1558         clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1559 #endif
1560         clear_all_latency_tracing(p);
1561 
1562         /* ok, now we should be set up.. */
1563         p->pid = pid_nr(pid);
1564         if (clone_flags & CLONE_THREAD) {
1565                 p->exit_signal = -1;
1566                 p->group_leader = current->group_leader;
1567                 p->tgid = current->tgid;
1568         } else {
1569                 if (clone_flags & CLONE_PARENT)
1570                         p->exit_signal = current->group_leader->exit_signal;
1571                 else
1572                         p->exit_signal = (clone_flags & CSIGNAL);
1573                 p->group_leader = p;
1574                 p->tgid = p->pid;
1575         }
1576 
1577         p->nr_dirtied = 0;
1578         p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
1579         p->dirty_paused_when = 0;
1580 
1581         p->pdeath_signal = 0;
1582         INIT_LIST_HEAD(&p->thread_group);
1583         p->task_works = NULL;
1584 
1585         threadgroup_change_begin(current);
1586         /*
1587          * Ensure that the cgroup subsystem policies allow the new process to be
1588          * forked. It should be noted the the new process's css_set can be changed
1589          * between here and cgroup_post_fork() if an organisation operation is in
1590          * progress.
1591          */
1592         retval = cgroup_can_fork(p);
1593         if (retval)
1594                 goto bad_fork_free_pid;
1595 
1596         /*
1597          * Make it visible to the rest of the system, but dont wake it up yet.
1598          * Need tasklist lock for parent etc handling!
1599          */
1600         write_lock_irq(&tasklist_lock);
1601 
1602         /* CLONE_PARENT re-uses the old parent */
1603         if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
1604                 p->real_parent = current->real_parent;
1605                 p->parent_exec_id = current->parent_exec_id;
1606         } else {
1607                 p->real_parent = current;
1608                 p->parent_exec_id = current->self_exec_id;
1609         }
1610 
1611         spin_lock(&current->sighand->siglock);
1612 
1613         /*
1614          * Copy seccomp details explicitly here, in case they were changed
1615          * before holding sighand lock.
1616          */
1617         copy_seccomp(p);
1618 
1619         /*
1620          * Process group and session signals need to be delivered to just the
1621          * parent before the fork or both the parent and the child after the
1622          * fork. Restart if a signal comes in before we add the new process to
1623          * it's process group.
1624          * A fatal signal pending means that current will exit, so the new
1625          * thread can't slip out of an OOM kill (or normal SIGKILL).
1626         */
1627         recalc_sigpending();
1628         if (signal_pending(current)) {
1629                 spin_unlock(&current->sighand->siglock);
1630                 write_unlock_irq(&tasklist_lock);
1631                 retval = -ERESTARTNOINTR;
1632                 goto bad_fork_cancel_cgroup;
1633         }
1634 
1635         if (likely(p->pid)) {
1636                 ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
1637 
1638                 init_task_pid(p, PIDTYPE_PID, pid);
1639                 if (thread_group_leader(p)) {
1640                         init_task_pid(p, PIDTYPE_PGID, task_pgrp(current));
1641                         init_task_pid(p, PIDTYPE_SID, task_session(current));
1642 
1643                         if (is_child_reaper(pid)) {
1644                                 ns_of_pid(pid)->child_reaper = p;
1645                                 p->signal->flags |= SIGNAL_UNKILLABLE;
1646                         }
1647 
1648                         p->signal->leader_pid = pid;
1649                         p->signal->tty = tty_kref_get(current->signal->tty);
1650                         list_add_tail(&p->sibling, &p->real_parent->children);
1651                         list_add_tail_rcu(&p->tasks, &init_task.tasks);
1652                         attach_pid(p, PIDTYPE_PGID);
1653                         attach_pid(p, PIDTYPE_SID);
1654                         __this_cpu_inc(process_counts);
1655                 } else {
1656                         current->signal->nr_threads++;
1657                         atomic_inc(&current->signal->live);
1658                         atomic_inc(&current->signal->sigcnt);
1659                         list_add_tail_rcu(&p->thread_group,
1660                                           &p->group_leader->thread_group);
