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

TOMOYO Linux Cross Reference
Linux/kernel/fork.c

Version: ~ [ linux-5.15-rc5 ] ~ [ linux-5.14.11 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.72 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.152 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.210 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.250 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.286 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.288 ] ~ [ 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 
 79 #include <asm/pgtable.h>
 80 #include <asm/pgalloc.h>
 81 #include <asm/uaccess.h>
 82 #include <asm/mmu_context.h>
 83 #include <asm/cacheflush.h>
 84 #include <asm/tlbflush.h>
 85 
 86 #include <trace/events/sched.h>
 87 
 88 #define CREATE_TRACE_POINTS
 89 #include <trace/events/task.h>
 90 
 91 /*
 92  * Minimum number of threads to boot the kernel
 93  */
 94 #define MIN_THREADS 20
 95 
 96 /*
 97  * Maximum number of threads
 98  */
 99 #define MAX_THREADS FUTEX_TID_MASK
100 
101 /*
102  * Protected counters by write_lock_irq(&tasklist_lock)
103  */
104 unsigned long total_forks;      /* Handle normal Linux uptimes. */
105 int nr_threads;                 /* The idle threads do not count.. */
106 
107 int max_threads;                /* tunable limit on nr_threads */
108 
109 DEFINE_PER_CPU(unsigned long, process_counts) = 0;
110 
111 __cacheline_aligned DEFINE_RWLOCK(tasklist_lock);  /* outer */
112 
113 #ifdef CONFIG_PROVE_RCU
114 int lockdep_tasklist_lock_is_held(void)
115 {
116         return lockdep_is_held(&tasklist_lock);
117 }
118 EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
119 #endif /* #ifdef CONFIG_PROVE_RCU */
120 
121 int nr_processes(void)
122 {
123         int cpu;
124         int total = 0;
125 
126         for_each_possible_cpu(cpu)
127                 total += per_cpu(process_counts, cpu);
128 
129         return total;
130 }
131 
132 void __weak arch_release_task_struct(struct task_struct *tsk)
133 {
134 }
135 
136 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
137 static struct kmem_cache *task_struct_cachep;
138 
139 static inline struct task_struct *alloc_task_struct_node(int node)
140 {
141         return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node);
142 }
143 
144 static inline void free_task_struct(struct task_struct *tsk)
145 {
146         kmem_cache_free(task_struct_cachep, tsk);
147 }
148 #endif
149 
150 void __weak arch_release_thread_info(struct thread_info *ti)
151 {
152 }
153 
154 #ifndef CONFIG_ARCH_THREAD_INFO_ALLOCATOR
155 
156 /*
157  * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a
158  * kmemcache based allocator.
159  */
160 # if THREAD_SIZE >= PAGE_SIZE
161 static struct thread_info *alloc_thread_info_node(struct task_struct *tsk,
162                                                   int node)
163 {
164         struct page *page = alloc_kmem_pages_node(node, THREADINFO_GFP,
165                                                   THREAD_SIZE_ORDER);
166 
167         return page ? page_address(page) : NULL;
168 }
169 
170 static inline void free_thread_info(struct thread_info *ti)
171 {
172         free_kmem_pages((unsigned long)ti, THREAD_SIZE_ORDER);
173 }
174 # else
175 static struct kmem_cache *thread_info_cache;
176 
177 static struct thread_info *alloc_thread_info_node(struct task_struct *tsk,
178                                                   int node)
179 {
180         return kmem_cache_alloc_node(thread_info_cache, THREADINFO_GFP, node);
181 }
182 
183 static void free_thread_info(struct thread_info *ti)
184 {
185         kmem_cache_free(thread_info_cache, ti);
186 }
187 
188 void thread_info_cache_init(void)
189 {
190         thread_info_cache = kmem_cache_create("thread_info", THREAD_SIZE,
191                                               THREAD_SIZE, 0, NULL);
192         BUG_ON(thread_info_cache == NULL);
193 }
194 # endif
195 #endif
196 
197 /* SLAB cache for signal_struct structures (tsk->signal) */
198 static struct kmem_cache *signal_cachep;
199 
200 /* SLAB cache for sighand_struct structures (tsk->sighand) */
201 struct kmem_cache *sighand_cachep;
202 
203 /* SLAB cache for files_struct structures (tsk->files) */
204 struct kmem_cache *files_cachep;
205 
206 /* SLAB cache for fs_struct structures (tsk->fs) */
207 struct kmem_cache *fs_cachep;
208 
209 /* SLAB cache for vm_area_struct structures */
210 struct kmem_cache *vm_area_cachep;
211 
212 /* SLAB cache for mm_struct structures (tsk->mm) */
213 static struct kmem_cache *mm_cachep;
214 
215 static void account_kernel_stack(struct thread_info *ti, int account)
216 {
217         struct zone *zone = page_zone(virt_to_page(ti));
218 
219         mod_zone_page_state(zone, NR_KERNEL_STACK, account);
220 }
221 
222 void free_task(struct task_struct *tsk)
223 {
224         account_kernel_stack(tsk->stack, -1);
225         arch_release_thread_info(tsk->stack);
226         free_thread_info(tsk->stack);
227         rt_mutex_debug_task_free(tsk);
228         ftrace_graph_exit_task(tsk);
229         put_seccomp_filter(tsk);
230         arch_release_task_struct(tsk);
231         free_task_struct(tsk);
232 }
233 EXPORT_SYMBOL(free_task);
234 
235 static inline void free_signal_struct(struct signal_struct *sig)
236 {
237         taskstats_tgid_free(sig);
238         sched_autogroup_exit(sig);
239         kmem_cache_free(signal_cachep, sig);
240 }
241 
242 static inline void put_signal_struct(struct signal_struct *sig)
243 {
244         if (atomic_dec_and_test(&sig->sigcnt))
245                 free_signal_struct(sig);
246 }
247 
248 void __put_task_struct(struct task_struct *tsk)
249 {
250         WARN_ON(!tsk->exit_state);
251         WARN_ON(atomic_read(&tsk->usage));
252         WARN_ON(tsk == current);
253 
254         task_numa_free(tsk);
255         security_task_free(tsk);
256         exit_creds(tsk);
257         delayacct_tsk_free(tsk);
258         put_signal_struct(tsk->signal);
259 
260         ccs_free_task_security(tsk);
261         if (!profile_handoff_task(tsk))
262                 free_task(tsk);
263 }
264 EXPORT_SYMBOL_GPL(__put_task_struct);
265 
266 void __init __weak arch_task_cache_init(void) { }
267 
268 /*
269  * set_max_threads
270  */
271 static void set_max_threads(unsigned int max_threads_suggested)
272 {
273         u64 threads;
274 
275         /*
276          * The number of threads shall be limited such that the thread
277          * structures may only consume a small part of the available memory.
278          */
279         if (fls64(totalram_pages) + fls64(PAGE_SIZE) > 64)
280                 threads = MAX_THREADS;
281         else
282                 threads = div64_u64((u64) totalram_pages * (u64) PAGE_SIZE,
283                                     (u64) THREAD_SIZE * 8UL);
284 
285         if (threads > max_threads_suggested)
286                 threads = max_threads_suggested;
287 
288         max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS);
289 }
290 
291 void __init fork_init(void)
292 {
293 #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
294 #ifndef ARCH_MIN_TASKALIGN
295 #define ARCH_MIN_TASKALIGN      L1_CACHE_BYTES
296 #endif
297         /* create a slab on which task_structs can be allocated */
298         task_struct_cachep =
299                 kmem_cache_create("task_struct", sizeof(struct task_struct),
300                         ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL);
301 #endif
302 
303         /* do the arch specific task caches init */
304         arch_task_cache_init();
305 
306         set_max_threads(MAX_THREADS);
307 
308         init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
309         init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
310         init_task.signal->rlim[RLIMIT_SIGPENDING] =
311                 init_task.signal->rlim[RLIMIT_NPROC];
312 }
313 
314 int __weak arch_dup_task_struct(struct task_struct *dst,
315                                                struct task_struct *src)
316 {
317         *dst = *src;
318         return 0;
319 }
320 
321 void set_task_stack_end_magic(struct task_struct *tsk)
322 {
323         unsigned long *stackend;
324 
325         stackend = end_of_stack(tsk);
326         *stackend = STACK_END_MAGIC;    /* for overflow detection */
327 }
328 
329 static struct task_struct *dup_task_struct(struct task_struct *orig)
330 {
331         struct task_struct *tsk;
332         struct thread_info *ti;
333         int node = tsk_fork_get_node(orig);
334         int err;
335 
336         tsk = alloc_task_struct_node(node);
337         if (!tsk)
338                 return NULL;
339 
340         ti = alloc_thread_info_node(tsk, node);
341         if (!ti)
342                 goto free_tsk;
343 
344         err = arch_dup_task_struct(tsk, orig);
345         if (err)
346                 goto free_ti;
347 
348         tsk->stack = ti;
349 #ifdef CONFIG_SECCOMP
350         /*
351          * We must handle setting up seccomp filters once we're under
352          * the sighand lock in case orig has changed between now and
353          * then. Until then, filter must be NULL to avoid messing up
354          * the usage counts on the error path calling free_task.