1661                         list_add_tail_rcu(&p->thread_node,
1662                                           &p->signal->thread_head);
1663                 }
1664                 attach_pid(p, PIDTYPE_PID);
1665                 nr_threads++;
1666         }
1667 
1668         total_forks++;
1669         spin_unlock(&current->sighand->siglock);
1670         syscall_tracepoint_update(p);
1671         write_unlock_irq(&tasklist_lock);
1672 
1673         proc_fork_connector(p);
1674         cgroup_post_fork(p);
1675         threadgroup_change_end(current);
1676         perf_event_fork(p);
1677 
1678         trace_task_newtask(p, clone_flags);
1679         uprobe_copy_process(p, clone_flags);
1680 
1681         return p;
1682 
1683 bad_fork_cancel_cgroup:
1684         cgroup_cancel_fork(p);
1685 bad_fork_free_pid:
1686         threadgroup_change_end(current);
1687         if (pid != &init_struct_pid)
1688                 free_pid(pid);
1689 bad_fork_cleanup_thread:
1690         exit_thread(p);
1691 bad_fork_cleanup_io:
1692         if (p->io_context)
1693                 exit_io_context(p);
1694 bad_fork_cleanup_namespaces:
1695         exit_task_namespaces(p);
1696 bad_fork_cleanup_mm:
1697         if (p->mm)
1698                 mmput(p->mm);
1699 bad_fork_cleanup_signal:
1700         if (!(clone_flags & CLONE_THREAD))
1701                 free_signal_struct(p->signal);
1702 bad_fork_cleanup_sighand:
1703         __cleanup_sighand(p->sighand);
1704 bad_fork_cleanup_fs:
1705         exit_fs(p); /* blocking */
1706 bad_fork_cleanup_files:
1707         exit_files(p); /* blocking */
1708 bad_fork_cleanup_semundo:
1709         exit_sem(p);
1710 bad_fork_cleanup_audit:
1711         audit_free(p);
1712         ccs_free_task_security(p);
1713 bad_fork_cleanup_perf:
1714         perf_event_free_task(p);
1715 bad_fork_cleanup_policy:
1716 #ifdef CONFIG_NUMA
1717         mpol_put(p->mempolicy);
1718 bad_fork_cleanup_threadgroup_lock:
1719 #endif
1720         delayacct_tsk_free(p);
1721 bad_fork_cleanup_count:
1722         atomic_dec(&p->cred->user->processes);
1723         exit_creds(p);
1724 bad_fork_free:
1725         free_task(p);
1726 fork_out:
1727         return ERR_PTR(retval);
1728 }
1729 
1730 static inline void init_idle_pids(struct pid_link *links)
1731 {
1732         enum pid_type type;
1733 
1734         for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1735                 INIT_HLIST_NODE(&links[type].node); /* not really needed */
1736                 links[type].pid = &init_struct_pid;
1737         }
1738 }
1739 
1740 struct task_struct *fork_idle(int cpu)
1741 {
1742         struct task_struct *task;
1743         task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0, 0,
1744                             cpu_to_node(cpu));
1745         if (!IS_ERR(task)) {
1746                 init_idle_pids(task->pids);
1747                 init_idle(task, cpu);
1748         }
1749 
1750         return task;
1751 }
1752 
1753 /*
1754  *  Ok, this is the main fork-routine.
1755  *
1756  * It copies the process, and if successful kick-starts
1757  * it and waits for it to finish using the VM if required.
1758  */
1759 long _do_fork(unsigned long clone_flags,
1760               unsigned long stack_start,
1761               unsigned long stack_size,
1762               int __user *parent_tidptr,
1763               int __user *child_tidptr,
1764               unsigned long tls)
1765 {
1766         struct task_struct *p;
1767         int trace = 0;
1768         long nr;
1769 
1770         /*
1771          * Determine whether and which event to report to ptracer.  When
1772          * called from kernel_thread or CLONE_UNTRACED is explicitly
1773          * requested, no event is reported; otherwise, report if the event
1774          * for the type of forking is enabled.
1775          */
1776         if (!(clone_flags & CLONE_UNTRACED)) {
1777                 if (clone_flags & CLONE_VFORK)
1778                         trace = PTRACE_EVENT_VFORK;
1779                 else if ((clone_flags & CSIGNAL) != SIGCHLD)
1780                         trace = PTRACE_EVENT_CLONE;
1781                 else
1782                         trace = PTRACE_EVENT_FORK;
1783 
1784                 if (likely(!ptrace_event_enabled(current, trace)))
1785                         trace = 0;
1786         }
1787 
1788         p = copy_process(clone_flags, stack_start, stack_size,
1789                          child_tidptr, NULL, trace, tls, NUMA_NO_NODE);
1790         /*
1791          * Do this prior waking up the new thread - the thread pointer
1792          * might get invalid after that point, if the thread exits quickly.
1793          */
1794         if (!IS_ERR(p)) {
1795                 struct completion vfork;
1796                 struct pid *pid;
1797 
1798                 trace_sched_process_fork(current, p);
1799 
1800                 pid = get_task_pid(p, PIDTYPE_PID);
1801                 nr = pid_vnr(pid);
1802 
1803                 if (clone_flags & CLONE_PARENT_SETTID)
1804                         put_user(nr, parent_tidptr);
1805 
1806                 if (clone_flags & CLONE_VFORK) {
1807                         p->vfork_done = &vfork;
1808                         init_completion(&vfork);
1809                         get_task_struct(p);
1810                 }
1811 
1812                 wake_up_new_task(p);
1813 
1814                 /* forking complete and child started to run, tell ptracer */
1815                 if (unlikely(trace))
1816                         ptrace_event_pid(trace, pid);
1817 
1818                 if (clone_flags & CLONE_VFORK) {
1819                         if (!wait_for_vfork_done(p, &vfork))
1820                                 ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid);
1821                 }
1822 
1823                 put_pid(pid);
1824         } else {
1825                 nr = PTR_ERR(p);
1826         }
1827         return nr;
1828 }
1829 
1830 #ifndef CONFIG_HAVE_COPY_THREAD_TLS
1831 /* For compatibility with architectures that call do_fork directly rather than
1832  * using the syscall entry points below. */
1833 long do_fork(unsigned long clone_flags,
1834               unsigned long stack_start,
1835               unsigned long stack_size,
1836               int __user *parent_tidptr,
1837               int __user *child_tidptr)
1838 {
1839         return _do_fork(clone_flags, stack_start, stack_size,
1840                         parent_tidptr, child_tidptr, 0);
1841 }
1842 #endif
1843 
1844 /*
1845  * Create a kernel thread.
1846  */
1847 pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
1848 {
1849         return _do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn,
1850                 (unsigned long)arg, NULL, NULL, 0);
1851 }
1852 
1853 #ifdef __ARCH_WANT_SYS_FORK
1854 SYSCALL_DEFINE0(fork)
1855 {
1856 #ifdef CONFIG_MMU
1857         return _do_fork(SIGCHLD, 0, 0, NULL, NULL, 0);
1858 #else
1859         /* can not support in nommu mode */
1860         return -EINVAL;
1861 #endif
1862 }
1863 #endif
1864 
1865 #ifdef __ARCH_WANT_SYS_VFORK
1866 SYSCALL_DEFINE0(vfork)
1867 {
1868         return _do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0,
1869                         0, NULL, NULL, 0);
1870 }
1871 #endif
1872 
1873 #ifdef __ARCH_WANT_SYS_CLONE
1874 #ifdef CONFIG_CLONE_BACKWARDS
1875 SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
1876                  int __user *, parent_tidptr,
1877                  unsigned long, tls,
1878                  int __user *, child_tidptr)
1879 #elif defined(CONFIG_CLONE_BACKWARDS2)
1880 SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags,
1881                  int __user *, parent_tidptr,
1882                  int __user *, child_tidptr,
1883                  unsigned long, tls)
1884 #elif defined(CONFIG_CLONE_BACKWARDS3)
1885 SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp,
1886                 int, stack_size,
1887                 int __user *, parent_tidptr,
1888                 int __user *, child_tidptr,
1889                 unsigned long, tls)
1890 #else
1891 SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
1892                  int __user *, parent_tidptr,
1893                  int __user *, child_tidptr,
1894                  unsigned long, tls)
1895 #endif
1896 {
1897         return _do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr, tls);
1898 }
1899 #endif
1900 
1901 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1902 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1903 #endif
1904 
1905 static void sighand_ctor(void *data)
1906 {
1907         struct sighand_struct *sighand = data;
1908 
1909         spin_lock_init(&sighand->siglock);
1910         init_waitqueue_head(&sighand->signalfd_wqh);
1911 }
1912 
1913 void __init proc_caches_init(void)
1914 {
1915         sighand_cachep = kmem_cache_create("sighand_cache",
1916                         sizeof(struct sighand_struct), 0,
1917                         SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
1918                         SLAB_NOTRACK|SLAB_ACCOUNT, sighand_ctor);
1919         signal_cachep = kmem_cache_create("signal_cache",
1920                         sizeof(struct signal_struct), 0,
1921                         SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
1922                         NULL);
1923         files_cachep = kmem_cache_create("files_cache",
1924                         sizeof(struct files_struct), 0,
1925                         SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
1926                         NULL);
1927         fs_cachep = kmem_cache_create("fs_cache",
1928                         sizeof(struct fs_struct), 0,
1929                         SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
1930                         NULL);
1931         /*
1932          * FIXME! The "sizeof(struct mm_struct)" currently includes the
1933          * whole struct cpumask for the OFFSTACK case. We could change
1934          * this to *only* allocate as much of it as required by the
1935          * maximum number of CPU's we can ever have.  The cpumask_allocation
1936          * is at the end of the structure, exactly for that reason.