355          */
356         tsk->seccomp.filter = NULL;
357 #endif
358 
359         setup_thread_stack(tsk, orig);
360         clear_user_return_notifier(tsk);
361         clear_tsk_need_resched(tsk);
362         set_task_stack_end_magic(tsk);
363 
364 #ifdef CONFIG_CC_STACKPROTECTOR
365         tsk->stack_canary = get_random_long();
366 #endif
367 
368         /*
369          * One for us, one for whoever does the "release_task()" (usually
370          * parent)
371          */
372         atomic_set(&tsk->usage, 2);
373 #ifdef CONFIG_BLK_DEV_IO_TRACE
374         tsk->btrace_seq = 0;
375 #endif
376         tsk->splice_pipe = NULL;
377         tsk->task_frag.page = NULL;
378 
379         account_kernel_stack(ti, 1);
380 
381         return tsk;
382 
383 free_ti:
384         free_thread_info(ti);
385 free_tsk:
386         free_task_struct(tsk);
387         return NULL;
388 }
389 
390 #ifdef CONFIG_MMU
391 static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
392 {
393         struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
394         struct rb_node **rb_link, *rb_parent;
395         int retval;
396         unsigned long charge;
397 
398         uprobe_start_dup_mmap();
399         down_write(&oldmm->mmap_sem);
400         flush_cache_dup_mm(oldmm);
401         uprobe_dup_mmap(oldmm, mm);
402         /*
403          * Not linked in yet - no deadlock potential:
404          */
405         down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
406 
407         /* No ordering required: file already has been exposed. */
408         RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
409 
410         mm->total_vm = oldmm->total_vm;
411         mm->shared_vm = oldmm->shared_vm;
412         mm->exec_vm = oldmm->exec_vm;
413         mm->stack_vm = oldmm->stack_vm;
414 
415         rb_link = &mm->mm_rb.rb_node;
416         rb_parent = NULL;
417         pprev = &mm->mmap;
418         retval = ksm_fork(mm, oldmm);
419         if (retval)
420                 goto out;
421         retval = khugepaged_fork(mm, oldmm);
422         if (retval)
423                 goto out;
424 
425         prev = NULL;
426         for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
427                 struct file *file;
428 
429                 if (mpnt->vm_flags & VM_DONTCOPY) {
430                         vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
431                                                         -vma_pages(mpnt));
432                         continue;
433                 }
434                 charge = 0;
435                 if (mpnt->vm_flags & VM_ACCOUNT) {
436                         unsigned long len = vma_pages(mpnt);
437 
438                         if (security_vm_enough_memory_mm(oldmm, len)) /* sic */
439                                 goto fail_nomem;
440                         charge = len;
441                 }
442                 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
443                 if (!tmp)
444                         goto fail_nomem;
445                 *tmp = *mpnt;
446                 INIT_LIST_HEAD(&tmp->anon_vma_chain);
447                 retval = vma_dup_policy(mpnt, tmp);
448                 if (retval)
449                         goto fail_nomem_policy;
450                 tmp->vm_mm = mm;
451                 if (anon_vma_fork(tmp, mpnt))
452                         goto fail_nomem_anon_vma_fork;
453                 tmp->vm_flags &= ~VM_LOCKED;
454                 tmp->vm_next = tmp->vm_prev = NULL;
455                 file = tmp->vm_file;
456                 if (file) {
457                         struct inode *inode = file_inode(file);
458                         struct address_space *mapping = file->f_mapping;
459 
460                         get_file(file);
461                         if (tmp->vm_flags & VM_DENYWRITE)
462                                 atomic_dec(&inode->i_writecount);
463                         i_mmap_lock_write(mapping);
464                         if (tmp->vm_flags & VM_SHARED)
465                                 atomic_inc(&mapping->i_mmap_writable);
466                         flush_dcache_mmap_lock(mapping);
467                         /* insert tmp into the share list, just after mpnt */
468                         vma_interval_tree_insert_after(tmp, mpnt,
469                                         &mapping->i_mmap);
470                         flush_dcache_mmap_unlock(mapping);
471                         i_mmap_unlock_write(mapping);
472                 }
473 
474                 /*
475                  * Clear hugetlb-related page reserves for children. This only
476                  * affects MAP_PRIVATE mappings. Faults generated by the child
477                  * are not guaranteed to succeed, even if read-only
478                  */
479                 if (is_vm_hugetlb_page(tmp))
480                         reset_vma_resv_huge_pages(tmp);
481 
482                 /*
483                  * Link in the new vma and copy the page table entries.
484                  */
485                 *pprev = tmp;
486                 pprev = &tmp->vm_next;
487                 tmp->vm_prev = prev;
488                 prev = tmp;
489 
490                 __vma_link_rb(mm, tmp, rb_link, rb_parent);
491                 rb_link = &tmp->vm_rb.rb_right;
492                 rb_parent = &tmp->vm_rb;
493 
494                 mm->map_count++;
495                 retval = copy_page_range(mm, oldmm, mpnt);
496 
497                 if (tmp->vm_ops && tmp->vm_ops->open)
498                         tmp->vm_ops->open(tmp);
499 
500                 if (retval)
501                         goto out;
502         }
503         /* a new mm has just been created */
504         arch_dup_mmap(oldmm, mm);
505         retval = 0;
506 out:
507         up_write(&mm->mmap_sem);
508         flush_tlb_mm(oldmm);
509         up_write(&oldmm->mmap_sem);
510         uprobe_end_dup_mmap();
511         return retval;
512 fail_nomem_anon_vma_fork:
513         mpol_put(vma_policy(tmp));
514 fail_nomem_policy:
515         kmem_cache_free(vm_area_cachep, tmp);
516 fail_nomem:
517         retval = -ENOMEM;
518         vm_unacct_memory(charge);
519         goto out;
520 }
521 
522 static inline int mm_alloc_pgd(struct mm_struct *mm)
523 {
524         mm->pgd = pgd_alloc(mm);
525         if (unlikely(!mm->pgd))
526                 return -ENOMEM;
527         return 0;
528 }
529 
530 static inline void mm_free_pgd(struct mm_struct *mm)
531 {
532         pgd_free(mm, mm->pgd);
533 }
534 #else
535 static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
536 {
537         down_write(&oldmm->mmap_sem);
538         RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
539         up_write(&oldmm->mmap_sem);
540         return 0;
541 }
542 #define mm_alloc_pgd(mm)        (0)
543 #define mm_free_pgd(mm)
544 #endif /* CONFIG_MMU */
545 
546 __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
547 
548 #define allocate_mm()   (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
549 #define free_mm(mm)     (kmem_cache_free(mm_cachep, (mm)))
550 
551 static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
552 
553 static int __init coredump_filter_setup(char *s)
554 {
555         default_dump_filter =
556                 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
557                 MMF_DUMP_FILTER_MASK;
558         return 1;
559 }
560 
561 __setup("coredump_filter=", coredump_filter_setup);
562 
563 #include <linux/init_task.h>
564 
565 static void mm_init_aio(struct mm_struct *mm)
566 {
567 #ifdef CONFIG_AIO
568         spin_lock_init(&mm->ioctx_lock);
569         mm->ioctx_table = NULL;
570 #endif
571 }
572 
573 static void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
574 {
575 #ifdef CONFIG_MEMCG
576         mm->owner = p;
577 #endif
578 }
579 
580 static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p)
581 {
582         mm->mmap = NULL;
583         mm->mm_rb = RB_ROOT;
584         mm->vmacache_seqnum = 0;
585         atomic_set(&mm->mm_users, 1);
586         atomic_set(&mm->mm_count, 1);
587         init_rwsem(&mm->mmap_sem);
588         INIT_LIST_HEAD(&mm->mmlist);
589         mm->core_state = NULL;
590         atomic_long_set(&mm->nr_ptes, 0);
591         mm_nr_pmds_init(mm);
592         mm->map_count = 0;
593         mm->locked_vm = 0;
594         mm->pinned_vm = 0;
595         memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
596         spin_lock_init(&mm->page_table_lock);
597         mm_init_cpumask(mm);
598         mm_init_aio(mm);
599         mm_init_owner(mm, p);
600         mmu_notifier_mm_init(mm);
601         clear_tlb_flush_pending(mm);
602 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
603         mm->pmd_huge_pte = NULL;
604 #endif
605 
606         if (current->mm) {
607                 mm->flags = current->mm->flags & MMF_INIT_MASK;
608                 mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK;
609         } else {
610                 mm->flags = default_dump_filter;
611                 mm->def_flags = 0;
612         }
613 
614         if (mm_alloc_pgd(mm))
615                 goto fail_nopgd;
616 
617         if (init_new_context(p, mm))
618                 goto fail_nocontext;
619 
620         return mm;
621 
622 fail_nocontext:
623         mm_free_pgd(mm);
624 fail_nopgd:
625         free_mm(mm);
626         return NULL;
627 }
628 
629 static void check_mm(struct mm_struct *mm)
630 {
631         int i;
632 
633         for (i = 0; i < NR_MM_COUNTERS; i++) {
634                 long x = atomic_long_read(&mm->rss_stat.count[i]);
635 
636                 if (unlikely(x))
637                         printk(KERN_ALERT "BUG: Bad rss-counter state "
638                                           "mm:%p idx:%d val:%ld\n", mm, i, x);
639         }
640 
641         if (atomic_long_read(&mm->nr_ptes))
642                 pr_alert("BUG: non-zero nr_ptes on freeing mm: %ld\n",
643                                 atomic_long_read(&mm->nr_ptes));
644         if (mm_nr_pmds(mm))
645                 pr_alert("BUG: non-zero nr_pmds on freeing mm: %ld\n",
646                                 mm_nr_pmds(mm));
647 
648 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
649         VM_BUG_ON_MM(mm->pmd_huge_pte, mm);
650 #endif
651 }
652 
653 /*
654  * Allocate and initialize an mm_struct.
655  */
656 struct mm_struct *mm_alloc(void)
657 {
658         struct mm_struct *mm;
659 
660         mm = allocate_mm();
661         if (!mm)
662                 return NULL;
663 
664         memset(mm, 0, sizeof(*mm));
665         return mm_init(mm, current);
666 }
667 
668 /*
669  * Called when the last reference to the mm
670  * is dropped: either by a lazy thread or by
671  * mmput. Free the page directory and the mm.
672  */
673 void __mmdrop(struct mm_struct *mm)
674 {
675         BUG_ON(mm == &init_mm);
676         mm_free_pgd(mm);
677         destroy_context(mm);
678         mmu_notifier_mm_destroy(mm);
679         check_mm(mm);
680         free_mm(mm);
681 }
682 EXPORT_SYMBOL_GPL(__mmdrop);
683 
684 /*
685  * Decrement the use count and release all resources for an mm.
686  */
687 void mmput(struct mm_struct *mm)
688 {
689         might_sleep();
690 
691         if (atomic_dec_and_test(&mm->mm_users)) {
692                 uprobe_clear_state(mm);
693                 exit_aio(mm);
694                 ksm_exit(mm);
695                 khugepaged_exit(mm); /* must run before exit_mmap */
696                 exit_mmap(mm);
697                 set_mm_exe_file(mm, NULL);
698                 if (!list_empty(&mm->mmlist)) {
699                         spin_lock(&mmlist_lock);
700                         list_del(&mm->mmlist);
701                         spin_unlock(&mmlist_lock);
702                 }
703                 if (mm->binfmt)
704                         module_put(mm->binfmt->module);
705                 mmdrop(mm);
706         }
707 }
708 EXPORT_SYMBOL_GPL(mmput);
709 
710 /**
711  * set_mm_exe_file - change a reference to the mm's executable file
712  *
713  * This changes mm's executable file (shown as symlink /proc/[pid]/exe).
714  *
715  * Main users are mmput() and sys_execve(). Callers prevent concurrent
716  * invocations: in mmput() nobody alive left, in execve task is single
717  * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the
718  * mm->exe_file, but does so without using set_mm_exe_file() in order
719  * to do avoid the need for any locks.
720  */
721 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
722 {
723         struct file *old_exe_file;
724 
725         /*
726          * It is safe to dereference the exe_file without RCU as
727          * this function is only called if nobody else can access
728          * this mm -- see comment above for justification.
729          */
730         old_exe_file = rcu_dereference_raw(mm->exe_file);
731 
732         if (new_exe_file)
733                 get_file(new_exe_file);
734         rcu_assign_pointer(mm->exe_file, new_exe_file);
735         if (old_exe_file)
736                 fput(old_exe_file);
737 }
738 
739 /**
740  * get_mm_exe_file - acquire a reference to the mm's executable file
741  *
742  * Returns %NULL if mm has no associated executable file.
743  * User must release file via fput().
744  */
745 struct file *get_mm_exe_file(struct mm_struct *mm)
746 {
747         struct file *exe_file;
748 
749         rcu_read_lock();
750         exe_file = rcu_dereference(mm->exe_file);
751         if (exe_file && !get_file_rcu(exe_file))
752                 exe_file = NULL;
753         rcu_read_unlock();
754         return exe_file;
755 }
756 EXPORT_SYMBOL(get_mm_exe_file);
757 
758 /**
759  * get_task_mm - acquire a reference to the task's mm
760  *
761  * Returns %NULL if the task has no mm.  Checks PF_KTHREAD (meaning
762  * this kernel workthread has transiently adopted a user mm with use_mm,
763  * to do its AIO) is not set and if so returns a reference to it, after
764  * bumping up the use count.  User must release the mm via mmput()
765  * after use.  Typically used by /proc and ptrace.
766  */
767 struct mm_struct *get_task_mm(struct task_struct *task)
768 {
769         struct mm_struct *mm;
770 
771         task_lock(task);
772         mm = task->mm;
773         if (mm) {
774                 if (task->flags & PF_KTHREAD)
775                         mm = NULL;
776                 else
777                         atomic_inc(&mm->mm_users);
778         }
779         task_unlock(task);
780         return mm;
781 }
782 EXPORT_SYMBOL_GPL(get_task_mm);
783 
784 struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
785 {
786         struct mm_struct *mm;
787         int err;
788 
789         err =  mutex_lock_killable(&task->signal->cred_guard_mutex);
790         if (err)
791                 return ERR_PTR(err);
792 
793         mm = get_task_mm(task);
794         if (mm && mm != current->mm &&
795                         !ptrace_may_access(task, mode)) {
796                 mmput(mm);
797                 mm = ERR_PTR(-EACCES);
798         }
799         mutex_unlock(&task->signal->cred_guard_mutex);
800 
801         return mm;
802 }
803 
804 static void complete_vfork_done(struct task_struct *tsk)
805 {
806         struct completion *vfork;
807 
808         task_lock(tsk);
809         vfork = tsk->vfork_done;
810         if (likely(vfork)) {
811                 tsk->vfork_done = NULL;
812                 complete(vfork);
813         }
814         task_unlock(tsk);
815 }
816 
817 static int wait_for_vfork_done(struct task_struct *child,
818                                 struct completion *vfork)
819 {
820         int killed;
821 
822         freezer_do_not_count();
823         killed = wait_for_completion_killable(vfork);
824         freezer_count();
825 
826         if (killed) {
827                 task_lock(child);
828                 child->vfork_done = NULL;
829                 task_unlock(child);
830         }
831 
832         put_task_struct(child);
833         return killed;
834 }
835 
836 /* Please note the differences between mmput and mm_release.
837  * mmput is called whenever we stop holding onto a mm_struct,
838  * error success whatever.
839  *
840  * mm_release is called after a mm_struct has been removed
841  * from the current process.
842  *
843  * This difference is important for error handling, when we
844  * only half set up a mm_struct for a new process and need to restore
845  * the old one.  Because we mmput the new mm_struct before
846  * restoring the old one. . .
847  * Eric Biederman 10 January 1998
848  */
849 void mm_release(struct task_struct *tsk, struct mm_struct *mm)
850 {
851         /* Get rid of any futexes when releasing the mm */
852 #ifdef CONFIG_FUTEX
853         if (unlikely(tsk->robust_list)) {
854                 exit_robust_list(tsk);
855                 tsk->robust_list = NULL;
856         }
857 #ifdef CONFIG_COMPAT
858         if (unlikely(tsk->compat_robust_list)) {
859                 compat_exit_robust_list(tsk);
860                 tsk->compat_robust_list = NULL;
861         }
862 #endif
863         if (unlikely(!list_empty(&tsk->pi_state_list)))
864                 exit_pi_state_list(tsk);
865 #endif
866 
867         uprobe_free_utask(tsk);
868 
869         /* Get rid of any cached register state */
870         deactivate_mm(tsk, mm);
871 
872         /*
873          * If we're exiting normally, clear a user-space tid field if
874          * requested.  We leave this alone when dying by signal, to leave
875          * the value intact in a core dump, and to save the unnecessary
876          * trouble, say, a killed vfork parent shouldn't touch this mm.
877          * Userland only wants this done for a sys_exit.
878          */
879         if (tsk->clear_child_tid) {
880                 if (!(tsk->flags & PF_SIGNALED) &&
881                     atomic_read(&mm->mm_users) > 1) {
882                         /*
883                          * We don't check the error code - if userspace has
884                          * not set up a proper pointer then tough luck.
885                          */
886                         put_user(0, tsk->clear_child_tid);
887                         sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
888                                         1, NULL, NULL, 0);
889                 }
890                 tsk->clear_child_tid = NULL;
891         }
892 
893         /*
894          * All done, finally we can wake up parent and return this mm to him.
895          * Also kthread_stop() uses this completion for synchronization.
896          */
897         if (tsk->vfork_done)
898                 complete_vfork_done(tsk);
899 }
900 
901 /*
902  * Allocate a new mm structure and copy contents from the
903  * mm structure of the passed in task structure.
904  */
905 static struct mm_struct *dup_mm(struct task_struct *tsk)
906 {
907         struct mm_struct *mm, *oldmm = current->mm;
908         int err;
909 
910         mm = allocate_mm();
911         if (!mm)
912                 goto fail_nomem;
913 
914         memcpy(mm, oldmm, sizeof(*mm));
915 
916         if (!mm_init(mm, tsk))
917                 goto fail_nomem;
918 
919         err = dup_mmap(mm, oldmm);
920         if (err)
921                 goto free_pt;
922 
923         mm->hiwater_rss = get_mm_rss(mm);
924         mm->hiwater_vm = mm->total_vm;
925 
926         if (mm->binfmt && !try_module_get(mm->binfmt->module))
927                 goto free_pt;
928 
929         return mm;
930 
931 free_pt:
932         /* don't put binfmt in mmput, we haven't got module yet */
933         mm->binfmt = NULL;
934         mmput(mm);
935 
936 fail_nomem:
937         return NULL;
938 }
939 
940 static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
941 {
942         struct mm_struct *mm, *oldmm;
943         int retval;
944 
945         tsk->min_flt = tsk->maj_flt = 0;
946         tsk->nvcsw = tsk->nivcsw = 0;
947 #ifdef CONFIG_DETECT_HUNG_TASK
948         tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
949 #endif
950 
951         tsk->mm = NULL;
952         tsk->active_mm = NULL;
953 
954         /*
955          * Are we cloning a kernel thread?
956          *
957          * We need to steal a active VM for that..
958          */
959         oldmm = current->mm;
960         if (!oldmm)
961                 return 0;
962 
963         /* initialize the new vmacache entries */
964         vmacache_flush(tsk);
965 
966         if (clone_flags & CLONE_VM) {
967                 atomic_inc(&oldmm->mm_users);
968                 mm = oldmm;
969                 goto good_mm;
970         }
971 
972         retval = -ENOMEM;
973         mm = dup_mm(tsk);
974         if (!mm)
975                 goto fail_nomem;
976 
977 good_mm:
978         tsk->mm = mm;
979         tsk->active_mm = mm;
980         return 0;
981 
982 fail_nomem:
983         return retval;
984 }
985 
986 static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
987 {
988         struct fs_struct *fs = current->fs;
989         if (clone_flags & CLONE_FS) {
990                 /* tsk->fs is already what we want */
991                 spin_lock(&fs->lock);
992                 if (fs->in_exec) {
993                         spin_unlock(&fs->lock);
994                         return -EAGAIN;
995                 }
996                 fs->users++;
997                 spin_unlock(&fs->lock);
998                 return 0;
999         }
1000         tsk->fs = copy_fs_struct(fs);
1001         if (!tsk->fs)
1002                 return -ENOMEM;
1003         return 0;
1004 }
1005 
1006 static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
1007 {
1008         struct files_struct *oldf, *newf;
1009         int error = 0;
1010 
1011         /*
1012          * A background process may not have any files ...
1013          */
1014         oldf = current->files;
1015         if (!oldf)
1016                 goto out;
1017 
1018         if (clone_flags & CLONE_FILES) {
1019                 atomic_inc(&oldf->count);
1020                 goto out;
1021         }
1022 
1023         newf = dup_fd(oldf, &error);
1024         if (!newf)
1025                 goto out;
1026 
1027         tsk->files = newf;
1028         error = 0;
1029 out:
1030         return error;
1031 }
1032 
1033 static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
1034 {
1035 #ifdef CONFIG_BLOCK
1036         struct io_context *ioc = current->io_context;
1037         struct io_context *new_ioc;
1038 
1039         if (!ioc)
1040                 return 0;
1041         /*
1042          * Share io context with parent, if CLONE_IO is set
1043          */
1044         if (clone_flags & CLONE_IO) {
1045                 ioc_task_link(ioc);
1046                 tsk->io_context = ioc;
1047         } else if (ioprio_valid(ioc->ioprio)) {
1048                 new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE);
1049                 if (unlikely(!new_ioc))
1050                         return -ENOMEM;
1051 
1052                 new_ioc->ioprio = ioc->ioprio;
1053                 put_io_context(new_ioc);
1054         }
1055 #endif
1056         return 0;
1057 }
1058 
1059 static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
1060 {
1061         struct sighand_struct *sig;
1062 
1063         if (clone_flags & CLONE_SIGHAND) {
1064                 atomic_inc(&current->sighand->count);
1065                 return 0;
1066         }
1067         sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
1068         rcu_assign_pointer(tsk->sighand, sig);
1069         if (!sig)
1070                 return -ENOMEM;
1071         atomic_set(&sig->count, 1);
1072         memcpy(sig->action, current->sighand->action, sizeof(sig->action));
1073         return 0;
1074 }
1075 
1076 void __cleanup_sighand(struct sighand_struct *sighand)
1077 {
1078         if (atomic_dec_and_test(&sighand->count)) {
1079                 signalfd_cleanup(sighand);
1080                 /*
1081                  * sighand_cachep is SLAB_DESTROY_BY_RCU so we can free it
1082                  * without an RCU grace period, see __lock_task_sighand().
1083                  */
1084                 kmem_cache_free(sighand_cachep, sighand);
1085         }
1086 }
1087 
1088 /*
1089  * Initialize POSIX timer handling for a thread group.
1090  */
1091 static void posix_cpu_timers_init_group(struct signal_struct *sig)
1092 {
1093         unsigned long cpu_limit;
1094 
1095         /* Thread group counters. */
1096         thread_group_cputime_init(sig);
1097 
1098         cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
1099         if (cpu_limit != RLIM_INFINITY) {
1100                 sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit);
1101                 sig->cputimer.running = 1;
1102         }
1103 
1104         /* The timer lists. */
1105         INIT_LIST_HEAD(&sig->cpu_timers[0]);
1106         INIT_LIST_HEAD(&sig->cpu_timers[1]);
1107         INIT_LIST_HEAD(&sig->cpu_timers[2]);
1108 }
1109 
1110 static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
1111 {
1112         struct signal_struct *sig;
1113 
1114         if (clone_flags & CLONE_THREAD)
1115                 return 0;
1116 
1117         sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
1118         tsk->signal = sig;
1119         if (!sig)
1120                 return -ENOMEM;
1121 
1122         sig->nr_threads = 1;
1123         atomic_set(&sig->live, 1);
1124         atomic_set(&sig->sigcnt, 1);
1125 
1126         /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */
1127         sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node);
1128         tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head);
1129 
1130         init_waitqueue_head(&sig->wait_chldexit);
1131         sig->curr_target = tsk;
1132         init_sigpending(&sig->shared_pending);
1133         INIT_LIST_HEAD(&sig->posix_timers);
1134         seqlock_init(&sig->stats_lock);
1135 
1136         hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
1137         sig->real_timer.function = it_real_fn;
1138 
1139         task_lock(current->group_leader);
1140         memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
1141         task_unlock(current->group_leader);
1142 
1143         posix_cpu_timers_init_group(sig);
1144 
1145         tty_audit_fork(sig);
1146         sched_autogroup_fork(sig);
1147 
1148 #ifdef CONFIG_CGROUPS
1149         init_rwsem(&sig->group_rwsem);
1150 #endif
1151 
1152         sig->oom_score_adj = current->signal->oom_score_adj;
1153         sig->oom_score_adj_min = current->signal->oom_score_adj_min;
1154 
1155         sig->has_child_subreaper = current->signal->has_child_subreaper ||
1156                                    current->signal->is_child_subreaper;
1157 
1158         mutex_init(&sig->cred_guard_mutex);
1159 
1160         return 0;
1161 }
1162 
1163 static void copy_seccomp(struct task_struct *p)
1164 {
1165 #ifdef CONFIG_SECCOMP
1166         /*
1167          * Must be called with sighand->lock held, which is common to
1168          * all threads in the group. Holding cred_guard_mutex is not
1169          * needed because this new task is not yet running and cannot
1170          * be racing exec.
1171          */
1172         assert_spin_locked(&current->sighand->siglock);
1173 
1174         /* Ref-count the new filter user, and assign it. */
1175         get_seccomp_filter(current);
1176         p->seccomp = current->seccomp;
1177 
1178         /*
1179          * Explicitly enable no_new_privs here in case it got set
1180          * between the task_struct being duplicated and holding the
1181          * sighand lock. The seccomp state and nnp must be in sync.
1182          */
1183         if (task_no_new_privs(current))
1184                 task_set_no_new_privs(p);
1185 
1186         /*
1187          * If the parent gained a seccomp mode after copying thread
1188          * flags and between before we held the sighand lock, we have
1189          * to manually enable the seccomp thread flag here.
1190          */
1191         if (p->seccomp.mode != SECCOMP_MODE_DISABLED)
1192                 set_tsk_thread_flag(p, TIF_SECCOMP);
1193 #endif
1194 }
1195 
1196 SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
1197 {
1198         current->clear_child_tid = tidptr;
1199 
1200         return task_pid_vnr(current);
1201 }
1202 
1203 static void rt_mutex_init_task(struct task_struct *p)
1204 {
1205         raw_spin_lock_init(&p->pi_lock);
1206 #ifdef CONFIG_RT_MUTEXES
1207         p->pi_waiters = RB_ROOT;
1208         p->pi_waiters_leftmost = NULL;
1209         p->pi_blocked_on = NULL;
1210 #endif
1211 }
1212 
1213 /*
1214  * Initialize POSIX timer handling for a single task.
1215  */
1216 static void posix_cpu_timers_init(struct task_struct *tsk)
1217 {
1218         tsk->cputime_expires.prof_exp = 0;
1219         tsk->cputime_expires.virt_exp = 0;
1220         tsk->cputime_expires.sched_exp = 0;
1221         INIT_LIST_HEAD(&tsk->cpu_timers[0]);
1222         INIT_LIST_HEAD(&tsk->cpu_timers[1]);
1223         INIT_LIST_HEAD(&tsk->cpu_timers[2]);
1224 }
1225 
1226 static inline void
1227 init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid)
1228 {
1229          task->pids[type].pid = pid;
1230 }
1231 
1232 /*
1233  * This creates a new process as a copy of the old one,
1234  * but does not actually start it yet.
1235  *
1236  * It copies the registers, and all the appropriate
1237  * parts of the process environment (as per the clone
1238  * flags). The actual kick-off is left to the caller.
1239  */
1240 static struct task_struct *copy_process(unsigned long clone_flags,
1241                                         unsigned long stack_start,
1242                                         unsigned long stack_size,
1243                                         int __user *child_tidptr,
1244                                         struct pid *pid,
1245                                         int trace)
1246 {
1247         int retval;
1248         struct task_struct *p;
1249 
1250         if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
1251                 return ERR_PTR(-EINVAL);
1252 
1253         if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS))
1254                 return ERR_PTR(-EINVAL);
1255 
1256         /*
1257          * Thread groups must share signals as well, and detached threads
1258          * can only be started up within the thread group.
1259          */
1260         if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1261                 return ERR_PTR(-EINVAL);
1262 
1263         /*
1264          * Shared signal handlers imply shared VM. By way of the above,
1265          * thread groups also imply shared VM. Blocking this case allows
1266          * for various simplifications in other code.
1267          */
1268         if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1269                 return ERR_PTR(-EINVAL);
1270 
1271         /*
1272          * Siblings of global init remain as zombies on exit since they are
1273          * not reaped by their parent (swapper). To solve this and to avoid
1274          * multi-rooted process trees, prevent global and container-inits
1275          * from creating siblings.
1276          */
1277         if ((clone_flags & CLONE_PARENT) &&
1278                                 current->signal->flags & SIGNAL_UNKILLABLE)
1279                 return ERR_PTR(-EINVAL);
1280 
1281         /*
1282          * If the new process will be in a different pid or user namespace
1283          * do not allow it to share a thread group or signal handlers or
1284          * parent with the forking task.
1285          */
1286         if (clone_flags & CLONE_SIGHAND) {
1287                 if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) ||
1288                     (task_active_pid_ns(current) !=
1289                                 current->nsproxy->pid_ns_for_children))
1290                         return ERR_PTR(-EINVAL);
1291         }
1292 
1293         retval = security_task_create(clone_flags);
1294         if (retval)
1295                 goto fork_out;
1296 
1297         retval = -ENOMEM;
1298         p = dup_task_struct(current);
1299         if (!p)
1300                 goto fork_out;
1301 
1302         ftrace_graph_init_task(p);
1303 
1304         rt_mutex_init_task(p);
1305 
1306 #ifdef CONFIG_PROVE_LOCKING
1307         DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1308         DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1309 #endif
1310         retval = -EAGAIN;
1311         if (atomic_read(&p->real_cred->user->processes) >=
1312                         task_rlimit(p, RLIMIT_NPROC)) {
1313                 if (p->real_cred->user != INIT_USER &&
1314                     !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN))
1315                         goto bad_fork_free;
1316         }
1317         current->flags &= ~PF_NPROC_EXCEEDED;
1318 
1319         retval = copy_creds(p, clone_flags);
1320         if (retval < 0)
1321                 goto bad_fork_free;
1322 
1323         /*
1324          * If multiple threads are within copy_process(), then this check
1325          * triggers too late. This doesn't hurt, the check is only there
1326          * to stop root fork bombs.
1327          */
1328         retval = -EAGAIN;
1329         if (nr_threads >= max_threads)
1330                 goto bad_fork_cleanup_count;
1331 
1332         delayacct_tsk_init(p);  /* Must remain after dup_task_struct() */
1333         p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER);
1334         p->flags |= PF_FORKNOEXEC;
1335         INIT_LIST_HEAD(&p->children);
1336         INIT_LIST_HEAD(&p->sibling);
1337         rcu_copy_process(p);
1338         p->vfork_done = NULL;
1339         spin_lock_init(&p->alloc_lock);
1340 
1341         init_sigpending(&p->pending);
1342 
1343         p->utime = p->stime = p->gtime = 0;
1344         p->utimescaled = p->stimescaled = 0;
1345 #ifndef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE
1346         p->prev_cputime.utime = p->prev_cputime.stime = 0;
1347 #endif
1348 #ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
1349         seqlock_init(&p->vtime_seqlock);
1350         p->vtime_snap = 0;
1351         p->vtime_snap_whence = VTIME_SLEEPING;
1352 #endif
1353 
1354 #if defined(SPLIT_RSS_COUNTING)
1355         memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1356 #endif
1357 
1358         p->default_timer_slack_ns = current->timer_slack_ns;
1359 
1360         task_io_accounting_init(&p->ioac);
1361         acct_clear_integrals(p);
1362 
1363         posix_cpu_timers_init(p);
1364 
1365         p->start_time = ktime_get_ns();
1366         p->real_start_time = ktime_get_boot_ns();
1367         p->io_context = NULL;
1368         p->audit_context = NULL;
1369         if (clone_flags & CLONE_THREAD)
1370                 threadgroup_change_begin(current);
1371         cgroup_fork(p);
1372 #ifdef CONFIG_NUMA
1373         p->mempolicy = mpol_dup(p->mempolicy);
1374         if (IS_ERR(p->mempolicy)) {
1375                 retval = PTR_ERR(p->mempolicy);
1376                 p->mempolicy = NULL;
1377                 goto bad_fork_cleanup_threadgroup_lock;
1378         }
1379 #endif
1380 #ifdef CONFIG_CPUSETS
1381         p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
1382         p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
1383         seqcount_init(&p->mems_allowed_seq);
1384 #endif
1385 #ifdef CONFIG_TRACE_IRQFLAGS
1386         p->irq_events = 0;
1387         p->hardirqs_enabled = 0;
1388         p->hardirq_enable_ip = 0;
1389         p->hardirq_enable_event = 0;
1390         p->hardirq_disable_ip = _THIS_IP_;
1391         p->hardirq_disable_event = 0;
1392         p->softirqs_enabled = 1;
1393         p->softirq_enable_ip = _THIS_IP_;
1394         p->softirq_enable_event = 0;
1395         p->softirq_disable_ip = 0;
1396         p->softirq_disable_event = 0;
1397         p->hardirq_context = 0;
1398         p->softirq_context = 0;
1399 #endif
1400 #ifdef CONFIG_LOCKDEP
1401         p->lockdep_depth = 0; /* no locks held yet */
1402         p->curr_chain_key = 0;
1403         p->lockdep_recursion = 0;
1404 #endif
1405 
1406 #ifdef CONFIG_DEBUG_MUTEXES
1407         p->blocked_on = NULL; /* not blocked yet */
1408 #endif
1409 #ifdef CONFIG_BCACHE
1410         p->sequential_io        = 0;
1411         p->sequential_io_avg    = 0;
1412 #endif
1413 
1414         /* Perform scheduler related setup. Assign this task to a CPU. */
1415         retval = sched_fork(clone_flags, p);
1416         if (retval)
1417                 goto bad_fork_cleanup_policy;
1418 
1419         retval = perf_event_init_task(p);
1420         if (retval)
1421                 goto bad_fork_cleanup_policy;
1422         retval = audit_alloc(p);
1423         if (retval)
1424                 goto bad_fork_cleanup_perf;
1425         /* copy all the process information */
1426         shm_init_task(p);
1427         retval = ccs_alloc_task_security(p);
1428         if (retval)
1429                 goto bad_fork_cleanup_audit;
1430         retval = copy_semundo(clone_flags, p);
1431         if (retval)
1432                 goto bad_fork_cleanup_audit;
1433         retval = copy_files(clone_flags, p);
1434         if (retval)
1435                 goto bad_fork_cleanup_semundo;
1436         retval = copy_fs(clone_flags, p);
1437         if (retval)
1438                 goto bad_fork_cleanup_files;
1439         retval = copy_sighand(clone_flags, p);
1440         if (retval)
1441                 goto bad_fork_cleanup_fs;
1442         retval = copy_signal(clone_flags, p);
1443         if (retval)
1444                 goto bad_fork_cleanup_sighand;
1445         retval = copy_mm(clone_flags, p);
1446         if (retval)
1447                 goto bad_fork_cleanup_signal;
1448         retval = copy_namespaces(clone_flags, p);
1449         if (retval)
1450                 goto bad_fork_cleanup_mm;
1451         retval = copy_io(clone_flags, p);
1452         if (retval)
1453                 goto bad_fork_cleanup_namespaces;
1454         retval = copy_thread(clone_flags, stack_start, stack_size, p);
1455         if (retval)
1456                 goto bad_fork_cleanup_io;
1457 
1458         if (pid != &init_struct_pid) {
1459                 pid = alloc_pid(p->nsproxy->pid_ns_for_children);
1460                 if (IS_ERR(pid)) {
1461                         retval = PTR_ERR(pid);
1462                         goto bad_fork_cleanup_io;
1463                 }
1464         }
1465 
1466         p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1467         /*
1468          * Clear TID on mm_release()?
1469          */
1470         p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL;
1471 #ifdef CONFIG_BLOCK
1472         p->plug = NULL;
1473 #endif
1474 #ifdef CONFIG_FUTEX
1475         p->robust_list = NULL;
1476 #ifdef CONFIG_COMPAT
1477         p->compat_robust_list = NULL;
1478 #endif
1479         INIT_LIST_HEAD(&p->pi_state_list);
1480         p->pi_state_cache = NULL;
1481 #endif
1482         /*
1483          * sigaltstack should be cleared when sharing the same VM
1484          */
1485         if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1486                 p->sas_ss_sp = p->sas_ss_size = 0;
1487 
1488         /*
1489          * Syscall tracing and stepping should be turned off in the
1490          * child regardless of CLONE_PTRACE.
1491          */
1492         user_disable_single_step(p);
1493         clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
1494 #ifdef TIF_SYSCALL_EMU
1495         clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1496 #endif
1497         clear_all_latency_tracing(p);
1498 
1499         /* ok, now we should be set up.. */
1500         p->pid = pid_nr(pid);
1501         if (clone_flags & CLONE_THREAD) {
1502                 p->exit_signal = -1;
1503                 p->group_leader = current->group_leader;
1504                 p->tgid = current->tgid;
1505         } else {
1506                 if (clone_flags & CLONE_PARENT)
1507                         p->exit_signal = current->group_leader->exit_signal;
1508                 else
1509                         p->exit_signal = (clone_flags & CSIGNAL);
1510                 p->group_leader = p;
1511                 p->tgid = p->pid;
1512         }
1513 
1514         p->nr_dirtied = 0;
1515         p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
1516         p->dirty_paused_when = 0;
1517 
1518         p->pdeath_signal = 0;
1519         INIT_LIST_HEAD(&p->thread_group);
1520         p->task_works = NULL;
1521 
1522         /*
1523          * Make it visible to the rest of the system, but dont wake it up yet.
1524          * Need tasklist lock for parent etc handling!
1525          */
1526         write_lock_irq(&tasklist_lock);
1527 
1528         /* CLONE_PARENT re-uses the old parent */
1529         if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
1530                 p->real_parent = current->real_parent;
1531                 p->parent_exec_id = current->parent_exec_id;
1532         } else {
1533                 p->real_parent = current;
1534                 p->parent_exec_id = current->self_exec_id;
1535         }
1536 
1537         spin_lock(&current->sighand->siglock);
1538 
1539         /*
1540          * Copy seccomp details explicitly here, in case they were changed
1541          * before holding sighand lock.
1542          */
1543         copy_seccomp(p);
1544 
1545         /*
1546          * Process group and session signals need to be delivered to just the
1547          * parent before the fork or both the parent and the child after the
1548          * fork. Restart if a signal comes in before we add the new process to
1549          * it's process group.
1550          * A fatal signal pending means that current will exit, so the new
1551          * thread can't slip out of an OOM kill (or normal SIGKILL).
1552         */
1553         recalc_sigpending();
1554         if (signal_pending(current)) {
1555                 spin_unlock(&current->sighand->siglock);
1556                 write_unlock_irq(&tasklist_lock);
1557                 retval = -ERESTARTNOINTR;
1558                 goto bad_fork_free_pid;
1559         }
1560 
1561         if (likely(p->pid)) {
1562                 ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
1563 
1564                 init_task_pid(p, PIDTYPE_PID, pid);
1565                 if (thread_group_leader(p)) {
1566                         init_task_pid(p, PIDTYPE_PGID, task_pgrp(current));
1567                         init_task_pid(p, PIDTYPE_SID, task_session(current));
1568 
1569                         if (is_child_reaper(pid)) {
1570                                 ns_of_pid(pid)->child_reaper = p;
1571                                 p->signal->flags |= SIGNAL_UNKILLABLE;
1572                         }
1573 
1574                         p->signal->leader_pid = pid;
1575                         p->signal->tty = tty_kref_get(current->signal->tty);
1576                         list_add_tail(&p->sibling, &p->real_parent->children);
1577                         list_add_tail_rcu(&p->tasks, &init_task.tasks);
1578                         attach_pid(p, PIDTYPE_PGID);
1579                         attach_pid(p, PIDTYPE_SID);
1580                         __this_cpu_inc(process_counts);
1581                 } else {
1582                         current->signal->nr_threads++;
1583                         atomic_inc(&current->signal->live);
1584                         atomic_inc(&current->signal->sigcnt);
1585                         list_add_tail_rcu(&p->thread_group,
1586                                           &p->group_leader->thread_group);
1587                         list_add_tail_rcu(&p->thread_node,
1588                                           &p->signal->thread_head);
1589                 }
1590                 attach_pid(p, PIDTYPE_PID);
1591                 nr_threads++;
1592         }
1593 
1594         total_forks++;
1595         spin_unlock(&current->sighand->siglock);
1596         syscall_tracepoint_update(p);
1597         write_unlock_irq(&tasklist_lock);
1598 
1599         proc_fork_connector(p);
1600         cgroup_post_fork(p);
1601         if (clone_flags & CLONE_THREAD)
1602                 threadgroup_change_end(current);
1603         perf_event_fork(p);
1604 
1605         trace_task_newtask(p, clone_flags);
1606         uprobe_copy_process(p, clone_flags);
1607 
1608         return p;
1609 
1610 bad_fork_free_pid:
1611         if (pid != &init_struct_pid)
1612                 free_pid(pid);
1613 bad_fork_cleanup_io:
1614         if (p->io_context)
1615                 exit_io_context(p);
1616 bad_fork_cleanup_namespaces:
1617         exit_task_namespaces(p);
1618 bad_fork_cleanup_mm:
1619         if (p->mm)
1620                 mmput(p->mm);
1621 bad_fork_cleanup_signal:
1622         if (!(clone_flags & CLONE_THREAD))
1623                 free_signal_struct(p->signal);
1624 bad_fork_cleanup_sighand:
1625         __cleanup_sighand(p->sighand);
1626 bad_fork_cleanup_fs:
1627         exit_fs(p); /* blocking */
1628 bad_fork_cleanup_files:
1629         exit_files(p); /* blocking */
1630 bad_fork_cleanup_semundo:
1631         exit_sem(p);
1632 bad_fork_cleanup_audit:
1633         audit_free(p);
1634         ccs_free_task_security(p);
1635 bad_fork_cleanup_perf:
1636         perf_event_free_task(p);
1637 bad_fork_cleanup_policy:
1638 #ifdef CONFIG_NUMA
1639         mpol_put(p->mempolicy);
1640 bad_fork_cleanup_threadgroup_lock:
1641 #endif
1642         if (clone_flags & CLONE_THREAD)
1643                 threadgroup_change_end(current);
1644         delayacct_tsk_free(p);
1645 bad_fork_cleanup_count:
1646         atomic_dec(&p->cred->user->processes);
1647         exit_creds(p);
1648 bad_fork_free:
1649         free_task(p);
1650 fork_out:
1651         return ERR_PTR(retval);
1652 }
1653 
1654 static inline void init_idle_pids(struct pid_link *links)
1655 {
1656         enum pid_type type;
1657 
1658         for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1659                 INIT_HLIST_NODE(&links[type].node); /* not really needed */
1660                 links[type].pid = &init_struct_pid;
1661         }
1662 }
1663 
1664 struct task_struct *fork_idle(int cpu)
1665 {
1666         struct task_struct *task;
1667         task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0);
1668         if (!IS_ERR(task)) {
1669                 init_idle_pids(task->pids);
1670                 init_idle(task, cpu);
1671         }
1672 
1673         return task;
1674 }
1675 
1676 /*
1677  *  Ok, this is the main fork-routine.
1678  *
1679  * It copies the process, and if successful kick-starts
1680  * it and waits for it to finish using the VM if required.
1681  */
1682 long do_fork(unsigned long clone_flags,
1683               unsigned long stack_start,
1684               unsigned long stack_size,
1685               int __user *parent_tidptr,
1686               int __user *child_tidptr)
1687 {
1688         struct task_struct *p;
1689         int trace = 0;
1690         long nr;
1691 
1692         /*
1693          * Determine whether and which event to report to ptracer.  When
1694          * called from kernel_thread or CLONE_UNTRACED is explicitly
1695          * requested, no event is reported; otherwise, report if the event
1696          * for the type of forking is enabled.
1697          */
1698         if (!(clone_flags & CLONE_UNTRACED)) {
1699                 if (clone_flags & CLONE_VFORK)
1700                         trace = PTRACE_EVENT_VFORK;
1701                 else if ((clone_flags & CSIGNAL) != SIGCHLD)
1702                         trace = PTRACE_EVENT_CLONE;
1703                 else
1704                         trace = PTRACE_EVENT_FORK;
1705 
1706                 if (likely(!ptrace_event_enabled(current, trace)))
1707                         trace = 0;
1708         }
1709 
1710         p = copy_process(clone_flags, stack_start, stack_size,
1711                          child_tidptr, NULL, trace);
1712         /*
1713          * Do this prior waking up the new thread - the thread pointer
1714          * might get invalid after that point, if the thread exits quickly.
1715          */
1716         if (!IS_ERR(p)) {
1717                 struct completion vfork;
1718                 struct pid *pid;
1719 
1720                 trace_sched_process_fork(current, p);
1721 
1722                 pid = get_task_pid(p, PIDTYPE_PID);
1723                 nr = pid_vnr(pid);
1724 
1725                 if (clone_flags & CLONE_PARENT_SETTID)
1726                         put_user(nr, parent_tidptr);
1727 
1728                 if (clone_flags & CLONE_VFORK) {
1729                         p->vfork_done = &vfork;
1730                         init_completion(&vfork);
1731                         get_task_struct(p);
1732                 }
1733 
1734                 wake_up_new_task(p);
1735 
1736                 /* forking complete and child started to run, tell ptracer */
1737                 if (unlikely(trace))
1738                         ptrace_event_pid(trace, pid);
1739 
1740                 if (clone_flags & CLONE_VFORK) {
1741                         if (!wait_for_vfork_done(p, &vfork))
1742                                 ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid);
1743                 }
1744 
1745                 put_pid(pid);
1746         } else {
1747                 nr = PTR_ERR(p);
1748         }
1749         return nr;
1750 }
1751 
1752 /*
1753  * Create a kernel thread.
1754  */
1755 pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
1756 {
1757         return do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn,
1758                 (unsigned long)arg, NULL, NULL);
1759 }
1760 
1761 #ifdef __ARCH_WANT_SYS_FORK
1762 SYSCALL_DEFINE0(fork)
1763 {
1764 #ifdef CONFIG_MMU
1765         return do_fork(SIGCHLD, 0, 0, NULL, NULL);
1766 #else
1767         /* can not support in nommu mode */
1768         return -EINVAL;
1769 #endif
1770 }
1771 #endif
1772 
1773 #ifdef __ARCH_WANT_SYS_VFORK
1774 SYSCALL_DEFINE0(vfork)
1775 {
1776         return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0,
1777                         0, NULL, NULL);
1778 }
1779 #endif
1780 
1781 #ifdef __ARCH_WANT_SYS_CLONE
1782 #ifdef CONFIG_CLONE_BACKWARDS
1783 SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
1784                  int __user *, parent_tidptr,
1785                  int, tls_val,
1786                  int __user *, child_tidptr)
1787 #elif defined(CONFIG_CLONE_BACKWARDS2)
1788 SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags,
1789                  int __user *, parent_tidptr,
1790                  int __user *, child_tidptr,
1791                  int, tls_val)
1792 #elif defined(CONFIG_CLONE_BACKWARDS3)
1793 SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp,
1794                 int, stack_size,
1795                 int __user *, parent_tidptr,
1796                 int __user *, child_tidptr,
1797                 int, tls_val)
1798 #else
1799 SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
1800                  int __user *, parent_tidptr,
1801                  int __user *, child_tidptr,
1802                  int, tls_val)
1803 #endif
1804 {
1805         return do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr);
1806 }
1807 #endif
1808 
1809 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1810 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1811 #endif
1812 
1813 static void sighand_ctor(void *data)
1814 {
1815         struct sighand_struct *sighand = data;
1816 
1817         spin_lock_init(&sighand->siglock);
1818         init_waitqueue_head(&sighand->signalfd_wqh);
1819 }
1820 
1821 void __init proc_caches_init(void)
1822 {
1823         sighand_cachep = kmem_cache_create("sighand_cache",
1824                         sizeof(struct sighand_struct), 0,
1825                         SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
1826                         SLAB_NOTRACK, sighand_ctor);
1827         signal_cachep = kmem_cache_create("signal_cache",
1828                         sizeof(struct signal_struct), 0,
1829                         SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1830         files_cachep = kmem_cache_create("files_cache",
1831                         sizeof(struct files_struct), 0,
1832                         SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1833         fs_cachep = kmem_cache_create("fs_cache",
1834                         sizeof(struct fs_struct), 0,
1835                         SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1836         /*
1837          * FIXME! The "sizeof(struct mm_struct)" currently includes the
1838          * whole struct cpumask for the OFFSTACK case. We could change
1839          * this to *only* allocate as much of it as required by the
1840          * maximum number of CPU's we can ever have.  The cpumask_allocation
1841          * is at the end of the structure, exactly for that reason.
1842          */
1843         mm_cachep = kmem_cache_create("mm_struct",
1844                         sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
1845                         SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1846         vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
1847         mmap_init();
1848         nsproxy_cache_init();
1849 }
1850 
1851 /*
1852  * Check constraints on flags passed to the unshare system call.
1853  */
1854 static int check_unshare_flags(unsigned long unshare_flags)
1855 {
1856         if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
1857                                 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
1858                                 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET|
1859                                 CLONE_NEWUSER|CLONE_NEWPID))
1860                 return -EINVAL;
1861         /*
1862          * Not implemented, but pretend it works if there is nothing
1863          * to unshare.  Note that unsharing the address space or the
1864          * signal handlers also need to unshare the signal queues (aka
1865          * CLONE_THREAD).
1866          */
1867         if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
1868                 if (!thread_group_empty(current))
1869                         return -EINVAL;
1870         }
1871         if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) {
1872                 if (atomic_read(&current->sighand->count) > 1)
1873                         return -EINVAL;
1874         }
1875         if (unshare_flags & CLONE_VM) {
1876                 if (!current_is_single_threaded())
1877                         return -EINVAL;
1878         }
1879 
1880         return 0;
1881 }
1882 
1883 /*
1884  * Unshare the filesystem structure if it is being shared
1885  */
1886 static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1887 {
1888         struct fs_struct *fs = current->fs;
1889 
1890         if (!(unshare_flags & CLONE_FS) || !fs)
1891                 return 0;
1892 
1893         /* don't need lock here; in the worst case we'll do useless copy */
1894         if (fs->users == 1)
1895                 return 0;
1896 
1897         *new_fsp = copy_fs_struct(fs);
1898         if (!*new_fsp)
1899                 return -ENOMEM;
1900 
1901         return 0;
1902 }
1903 
1904 /*
1905  * Unshare file descriptor table if it is being shared
1906  */
1907 static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1908 {
1909         struct files_struct *fd = current->files;
1910         int error = 0;
1911 
1912         if ((unshare_flags & CLONE_FILES) &&
1913             (fd && atomic_read(&fd->count) > 1)) {
1914                 *new_fdp = dup_fd(fd, &error);
1915                 if (!*new_fdp)
1916                         return error;
1917         }
1918 
1919         return 0;
1920 }
1921 
1922 /*
1923  * unshare allows a process to 'unshare' part of the process
1924  * context which was originally shared using clone.  copy_*
1925  * functions used by do_fork() cannot be used here directly
1926  * because they modify an inactive task_struct that is being
1927  * constructed. Here we are modifying the current, active,
1928  * task_struct.
1929  */
1930 SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
1931 {
1932         struct fs_struct *fs, *new_fs = NULL;
1933         struct files_struct *fd, *new_fd = NULL;
1934         struct cred *new_cred = NULL;
1935         struct nsproxy *new_nsproxy = NULL;
1936         int do_sysvsem = 0;
1937         int err;
1938 
1939         /*
1940          * If unsharing a user namespace must also unshare the thread.
1941          */
1942         if (unshare_flags & CLONE_NEWUSER)
1943                 unshare_flags |= CLONE_THREAD | CLONE_FS;
1944         /*
1945          * If unsharing vm, must also unshare signal handlers.
1946          */
1947         if (unshare_flags & CLONE_VM)
1948                 unshare_flags |= CLONE_SIGHAND;
1949         /*
1950          * If unsharing a signal handlers, must also unshare the signal queues.
1951          */
1952         if (unshare_flags & CLONE_SIGHAND)
1953                 unshare_flags |= CLONE_THREAD;
1954         /*
1955          * If unsharing namespace, must also unshare filesystem information.
1956          */
1957         if (unshare_flags & CLONE_NEWNS)
1958                 unshare_flags |= CLONE_FS;
1959 
1960         err = check_unshare_flags(unshare_flags);
1961         if (err)
1962                 goto bad_unshare_out;
1963         /*
1964          * CLONE_NEWIPC must also detach from the undolist: after switching
1965          * to a new ipc namespace, the semaphore arrays from the old
1966          * namespace are unreachable.
1967          */
1968         if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
1969                 do_sysvsem = 1;
1970         err = unshare_fs(unshare_flags, &new_fs);
1971         if (err)
1972                 goto bad_unshare_out;
1973         err = unshare_fd(unshare_flags, &new_fd);
1974         if (err)
1975                 goto bad_unshare_cleanup_fs;
1976         err = unshare_userns(unshare_flags, &new_cred);
1977         if (err)
1978                 goto bad_unshare_cleanup_fd;
1979         err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
1980                                          new_cred, new_fs);
1981         if (err)
1982                 goto bad_unshare_cleanup_cred;
1983 
1984         if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) {
1985                 if (do_sysvsem) {
1986                         /*
1987                          * CLONE_SYSVSEM is equivalent to sys_exit().
1988                          */
1989                         exit_sem(current);
1990                 }
1991                 if (unshare_flags & CLONE_NEWIPC) {
1992                         /* Orphan segments in old ns (see sem above). */
1993                         exit_shm(current);
1994                         shm_init_task(current);
1995                 }
1996 
1997                 if (new_nsproxy)
1998                         switch_task_namespaces(current, new_nsproxy);
1999 
2000                 task_lock(current);
2001 
2002                 if (new_fs) {
2003                         fs = current->fs;
2004                         spin_lock(&fs->lock);
2005                         current->fs = new_fs;
2006                         if (--fs->users)
2007                                 new_fs = NULL;
2008                         else
2009                                 new_fs = fs;
2010                         spin_unlock(&fs->lock);
2011                 }
2012 
2013                 if (new_fd) {
2014                         fd = current->files;
2015                         current->files = new_fd;
2016                         new_fd = fd;
2017                 }
2018 
2019                 task_unlock(current);
2020 
2021                 if (new_cred) {
2022                         /* Install the new user namespace */
2023                         commit_creds(new_cred);
2024                         new_cred = NULL;
2025                 }
2026         }
2027 
2028 bad_unshare_cleanup_cred:
2029         if (new_cred)
2030                 put_cred(new_cred);
2031 bad_unshare_cleanup_fd:
2032         if (new_fd)
2033                 put_files_struct(new_fd);
2034 
2035 bad_unshare_cleanup_fs:
2036         if (new_fs)
2037                 free_fs_struct(new_fs);
2038 
2039 bad_unshare_out:
2040         return err;
2041 }
2042 
2043 /*
2044  *      Helper to unshare the files of the current task.
2045  *      We don't want to expose copy_files internals to
2046  *      the exec layer of the kernel.
2047  */
2048 
2049 int unshare_files(struct files_struct **displaced)
2050 {
2051         struct task_struct *task = current;
2052         struct files_struct *copy = NULL;
2053         int error;
2054 
2055         error = unshare_fd(CLONE_FILES, &copy);
2056         if (error || !copy) {
2057                 *displaced = NULL;
2058                 return error;
2059         }
2060         *displaced = task->files;
2061         task_lock(task);
2062         task->files = copy;
2063         task_unlock(task);
2064         return 0;
2065 }
2066 
2067 int sysctl_max_threads(struct ctl_table *table, int write,
2068                        void __user *buffer, size_t *lenp, loff_t *ppos)
2069 {
2070         struct ctl_table t;
2071         int ret;
2072         int threads = max_threads;
2073         int min = MIN_THREADS;
2074         int max = MAX_THREADS;
2075 
2076         t = *table;
2077         t.data = &threads;
2078         t.extra1 = &min;
2079         t.extra2 = &max;
2080 
2081         ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos);
2082         if (ret || !write)
2083                 return ret;
2084 
2085         set_max_threads(threads);
2086 
2087         return 0;
2088 }
2089 

~ [ 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