1937          */
1938         mm_cachep = kmem_cache_create("mm_struct",
1939                         sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
1940                         SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
1941                         NULL);
1942         vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT);
1943         mmap_init();
1944         nsproxy_cache_init();
1945 }
1946 
1947 /*
1948  * Check constraints on flags passed to the unshare system call.
1949  */
1950 static int check_unshare_flags(unsigned long unshare_flags)
1951 {
1952         if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
1953                                 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
1954                                 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET|
1955                                 CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP))
1956                 return -EINVAL;
1957         /*
1958          * Not implemented, but pretend it works if there is nothing
1959          * to unshare.  Note that unsharing the address space or the
1960          * signal handlers also need to unshare the signal queues (aka
1961          * CLONE_THREAD).
1962          */
1963         if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
1964                 if (!thread_group_empty(current))
1965                         return -EINVAL;
1966         }
1967         if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) {
1968                 if (atomic_read(&current->sighand->count) > 1)
1969                         return -EINVAL;
1970         }
1971         if (unshare_flags & CLONE_VM) {
1972                 if (!current_is_single_threaded())
1973                         return -EINVAL;
1974         }
1975 
1976         return 0;
1977 }
1978 
1979 /*
1980  * Unshare the filesystem structure if it is being shared
1981  */
1982 static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1983 {
1984         struct fs_struct *fs = current->fs;
1985 
1986         if (!(unshare_flags & CLONE_FS) || !fs)
1987                 return 0;
1988 
1989         /* don't need lock here; in the worst case we'll do useless copy */
1990         if (fs->users == 1)
1991                 return 0;
1992 
1993         *new_fsp = copy_fs_struct(fs);
1994         if (!*new_fsp)
1995                 return -ENOMEM;
1996 
1997         return 0;
1998 }
1999 
2000 /*
2001  * Unshare file descriptor table if it is being shared
2002  */
2003 static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
2004 {
2005         struct files_struct *fd = current->files;
2006         int error = 0;
2007 
2008         if ((unshare_flags & CLONE_FILES) &&
2009             (fd && atomic_read(&fd->count) > 1)) {
2010                 *new_fdp = dup_fd(fd, &error);
2011                 if (!*new_fdp)
2012                         return error;
2013         }
2014 
2015         return 0;
2016 }
2017 
2018 /*
2019  * unshare allows a process to 'unshare' part of the process
2020  * context which was originally shared using clone.  copy_*
2021  * functions used by do_fork() cannot be used here directly
2022  * because they modify an inactive task_struct that is being
2023  * constructed. Here we are modifying the current, active,
2024  * task_struct.
2025  */
2026 SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
2027 {
2028         struct fs_struct *fs, *new_fs = NULL;
2029         struct files_struct *fd, *new_fd = NULL;
2030         struct cred *new_cred = NULL;
2031         struct nsproxy *new_nsproxy = NULL;
2032         int do_sysvsem = 0;
2033         int err;
2034 
2035         /*
2036          * If unsharing a user namespace must also unshare the thread group
2037          * and unshare the filesystem root and working directories.
2038          */
2039         if (unshare_flags & CLONE_NEWUSER)
2040                 unshare_flags |= CLONE_THREAD | CLONE_FS;
2041         /*
2042          * If unsharing vm, must also unshare signal handlers.
2043          */
2044         if (unshare_flags & CLONE_VM)
2045                 unshare_flags |= CLONE_SIGHAND;
2046         /*
2047          * If unsharing a signal handlers, must also unshare the signal queues.
2048          */
2049         if (unshare_flags & CLONE_SIGHAND)
2050                 unshare_flags |= CLONE_THREAD;
2051         /*
2052          * If unsharing namespace, must also unshare filesystem information.
2053          */
2054         if (unshare_flags & CLONE_NEWNS)
2055                 unshare_flags |= CLONE_FS;
2056 
2057         err = check_unshare_flags(unshare_flags);
2058         if (err)
2059                 goto bad_unshare_out;
2060         /*
2061          * CLONE_NEWIPC must also detach from the undolist: after switching
2062          * to a new ipc namespace, the semaphore arrays from the old
2063          * namespace are unreachable.
2064          */
2065         if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
2066                 do_sysvsem = 1;
2067         err = unshare_fs(unshare_flags, &new_fs);
2068         if (err)
2069                 goto bad_unshare_out;
2070         err = unshare_fd(unshare_flags, &new_fd);
2071         if (err)
2072                 goto bad_unshare_cleanup_fs;
2073         err = unshare_userns(unshare_flags, &new_cred);
2074         if (err)
2075                 goto bad_unshare_cleanup_fd;
2076         err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
2077                                          new_cred, new_fs);
2078         if (err)
2079                 goto bad_unshare_cleanup_cred;
2080 
2081         if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) {
2082                 if (do_sysvsem) {
2083                         /*
2084                          * CLONE_SYSVSEM is equivalent to sys_exit().
2085                          */
2086                         exit_sem(current);
2087                 }
2088                 if (unshare_flags & CLONE_NEWIPC) {
2089                         /* Orphan segments in old ns (see sem above). */
2090                         exit_shm(current);
2091                         shm_init_task(current);
2092                 }
2093 
2094                 if (new_nsproxy)
2095                         switch_task_namespaces(current, new_nsproxy);
2096 
2097                 task_lock(current);
2098 
2099                 if (new_fs) {
2100                         fs = current->fs;
2101                         spin_lock(&fs->lock);
2102                         current->fs = new_fs;
2103                         if (--fs->users)
2104                                 new_fs = NULL;
2105                         else
2106                                 new_fs = fs;
2107                         spin_unlock(&fs->lock);
2108                 }
2109 
2110                 if (new_fd) {
2111                         fd = current->files;
2112                         current->files = new_fd;
2113                         new_fd = fd;
2114                 }
2115 
2116                 task_unlock(current);
2117 
2118                 if (new_cred) {
2119                         /* Install the new user namespace */
2120                         commit_creds(new_cred);
2121                         new_cred = NULL;
2122                 }
2123         }
2124 
2125 bad_unshare_cleanup_cred:
2126         if (new_cred)
2127                 put_cred(new_cred);
2128 bad_unshare_cleanup_fd:
2129         if (new_fd)
2130                 put_files_struct(new_fd);
2131 
2132 bad_unshare_cleanup_fs:
2133         if (new_fs)
2134                 free_fs_struct(new_fs);
2135 
2136 bad_unshare_out:
2137         return err;
2138 }
2139 
2140 /*
2141  *      Helper to unshare the files of the current task.
2142  *      We don't want to expose copy_files internals to
2143  *      the exec layer of the kernel.
2144  */
2145 
2146 int unshare_files(struct files_struct **displaced)
2147 {
2148         struct task_struct *task = current;
2149         struct files_struct *copy = NULL;
2150         int error;
2151 
2152         error = unshare_fd(CLONE_FILES, &copy);
2153         if (error || !copy) {
2154                 *displaced = NULL;
2155                 return error;
2156         }
2157         *displaced = task->files;
2158         task_lock(task);
2159         task->files = copy;
2160         task_unlock(task);
2161         return 0;
2162 }
2163 
2164 int sysctl_max_threads(struct ctl_table *table, int write,
2165                        void __user *buffer, size_t *lenp, loff_t *ppos)
2166 {
2167         struct ctl_table t;
2168         int ret;
2169         int threads = max_threads;
2170         int min = MIN_THREADS;
2171         int max = MAX_THREADS;
2172 
2173         t = *table;
2174         t.data = &threads;
2175         t.extra1 = &min;
2176         t.extra2 = &max;
2177 
2178         ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos);
2179         if (ret || !write)
2180                 return ret;
2181 
2182         set_max_threads(threads);
2183 
2184         return 0;
2185 }
2186 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp