~ [ 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/nsproxy.h>
 32 #include <linux/capability.h>
 33 #include <linux/cpu.h>
 34 #include <linux/cgroup.h>
 35 #include <linux/security.h>
 36 #include <linux/hugetlb.h>
 37 #include <linux/swap.h>
 38 #include <linux/syscalls.h>
 39 #include <linux/jiffies.h>
 40 #include <linux/tracehook.h>
 41 #include <linux/futex.h>
 42 #include <linux/compat.h>
 43 #include <linux/task_io_accounting_ops.h>
 44 #include <linux/rcupdate.h>
 45 #include <linux/ptrace.h>
 46 #include <linux/mount.h>
 47 #include <linux/audit.h>
 48 #include <linux/memcontrol.h>
 49 #include <linux/ftrace.h>
 50 #include <linux/profile.h>
 51 #include <linux/rmap.h>
 52 #include <linux/ksm.h>
 53 #include <linux/acct.h>
 54 #include <linux/tsacct_kern.h>
 55 #include <linux/cn_proc.h>
 56 #include <linux/freezer.h>
 57 #include <linux/delayacct.h>
 58 #include <linux/taskstats_kern.h>
 59 #include <linux/random.h>
 60 #include <linux/tty.h>
 61 #include <linux/proc_fs.h>
 62 #include <linux/blkdev.h>
 63 #include <linux/fs_struct.h>
 64 #include <linux/magic.h>
 65 #include <linux/perf_event.h>
 66 #include <linux/posix-timers.h>
 67 #include <linux/signalfd.h>
 68 
 69 #include <asm/pgtable.h>
 70 #include <asm/pgalloc.h>
 71 #include <asm/uaccess.h>
 72 #include <asm/mmu_context.h>
 73 #include <asm/cacheflush.h>
 74 #include <asm/tlbflush.h>
 75 
 76 #include <trace/events/sched.h>
 77 
 78 /*
 79  * Protected counters by write_lock_irq(&tasklist_lock)
 80  */
 81 unsigned long total_forks;      /* Handle normal Linux uptimes. */
 82 int nr_threads;                 /* The idle threads do not count.. */
 83 
 84 int max_threads;                /* tunable limit on nr_threads */
 85 
 86 DEFINE_PER_CPU(unsigned long, process_counts) = 0;
 87 
 88 __cacheline_aligned DEFINE_RWLOCK(tasklist_lock);  /* outer */
 89 
 90 int nr_processes(void)
 91 {
 92         int cpu;
 93         int total = 0;
 94 
 95         for_each_possible_cpu(cpu)
 96                 total += per_cpu(process_counts, cpu);
 97 
 98         return total;
 99 }
100 
101 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
102 # define alloc_task_struct()    kmem_cache_alloc(task_struct_cachep, GFP_KERNEL)
103 # define free_task_struct(tsk)  kmem_cache_free(task_struct_cachep, (tsk))
104 static struct kmem_cache *task_struct_cachep;
105 #endif
106 
107 #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
108 static inline struct thread_info *alloc_thread_info(struct task_struct *tsk)
109 {
110 #ifdef CONFIG_DEBUG_STACK_USAGE
111         gfp_t mask = GFP_KERNEL | __GFP_ZERO;
112 #else
113         gfp_t mask = GFP_KERNEL;
114 #endif
115         return (struct thread_info *)__get_free_pages(mask, THREAD_SIZE_ORDER);
116 }
117 
118 static inline void free_thread_info(struct thread_info *ti)
119 {
120         free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
121 }
122 #endif
123 
124 /* SLAB cache for signal_struct structures (tsk->signal) */
125 static struct kmem_cache *signal_cachep;
126 
127 /* SLAB cache for sighand_struct structures (tsk->sighand) */
128 struct kmem_cache *sighand_cachep;
129 
130 /* SLAB cache for files_struct structures (tsk->files) */
131 struct kmem_cache *files_cachep;
132 
133 /* SLAB cache for fs_struct structures (tsk->fs) */
134 struct kmem_cache *fs_cachep;
135 
136 /* SLAB cache for vm_area_struct structures */
137 struct kmem_cache *vm_area_cachep;
138 
139 /* SLAB cache for mm_struct structures (tsk->mm) */
140 static struct kmem_cache *mm_cachep;
141 
142 static void account_kernel_stack(struct thread_info *ti, int account)
143 {
144         struct zone *zone = page_zone(virt_to_page(ti));
145 
146         mod_zone_page_state(zone, NR_KERNEL_STACK, account);
147 }
148 
149 void free_task(struct task_struct *tsk)
150 {
151         prop_local_destroy_single(&tsk->dirties);
152         account_kernel_stack(tsk->stack, -1);
153         free_thread_info(tsk->stack);
154         rt_mutex_debug_task_free(tsk);
155         ftrace_graph_exit_task(tsk);
156         free_task_struct(tsk);
157 }
158 EXPORT_SYMBOL(free_task);
159 
160 void __put_task_struct(struct task_struct *tsk)
161 {
162         WARN_ON(!tsk->exit_state);
163         WARN_ON(atomic_read(&tsk->usage));
164         WARN_ON(tsk == current);
165 
166         exit_creds(tsk);
167         delayacct_tsk_free(tsk);
168 
169         ccs_free_task_security(tsk);
170         if (!profile_handoff_task(tsk))
171                 free_task(tsk);
172 }
173 
174 /*
175  * macro override instead of weak attribute alias, to workaround
176  * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
177  */
178 #ifndef arch_task_cache_init
179 #define arch_task_cache_init()
180 #endif
181 
182 void __init fork_init(unsigned long mempages)
183 {
184 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
185 #ifndef ARCH_MIN_TASKALIGN
186 #define ARCH_MIN_TASKALIGN      L1_CACHE_BYTES
187 #endif
188         /* create a slab on which task_structs can be allocated */
189         task_struct_cachep =
190                 kmem_cache_create("task_struct", sizeof(struct task_struct),
191                         ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL);
192 #endif
193 
194         /* do the arch specific task caches init */
195         arch_task_cache_init();
196 
197         /*
198          * The default maximum number of threads is set to a safe
199          * value: the thread structures can take up at most half
200          * of memory.
201          */
202         max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
203 
204         /*
205          * we need to allow at least 20 threads to boot a system
206          */
207         if(max_threads < 20)
208                 max_threads = 20;
209 
210         init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
211         init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
212         init_task.signal->rlim[RLIMIT_SIGPENDING] =
213                 init_task.signal->rlim[RLIMIT_NPROC];
214 }
215 
216 int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst,
217                                                struct task_struct *src)
218 {
219         *dst = *src;
220         return 0;
221 }
222 
223 static struct task_struct *dup_task_struct(struct task_struct *orig)
224 {
225         struct task_struct *tsk;
226         struct thread_info *ti;
227         unsigned long *stackend;
228 
229         int err;
230 
231         prepare_to_copy(orig);
232 
233         tsk = alloc_task_struct();
234         if (!tsk)
235                 return NULL;
236 
237         ti = alloc_thread_info(tsk);
238         if (!ti) {
239                 free_task_struct(tsk);
240                 return NULL;
241         }
242 
243         err = arch_dup_task_struct(tsk, orig);
244         if (err)
245                 goto out;
246 
247         tsk->stack = ti;
248 
249         err = prop_local_init_single(&tsk->dirties);
250         if (err)
251                 goto out;
252 
253         setup_thread_stack(tsk, orig);
254         stackend = end_of_stack(tsk);
255         *stackend = STACK_END_MAGIC;    /* for overflow detection */
256 
257 #ifdef CONFIG_CC_STACKPROTECTOR
258         tsk->stack_canary = get_random_int();
259 #endif
260 
261         /* One for us, one for whoever does the "release_task()" (usually parent) */
262         atomic_set(&tsk->usage,2);
263         atomic_set(&tsk->fs_excl, 0);
264 #ifdef CONFIG_BLK_DEV_IO_TRACE
265         tsk->btrace_seq = 0;
266 #endif
267         tsk->splice_pipe = NULL;
268 
269         account_kernel_stack(ti, 1);
270 
271         return tsk;
272 
273 out:
274         free_thread_info(ti);
275         free_task_struct(tsk);
276         return NULL;
277 }
278 
279 #ifdef CONFIG_MMU
280 static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
281 {
282         struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
283         struct rb_node **rb_link, *rb_parent;
284         int retval;
285         unsigned long charge;
286         struct mempolicy *pol;
287 
288         down_write(&oldmm->mmap_sem);
289         flush_cache_dup_mm(oldmm);
290         /*
291          * Not linked in yet - no deadlock potential:
292          */
293         down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
294 
295         mm->locked_vm = 0;
296         mm->mmap = NULL;
297         mm->mmap_cache = NULL;
298         mm->free_area_cache = oldmm->mmap_base;
299         mm->cached_hole_size = ~0UL;
300         mm->map_count = 0;
301         cpumask_clear(mm_cpumask(mm));
302         mm->mm_rb = RB_ROOT;
303         rb_link = &mm->mm_rb.rb_node;
304         rb_parent = NULL;
305         pprev = &mm->mmap;
306         retval = ksm_fork(mm, oldmm);
307         if (retval)
308                 goto out;
309 
310         prev = NULL;
311         for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
312                 struct file *file;
313 
314                 if (mpnt->vm_flags & VM_DONTCOPY) {
315                         long pages = vma_pages(mpnt);
316                         mm->total_vm -= pages;
317                         vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
318                                                                 -pages);
319                         continue;
320                 }
321                 charge = 0;
322                 if (mpnt->vm_flags & VM_ACCOUNT) {
323                         unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
324                         if (security_vm_enough_memory(len))
325                                 goto fail_nomem;
326                         charge = len;
327                 }
328                 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
329                 if (!tmp)
330                         goto fail_nomem;
331                 *tmp = *mpnt;
332                 pol = mpol_dup(vma_policy(mpnt));
333                 retval = PTR_ERR(pol);
334                 if (IS_ERR(pol))
335                         goto fail_nomem_policy;
336                 vma_set_policy(tmp, pol);
337                 tmp->vm_flags &= ~VM_LOCKED;
338                 tmp->vm_mm = mm;
339                 tmp->vm_next = tmp->vm_prev = NULL;
340                 anon_vma_link(tmp);
341                 file = tmp->vm_file;
342                 if (file) {
343                         struct inode *inode = file->f_path.dentry->d_inode;
344                         struct address_space *mapping = file->f_mapping;
345 
346                         get_file(file);
347                         if (tmp->vm_flags & VM_DENYWRITE)
348                                 atomic_dec(&inode->i_writecount);
349                         spin_lock(&mapping->i_mmap_lock);
350                         if (tmp->vm_flags & VM_SHARED)
351                                 mapping->i_mmap_writable++;
352                         tmp->vm_truncate_count = mpnt->vm_truncate_count;
353                         flush_dcache_mmap_lock(mapping);
354                         /* insert tmp into the share list, just after mpnt */
355                         vma_prio_tree_add(tmp, mpnt);
356                         flush_dcache_mmap_unlock(mapping);
357                         spin_unlock(&mapping->i_mmap_lock);
358                 }
359 
360                 /*
361                  * Clear hugetlb-related page reserves for children. This only
362                  * affects MAP_PRIVATE mappings. Faults generated by the child
363                  * are not guaranteed to succeed, even if read-only
364                  */
365                 if (is_vm_hugetlb_page(tmp))
366                         reset_vma_resv_huge_pages(tmp);
367 
368                 /*
369                  * Link in the new vma and copy the page table entries.
370                  */
371                 *pprev = tmp;
372                 pprev = &tmp->vm_next;
373                 tmp->vm_prev = prev;
374                 prev = tmp;
375 
376                 __vma_link_rb(mm, tmp, rb_link, rb_parent);
377                 rb_link = &tmp->vm_rb.rb_right;
378                 rb_parent = &tmp->vm_rb;
379 
380                 mm->map_count++;
381                 retval = copy_page_range(mm, oldmm, mpnt);
382 
383                 if (tmp->vm_ops && tmp->vm_ops->open)
384                         tmp->vm_ops->open(tmp);
385 
386                 if (retval)
387                         goto out;
388         }
389         /* a new mm has just been created */
390         arch_dup_mmap(oldmm, mm);
391         retval = 0;
392 out:
393         up_write(&mm->mmap_sem);
394         flush_tlb_mm(oldmm);
395         up_write(&oldmm->mmap_sem);
396         return retval;
397 fail_nomem_policy:
398         kmem_cache_free(vm_area_cachep, tmp);
399 fail_nomem:
400         retval = -ENOMEM;
401         vm_unacct_memory(charge);
402         goto out;
403 }
404 
405 static inline int mm_alloc_pgd(struct mm_struct * mm)
406 {
407         mm->pgd = pgd_alloc(mm);
408         if (unlikely(!mm->pgd))
409                 return -ENOMEM;
410         return 0;
411 }
412 
413 static inline void mm_free_pgd(struct mm_struct * mm)
414 {
415         pgd_free(mm, mm->pgd);
416 }
417 #else
418 #define dup_mmap(mm, oldmm)     (0)
419 #define mm_alloc_pgd(mm)        (0)
420 #define mm_free_pgd(mm)
421 #endif /* CONFIG_MMU */
422 
423 __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
424 
425 #define allocate_mm()   (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
426 #define free_mm(mm)     (kmem_cache_free(mm_cachep, (mm)))
427 
428 static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
429 
430 static int __init coredump_filter_setup(char *s)
431 {
432         default_dump_filter =
433                 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
434                 MMF_DUMP_FILTER_MASK;
435         return 1;
436 }
437 
438 __setup("coredump_filter=", coredump_filter_setup);
439 
440 #include <linux/init_task.h>
441 
442 static void mm_init_aio(struct mm_struct *mm)
443 {
444 #ifdef CONFIG_AIO
445         spin_lock_init(&mm->ioctx_lock);
446         INIT_HLIST_HEAD(&mm->ioctx_list);
447 #endif
448 }
449 
450 static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p)
451 {
452         atomic_set(&mm->mm_users, 1);
453         atomic_set(&mm->mm_count, 1);
454         init_rwsem(&mm->mmap_sem);
455         INIT_LIST_HEAD(&mm->mmlist);
456         mm->flags = (current->mm) ?
457                 (current->mm->flags & MMF_INIT_MASK) : default_dump_filter;
458         mm->core_state = NULL;
459         mm->nr_ptes = 0;
460         set_mm_counter(mm, file_rss, 0);
461         set_mm_counter(mm, anon_rss, 0);
462         spin_lock_init(&mm->page_table_lock);
463         mm->free_area_cache = TASK_UNMAPPED_BASE;
464         mm->cached_hole_size = ~0UL;
465         mm_init_aio(mm);
466         mm_init_owner(mm, p);
467 
468         if (likely(!mm_alloc_pgd(mm))) {
469                 mm->def_flags = 0;
470                 mmu_notifier_mm_init(mm);
471                 return mm;
472         }
473 
474         free_mm(mm);
475         return NULL;
476 }
477 
478 /*
479  * Allocate and initialize an mm_struct.
480  */
481 struct mm_struct * mm_alloc(void)
482 {
483         struct mm_struct * mm;
484 
485         mm = allocate_mm();
486         if (mm) {
487                 memset(mm, 0, sizeof(*mm));
488                 mm = mm_init(mm, current);
489         }
490         return mm;
491 }
492 
493 /*
494  * Called when the last reference to the mm
495  * is dropped: either by a lazy thread or by
496  * mmput. Free the page directory and the mm.
497  */
498 void __mmdrop(struct mm_struct *mm)
499 {
500         BUG_ON(mm == &init_mm);
501         mm_free_pgd(mm);
502         destroy_context(mm);
503         mmu_notifier_mm_destroy(mm);
504         free_mm(mm);
505 }
506 EXPORT_SYMBOL_GPL(__mmdrop);
507 
508 /*
509  * Decrement the use count and release all resources for an mm.
510  */
511 void mmput(struct mm_struct *mm)
512 {
513         might_sleep();
514 
515         if (atomic_dec_and_test(&mm->mm_users)) {
516                 exit_aio(mm);
517                 ksm_exit(mm);
518                 exit_mmap(mm);
519                 set_mm_exe_file(mm, NULL);
520                 if (!list_empty(&mm->mmlist)) {
521                         spin_lock(&mmlist_lock);
522                         list_del(&mm->mmlist);
523                         spin_unlock(&mmlist_lock);
524                 }
525                 put_swap_token(mm);
526                 if (mm->binfmt)
527                         module_put(mm->binfmt->module);
528                 mmdrop(mm);
529         }
530 }
531 EXPORT_SYMBOL_GPL(mmput);
532 
533 /**
534  * get_task_mm - acquire a reference to the task's mm
535  *
536  * Returns %NULL if the task has no mm.  Checks PF_KTHREAD (meaning
537  * this kernel workthread has transiently adopted a user mm with use_mm,
538  * to do its AIO) is not set and if so returns a reference to it, after
539  * bumping up the use count.  User must release the mm via mmput()
540  * after use.  Typically used by /proc and ptrace.
541  */
542 struct mm_struct *get_task_mm(struct task_struct *task)
543 {
544         struct mm_struct *mm;
545 
546         task_lock(task);
547         mm = task->mm;
548         if (mm) {
549                 if (task->flags & PF_KTHREAD)
550                         mm = NULL;
551                 else
552                         atomic_inc(&mm->mm_users);
553         }
554         task_unlock(task);
555         return mm;
556 }
557 EXPORT_SYMBOL_GPL(get_task_mm);
558 
559 /* Please note the differences between mmput and mm_release.
560  * mmput is called whenever we stop holding onto a mm_struct,
561  * error success whatever.
562  *
563  * mm_release is called after a mm_struct has been removed
564  * from the current process.
565  *
566  * This difference is important for error handling, when we
567  * only half set up a mm_struct for a new process and need to restore
568  * the old one.  Because we mmput the new mm_struct before
569  * restoring the old one. . .
570  * Eric Biederman 10 January 1998
571  */
572 void mm_release(struct task_struct *tsk, struct mm_struct *mm)
573 {
574         struct completion *vfork_done = tsk->vfork_done;
575 
576         /* Get rid of any futexes when releasing the mm */
577 #ifdef CONFIG_FUTEX
578         if (unlikely(tsk->robust_list)) {
579                 exit_robust_list(tsk);
580                 tsk->robust_list = NULL;
581         }
582 #ifdef CONFIG_COMPAT
583         if (unlikely(tsk->compat_robust_list)) {
584                 compat_exit_robust_list(tsk);
585                 tsk->compat_robust_list = NULL;
586         }
587 #endif
588         if (unlikely(!list_empty(&tsk->pi_state_list)))
589                 exit_pi_state_list(tsk);
590 #endif
591 
592         /* Get rid of any cached register state */
593         deactivate_mm(tsk, mm);
594 
595         /* notify parent sleeping on vfork() */
596         if (vfork_done) {
597                 tsk->vfork_done = NULL;
598                 complete(vfork_done);
599         }
600 
601         /*
602          * If we're exiting normally, clear a user-space tid field if
603          * requested.  We leave this alone when dying by signal, to leave
604          * the value intact in a core dump, and to save the unnecessary
605          * trouble otherwise.  Userland only wants this done for a sys_exit.
606          */
607         if (tsk->clear_child_tid) {
608                 if (!(tsk->flags & PF_SIGNALED) &&
609                     atomic_read(&mm->mm_users) > 1) {
610                         /*
611                          * We don't check the error code - if userspace has
612                          * not set up a proper pointer then tough luck.
613                          */
614                         put_user(0, tsk->clear_child_tid);
615                         sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
616                                         1, NULL, NULL, 0);
617                 }
618                 tsk->clear_child_tid = NULL;
619         }
620 }
621 
622 /*
623  * Allocate a new mm structure and copy contents from the
624  * mm structure of the passed in task structure.
625  */
626 struct mm_struct *dup_mm(struct task_struct *tsk)
627 {
628         struct mm_struct *mm, *oldmm = current->mm;
629         int err;
630 
631         if (!oldmm)
632                 return NULL;
633 
634         mm = allocate_mm();
635         if (!mm)
636                 goto fail_nomem;
637 
638         memcpy(mm, oldmm, sizeof(*mm));
639 
640         /* Initializing for Swap token stuff */
641         mm->token_priority = 0;
642         mm->last_interval = 0;
643 
644         if (!mm_init(mm, tsk))
645                 goto fail_nomem;
646 
647         if (init_new_context(tsk, mm))
648                 goto fail_nocontext;
649 
650         dup_mm_exe_file(oldmm, mm);
651 
652         err = dup_mmap(mm, oldmm);
653         if (err)
654                 goto free_pt;
655 
656         mm->hiwater_rss = get_mm_rss(mm);
657         mm->hiwater_vm = mm->total_vm;
658 
659         if (mm->binfmt && !try_module_get(mm->binfmt->module))
660                 goto free_pt;
661 
662         return mm;
663 
664 free_pt:
665         /* don't put binfmt in mmput, we haven't got module yet */
666         mm->binfmt = NULL;
667         mmput(mm);
668 
669 fail_nomem:
670         return NULL;
671 
672 fail_nocontext:
673         /*
674          * If init_new_context() failed, we cannot use mmput() to free the mm
675          * because it calls destroy_context()
676          */
677         mm_free_pgd(mm);
678         free_mm(mm);
679         return NULL;
680 }
681 
682 static int copy_mm(unsigned long clone_flags, struct task_struct * tsk)
683 {
684         struct mm_struct * mm, *oldmm;
685         int retval;
686 
687         tsk->min_flt = tsk->maj_flt = 0;
688         tsk->nvcsw = tsk->nivcsw = 0;
689 #ifdef CONFIG_DETECT_HUNG_TASK
690         tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
691 #endif
692 
693         tsk->mm = NULL;
694         tsk->active_mm = NULL;
695 
696         /*
697          * Are we cloning a kernel thread?
698          *
699          * We need to steal a active VM for that..
700          */
701         oldmm = current->mm;
702         if (!oldmm)
703                 return 0;
704 
705         if (clone_flags & CLONE_VM) {
706                 atomic_inc(&oldmm->mm_users);
707                 mm = oldmm;
708                 goto good_mm;
709         }
710 
711         retval = -ENOMEM;
712         mm = dup_mm(tsk);
713         if (!mm)
714                 goto fail_nomem;
715 
716 good_mm:
717         /* Initializing for Swap token stuff */
718         mm->token_priority = 0;
719         mm->last_interval = 0;
720 
721         tsk->mm = mm;
722         tsk->active_mm = mm;
723         return 0;
724 
725 fail_nomem:
726         return retval;
727 }
728 
729 static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
730 {
731         struct fs_struct *fs = current->fs;
732         if (clone_flags & CLONE_FS) {
733                 /* tsk->fs is already what we want */
734                 write_lock(&fs->lock);
735                 if (fs->in_exec) {
736                         write_unlock(&fs->lock);
737                         return -EAGAIN;
738                 }
739                 fs->users++;
740                 write_unlock(&fs->lock);
741                 return 0;
742         }
743         tsk->fs = copy_fs_struct(fs);
744         if (!tsk->fs)
745                 return -ENOMEM;
746         return 0;
747 }
748 
749 static int copy_files(unsigned long clone_flags, struct task_struct * tsk)
750 {
751         struct files_struct *oldf, *newf;
752         int error = 0;
753 
754         /*
755          * A background process may not have any files ...
756          */
757         oldf = current->files;
758         if (!oldf)
759                 goto out;
760 
761         if (clone_flags & CLONE_FILES) {
762                 atomic_inc(&oldf->count);
763                 goto out;
764         }
765 
766         newf = dup_fd(oldf, &error);
767         if (!newf)
768                 goto out;
769 
770         tsk->files = newf;
771         error = 0;
772 out:
773         return error;
774 }
775 
776 static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
777 {
778 #ifdef CONFIG_BLOCK
779         struct io_context *ioc = current->io_context;
780 
781         if (!ioc)
782                 return 0;
783         /*
784          * Share io context with parent, if CLONE_IO is set
785          */
786         if (clone_flags & CLONE_IO) {
787                 tsk->io_context = ioc_task_link(ioc);
788                 if (unlikely(!tsk->io_context))
789                         return -ENOMEM;
790         } else if (ioprio_valid(ioc->ioprio)) {
791                 tsk->io_context = alloc_io_context(GFP_KERNEL, -1);
792                 if (unlikely(!tsk->io_context))
793                         return -ENOMEM;
794 
795                 tsk->io_context->ioprio = ioc->ioprio;
796         }
797 #endif
798         return 0;
799 }
800 
801 static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
802 {
803         struct sighand_struct *sig;
804 
805         if (clone_flags & CLONE_SIGHAND) {
806                 atomic_inc(&current->sighand->count);
807                 return 0;
808         }
809         sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
810         rcu_assign_pointer(tsk->sighand, sig);
811         if (!sig)
812                 return -ENOMEM;
813         atomic_set(&sig->count, 1);
814         memcpy(sig->action, current->sighand->action, sizeof(sig->action));
815         return 0;
816 }
817 
818 void __cleanup_sighand(struct sighand_struct *sighand)
819 {
820         if (atomic_dec_and_test(&sighand->count)) {
821                 signalfd_cleanup(sighand);
822                 kmem_cache_free(sighand_cachep, sighand);
823         }
824 }
825 
826 
827 /*
828  * Initialize POSIX timer handling for a thread group.
829  */
830 static void posix_cpu_timers_init_group(struct signal_struct *sig)
831 {
832         /* Thread group counters. */
833         thread_group_cputime_init(sig);
834 
835         /* Expiration times and increments. */
836         sig->it[CPUCLOCK_PROF].expires = cputime_zero;
837         sig->it[CPUCLOCK_PROF].incr = cputime_zero;
838         sig->it[CPUCLOCK_VIRT].expires = cputime_zero;
839         sig->it[CPUCLOCK_VIRT].incr = cputime_zero;
840 
841         /* Cached expiration times. */
842         sig->cputime_expires.prof_exp = cputime_zero;
843         sig->cputime_expires.virt_exp = cputime_zero;
844         sig->cputime_expires.sched_exp = 0;
845 
846         if (sig->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) {
847                 sig->cputime_expires.prof_exp =
848                         secs_to_cputime(sig->rlim[RLIMIT_CPU].rlim_cur);
849                 sig->cputimer.running = 1;
850         }
851 
852         /* The timer lists. */
853         INIT_LIST_HEAD(&sig->cpu_timers[0]);
854         INIT_LIST_HEAD(&sig->cpu_timers[1]);
855         INIT_LIST_HEAD(&sig->cpu_timers[2]);
856 }
857 
858 static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
859 {
860         struct signal_struct *sig;
861 
862         if (clone_flags & CLONE_THREAD)
863                 return 0;
864 
865         sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL);
866         tsk->signal = sig;
867         if (!sig)
868                 return -ENOMEM;
869 
870         atomic_set(&sig->count, 1);
871         atomic_set(&sig->live, 1);
872         init_waitqueue_head(&sig->wait_chldexit);
873         sig->flags = 0;
874         if (clone_flags & CLONE_NEWPID)
875                 sig->flags |= SIGNAL_UNKILLABLE;
876         sig->group_exit_code = 0;
877         sig->group_exit_task = NULL;
878         sig->group_stop_count = 0;
879         sig->curr_target = tsk;
880         init_sigpending(&sig->shared_pending);
881         INIT_LIST_HEAD(&sig->posix_timers);
882 
883         hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
884         sig->it_real_incr.tv64 = 0;
885         sig->real_timer.function = it_real_fn;
886 
887         sig->leader = 0;        /* session leadership doesn't inherit */
888         sig->tty_old_pgrp = NULL;
889         sig->tty = NULL;
890 
891         sig->utime = sig->stime = sig->cutime = sig->cstime = cputime_zero;
892         sig->gtime = cputime_zero;
893         sig->cgtime = cputime_zero;
894 #ifndef CONFIG_VIRT_CPU_ACCOUNTING
895         sig->prev_utime = sig->prev_stime = cputime_zero;
896 #endif
897         sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0;
898         sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0;
899         sig->inblock = sig->oublock = sig->cinblock = sig->coublock = 0;
900         sig->maxrss = sig->cmaxrss = 0;
901         task_io_accounting_init(&sig->ioac);
902         sig->sum_sched_runtime = 0;
903         taskstats_tgid_init(sig);
904 
905         task_lock(current->group_leader);
906         memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
907         task_unlock(current->group_leader);
908 
909         posix_cpu_timers_init_group(sig);
910 
911         acct_init_pacct(&sig->pacct);
912 
913         tty_audit_fork(sig);
914 
915         sig->oom_adj = current->signal->oom_adj;
916 
917         return 0;
918 }
919 
920 void __cleanup_signal(struct signal_struct *sig)
921 {
922         thread_group_cputime_free(sig);
923         tty_kref_put(sig->tty);
924         kmem_cache_free(signal_cachep, sig);
925 }
926 
927 static void copy_flags(unsigned long clone_flags, struct task_struct *p)
928 {
929         unsigned long new_flags = p->flags;
930 
931         new_flags &= ~PF_SUPERPRIV;
932         new_flags |= PF_FORKNOEXEC;
933         new_flags |= PF_STARTING;
934         p->flags = new_flags;
935         clear_freeze_flag(p);
936 }
937 
938 SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
939 {
940         current->clear_child_tid = tidptr;
941 
942         return task_pid_vnr(current);
943 }
944 
945 static void rt_mutex_init_task(struct task_struct *p)
946 {
947         spin_lock_init(&p->pi_lock);
948 #ifdef CONFIG_RT_MUTEXES
949         plist_head_init(&p->pi_waiters, &p->pi_lock);
950         p->pi_blocked_on = NULL;
951 #endif
952 }
953 
954 #ifdef CONFIG_MM_OWNER
955 void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
956 {
957         mm->owner = p;
958 }
959 #endif /* CONFIG_MM_OWNER */
960 
961 /*
962  * Initialize POSIX timer handling for a single task.
963  */
964 static void posix_cpu_timers_init(struct task_struct *tsk)
965 {
966         tsk->cputime_expires.prof_exp = cputime_zero;
967         tsk->cputime_expires.virt_exp = cputime_zero;
968         tsk->cputime_expires.sched_exp = 0;
969         INIT_LIST_HEAD(&tsk->cpu_timers[0]);
970         INIT_LIST_HEAD(&tsk->cpu_timers[1]);
971         INIT_LIST_HEAD(&tsk->cpu_timers[2]);
972 }
973 
974 /*
975  * This creates a new process as a copy of the old one,
976  * but does not actually start it yet.
977  *
978  * It copies the registers, and all the appropriate
979  * parts of the process environment (as per the clone
980  * flags). The actual kick-off is left to the caller.
981  */
982 static struct task_struct *copy_process(unsigned long clone_flags,
983                                         unsigned long stack_start,
984                                         struct pt_regs *regs,
985                                         unsigned long stack_size,
986                                         int __user *child_tidptr,
987                                         struct pid *pid,
988                                         int trace)
989 {
990         int retval;
991         struct task_struct *p;
992         int cgroup_callbacks_done = 0;
993 
994         if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
995                 return ERR_PTR(-EINVAL);
996 
997         /*
998          * Thread groups must share signals as well, and detached threads
999          * can only be started up within the thread group.
1000          */
1001         if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1002                 return ERR_PTR(-EINVAL);
1003 
1004         /*
1005          * Shared signal handlers imply shared VM. By way of the above,
1006          * thread groups also imply shared VM. Blocking this case allows
1007          * for various simplifications in other code.
1008          */
1009         if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1010                 return ERR_PTR(-EINVAL);
1011 
1012         /*
1013          * Siblings of global init remain as zombies on exit since they are
1014          * not reaped by their parent (swapper). To solve this and to avoid
1015          * multi-rooted process trees, prevent global and container-inits
1016          * from creating siblings.
1017          */
1018         if ((clone_flags & CLONE_PARENT) &&
1019                                 current->signal->flags & SIGNAL_UNKILLABLE)
1020                 return ERR_PTR(-EINVAL);
1021 
1022         retval = security_task_create(clone_flags);
1023         if (retval)
1024                 goto fork_out;
1025 
1026         retval = -ENOMEM;
1027         p = dup_task_struct(current);
1028         if (!p)
1029                 goto fork_out;
1030 
1031         ftrace_graph_init_task(p);
1032 
1033         rt_mutex_init_task(p);
1034 
1035 #ifdef CONFIG_PROVE_LOCKING
1036         DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1037         DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1038 #endif
1039         retval = -EAGAIN;
1040         if (atomic_read(&p->real_cred->user->processes) >=
1041                         p->signal->rlim[RLIMIT_NPROC].rlim_cur) {
1042                 if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
1043                     p->real_cred->user != INIT_USER)
1044                         goto bad_fork_free;
1045         }
1046 
1047         retval = copy_creds(p, clone_flags);
1048         if (retval < 0)
1049                 goto bad_fork_free;
1050 
1051         /*
1052          * If multiple threads are within copy_process(), then this check
1053          * triggers too late. This doesn't hurt, the check is only there
1054          * to stop root fork bombs.
1055          */
1056         retval = -EAGAIN;
1057         if (nr_threads >= max_threads)
1058                 goto bad_fork_cleanup_count;
1059 
1060         if (!try_module_get(task_thread_info(p)->exec_domain->module))
1061                 goto bad_fork_cleanup_count;
1062 
1063         p->did_exec = 0;
1064         delayacct_tsk_init(p);  /* Must remain after dup_task_struct() */
1065         copy_flags(clone_flags, p);
1066         INIT_LIST_HEAD(&p->children);
1067         INIT_LIST_HEAD(&p->sibling);
1068         rcu_copy_process(p);
1069         p->vfork_done = NULL;
1070         spin_lock_init(&p->alloc_lock);
1071 
1072         init_sigpending(&p->pending);
1073 
1074         p->utime = cputime_zero;
1075         p->stime = cputime_zero;
1076         p->gtime = cputime_zero;
1077         p->utimescaled = cputime_zero;
1078         p->stimescaled = cputime_zero;
1079         p->prev_utime = cputime_zero;
1080         p->prev_stime = cputime_zero;
1081 
1082         p->default_timer_slack_ns = current->timer_slack_ns;
1083 
1084         task_io_accounting_init(&p->ioac);
1085         acct_clear_integrals(p);
1086 
1087         posix_cpu_timers_init(p);
1088 
1089         p->lock_depth = -1;             /* -1 = no lock */
1090         do_posix_clock_monotonic_gettime(&p->start_time);
1091         p->real_start_time = p->start_time;
1092         monotonic_to_bootbased(&p->real_start_time);
1093         p->io_context = NULL;
1094         p->audit_context = NULL;
1095         cgroup_fork(p);
1096 #ifdef CONFIG_NUMA
1097         p->mempolicy = mpol_dup(p->mempolicy);
1098         if (IS_ERR(p->mempolicy)) {
1099                 retval = PTR_ERR(p->mempolicy);
1100                 p->mempolicy = NULL;
1101                 goto bad_fork_cleanup_cgroup;
1102         }
1103         mpol_fix_fork_child_flag(p);
1104 #endif
1105 #ifdef CONFIG_TRACE_IRQFLAGS
1106         p->irq_events = 0;
1107 #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1108         p->hardirqs_enabled = 1;
1109 #else
1110         p->hardirqs_enabled = 0;
1111 #endif
1112         p->hardirq_enable_ip = 0;
1113         p->hardirq_enable_event = 0;
1114         p->hardirq_disable_ip = _THIS_IP_;
1115         p->hardirq_disable_event = 0;
1116         p->softirqs_enabled = 1;
1117         p->softirq_enable_ip = _THIS_IP_;
1118         p->softirq_enable_event = 0;
1119         p->softirq_disable_ip = 0;
1120         p->softirq_disable_event = 0;
1121         p->hardirq_context = 0;
1122         p->softirq_context = 0;
1123 #endif
1124 #ifdef CONFIG_LOCKDEP
1125         p->lockdep_depth = 0; /* no locks held yet */
1126         p->curr_chain_key = 0;
1127         p->lockdep_recursion = 0;
1128 #endif
1129 
1130 #ifdef CONFIG_DEBUG_MUTEXES
1131         p->blocked_on = NULL; /* not blocked yet */
1132 #endif
1133 
1134         p->bts = NULL;
1135 
1136         /* Perform scheduler related setup. Assign this task to a CPU. */
1137         sched_fork(p, clone_flags);
1138 
1139         retval = perf_event_init_task(p);
1140         if (retval)
1141                 goto bad_fork_cleanup_policy;
1142 
1143         if ((retval = audit_alloc(p)))
1144                 goto bad_fork_cleanup_policy;
1145         retval = ccs_alloc_task_security(p);
1146         if (retval)
1147                 goto bad_fork_cleanup_audit;
1148         /* copy all the process information */
1149         if ((retval = copy_semundo(clone_flags, p)))
1150                 goto bad_fork_cleanup_audit;
1151         if ((retval = copy_files(clone_flags, p)))
1152                 goto bad_fork_cleanup_semundo;
1153         if ((retval = copy_fs(clone_flags, p)))
1154                 goto bad_fork_cleanup_files;
1155         if ((retval = copy_sighand(clone_flags, p)))
1156                 goto bad_fork_cleanup_fs;
1157         if ((retval = copy_signal(clone_flags, p)))
1158                 goto bad_fork_cleanup_sighand;
1159         if ((retval = copy_mm(clone_flags, p)))
1160                 goto bad_fork_cleanup_signal;
1161         if ((retval = copy_namespaces(clone_flags, p)))
1162                 goto bad_fork_cleanup_mm;
1163         if ((retval = copy_io(clone_flags, p)))
1164                 goto bad_fork_cleanup_namespaces;
1165         retval = copy_thread(clone_flags, stack_start, stack_size, p, regs);
1166         if (retval)
1167                 goto bad_fork_cleanup_io;
1168 
1169         if (pid != &init_struct_pid) {
1170                 retval = -ENOMEM;
1171                 pid = alloc_pid(p->nsproxy->pid_ns);
1172                 if (!pid)
1173                         goto bad_fork_cleanup_io;
1174 
1175                 if (clone_flags & CLONE_NEWPID) {
1176                         retval = pid_ns_prepare_proc(p->nsproxy->pid_ns);
1177                         if (retval < 0)
1178                                 goto bad_fork_free_pid;
1179                 }
1180         }
1181 
1182         p->pid = pid_nr(pid);
1183         p->tgid = p->pid;
1184         if (clone_flags & CLONE_THREAD)
1185                 p->tgid = current->tgid;
1186 
1187         if (current->nsproxy != p->nsproxy) {
1188                 retval = ns_cgroup_clone(p, pid);
1189                 if (retval)
1190                         goto bad_fork_free_pid;
1191         }
1192 
1193         p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1194         /*
1195          * Clear TID on mm_release()?
1196          */
1197         p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL;
1198 #ifdef CONFIG_FUTEX
1199         p->robust_list = NULL;
1200 #ifdef CONFIG_COMPAT
1201         p->compat_robust_list = NULL;
1202 #endif
1203         INIT_LIST_HEAD(&p->pi_state_list);
1204         p->pi_state_cache = NULL;
1205 #endif
1206         /*
1207          * sigaltstack should be cleared when sharing the same VM
1208          */
1209         if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1210                 p->sas_ss_sp = p->sas_ss_size = 0;
1211 
1212         /*
1213          * Syscall tracing should be turned off in the child regardless
1214          * of CLONE_PTRACE.
1215          */
1216         clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
1217 #ifdef TIF_SYSCALL_EMU
1218         clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1219 #endif
1220         clear_all_latency_tracing(p);
1221 
1222         /* ok, now we should be set up.. */
1223         p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
1224         p->pdeath_signal = 0;
1225         p->exit_state = 0;
1226 
1227         /*
1228          * Ok, make it visible to the rest of the system.
1229          * We dont wake it up yet.
1230          */
1231         p->group_leader = p;
1232         INIT_LIST_HEAD(&p->thread_group);
1233 
1234         /* Now that the task is set up, run cgroup callbacks if
1235          * necessary. We need to run them before the task is visible
1236          * on the tasklist. */
1237         cgroup_fork_callbacks(p);
1238         cgroup_callbacks_done = 1;
1239 
1240         /* Need tasklist lock for parent etc handling! */
1241         write_lock_irq(&tasklist_lock);
1242 
1243         /* CLONE_PARENT re-uses the old parent */
1244         if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
1245                 p->real_parent = current->real_parent;
1246                 p->parent_exec_id = current->parent_exec_id;
1247         } else {
1248                 p->real_parent = current;
1249                 p->parent_exec_id = current->self_exec_id;
1250         }
1251 
1252         spin_lock(&current->sighand->siglock);
1253 
1254         /*
1255          * Process group and session signals need to be delivered to just the
1256          * parent before the fork or both the parent and the child after the
1257          * fork. Restart if a signal comes in before we add the new process to
1258          * it's process group.
1259          * A fatal signal pending means that current will exit, so the new
1260          * thread can't slip out of an OOM kill (or normal SIGKILL).
1261          */
1262         recalc_sigpending();
1263         if (signal_pending(current)) {
1264                 spin_unlock(&current->sighand->siglock);
1265                 write_unlock_irq(&tasklist_lock);
1266                 retval = -ERESTARTNOINTR;
1267                 goto bad_fork_free_pid;
1268         }
1269 
1270         if (clone_flags & CLONE_THREAD) {
1271                 atomic_inc(&current->signal->count);
1272                 atomic_inc(&current->signal->live);
1273                 p->group_leader = current->group_leader;
1274                 list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
1275         }
1276 
1277         if (likely(p->pid)) {
1278                 list_add_tail(&p->sibling, &p->real_parent->children);
1279                 tracehook_finish_clone(p, clone_flags, trace);
1280 
1281                 if (thread_group_leader(p)) {
1282                         if (clone_flags & CLONE_NEWPID)
1283                                 p->nsproxy->pid_ns->child_reaper = p;
1284 
1285                         p->signal->leader_pid = pid;
1286                         tty_kref_put(p->signal->tty);
1287                         p->signal->tty = tty_kref_get(current->signal->tty);
1288                         attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
1289                         attach_pid(p, PIDTYPE_SID, task_session(current));
1290                         list_add_tail_rcu(&p->tasks, &init_task.tasks);
1291                         __get_cpu_var(process_counts)++;
1292                 }
1293                 attach_pid(p, PIDTYPE_PID, pid);
1294                 nr_threads++;
1295         }
1296 
1297         total_forks++;
1298         spin_unlock(&current->sighand->siglock);
1299         write_unlock_irq(&tasklist_lock);
1300         proc_fork_connector(p);
1301         cgroup_post_fork(p);
1302         perf_event_fork(p);
1303         return p;
1304 
1305 bad_fork_free_pid:
1306         if (pid != &init_struct_pid)
1307                 free_pid(pid);
1308 bad_fork_cleanup_io:
1309         if (p->io_context)
1310                 exit_io_context(p);
1311 bad_fork_cleanup_namespaces:
1312         exit_task_namespaces(p);
1313 bad_fork_cleanup_mm:
1314         if (p->mm)
1315                 mmput(p->mm);
1316 bad_fork_cleanup_signal:
1317         if (!(clone_flags & CLONE_THREAD))
1318                 __cleanup_signal(p->signal);
1319 bad_fork_cleanup_sighand:
1320         __cleanup_sighand(p->sighand);
1321 bad_fork_cleanup_fs:
1322         exit_fs(p); /* blocking */
1323 bad_fork_cleanup_files:
1324         exit_files(p); /* blocking */
1325 bad_fork_cleanup_semundo:
1326         exit_sem(p);
1327 bad_fork_cleanup_audit:
1328         audit_free(p);
1329         ccs_free_task_security(p);
1330 bad_fork_cleanup_policy:
1331         perf_event_free_task(p);
1332 #ifdef CONFIG_NUMA
1333         mpol_put(p->mempolicy);
1334 bad_fork_cleanup_cgroup:
1335 #endif
1336         cgroup_exit(p, cgroup_callbacks_done);
1337         delayacct_tsk_free(p);
1338         module_put(task_thread_info(p)->exec_domain->module);
1339 bad_fork_cleanup_count:
1340         atomic_dec(&p->cred->user->processes);
1341         exit_creds(p);
1342 bad_fork_free:
1343         free_task(p);
1344 fork_out:
1345         return ERR_PTR(retval);
1346 }
1347 
1348 noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
1349 {
1350         memset(regs, 0, sizeof(struct pt_regs));
1351         return regs;
1352 }
1353 
1354 struct task_struct * __cpuinit fork_idle(int cpu)
1355 {
1356         struct task_struct *task;
1357         struct pt_regs regs;
1358 
1359         task = copy_process(CLONE_VM, 0, idle_regs(&regs), 0, NULL,
1360                             &init_struct_pid, 0);
1361         if (!IS_ERR(task))
1362                 init_idle(task, cpu);
1363 
1364         return task;
1365 }
1366 
1367 /*
1368  *  Ok, this is the main fork-routine.
1369  *
1370  * It copies the process, and if successful kick-starts
1371  * it and waits for it to finish using the VM if required.
1372  */
1373 long do_fork(unsigned long clone_flags,
1374               unsigned long stack_start,
1375               struct pt_regs *regs,
1376               unsigned long stack_size,
1377               int __user *parent_tidptr,
1378               int __user *child_tidptr)
1379 {
1380         struct task_struct *p;
1381         int trace = 0;
1382         long nr;
1383 
1384         /*
1385          * Do some preliminary argument and permissions checking before we
1386          * actually start allocating stuff
1387          */
1388         if (clone_flags & CLONE_NEWUSER) {
1389                 if (clone_flags & CLONE_THREAD)
1390                         return -EINVAL;
1391                 /* hopefully this check will go away when userns support is
1392                  * complete
1393                  */
1394                 if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) ||
1395                                 !capable(CAP_SETGID))
1396                         return -EPERM;
1397         }
1398 
1399         /*
1400          * We hope to recycle these flags after 2.6.26
1401          */
1402         if (unlikely(clone_flags & CLONE_STOPPED)) {
1403                 static int __read_mostly count = 100;
1404 
1405                 if (count > 0 && printk_ratelimit()) {
1406                         char comm[TASK_COMM_LEN];
1407 
1408                         count--;
1409                         printk(KERN_INFO "fork(): process `%s' used deprecated "
1410                                         "clone flags 0x%lx\n",
1411                                 get_task_comm(comm, current),
1412                                 clone_flags & CLONE_STOPPED);
1413                 }
1414         }
1415 
1416         /*
1417          * When called from kernel_thread, don't do user tracing stuff.
1418          */
1419         if (likely(user_mode(regs)))
1420                 trace = tracehook_prepare_clone(clone_flags);
1421 
1422         p = copy_process(clone_flags, stack_start, regs, stack_size,
1423                          child_tidptr, NULL, trace);
1424         /*
1425          * Do this prior waking up the new thread - the thread pointer
1426          * might get invalid after that point, if the thread exits quickly.
1427          */
1428         if (!IS_ERR(p)) {
1429                 struct completion vfork;
1430 
1431                 trace_sched_process_fork(current, p);
1432 
1433                 nr = task_pid_vnr(p);
1434 
1435                 if (clone_flags & CLONE_PARENT_SETTID)
1436                         put_user(nr, parent_tidptr);
1437 
1438                 if (clone_flags & CLONE_VFORK) {
1439                         p->vfork_done = &vfork;
1440                         init_completion(&vfork);
1441                 }
1442 
1443                 audit_finish_fork(p);
1444                 tracehook_report_clone(regs, clone_flags, nr, p);
1445 
1446                 /*
1447                  * We set PF_STARTING at creation in case tracing wants to
1448                  * use this to distinguish a fully live task from one that
1449                  * hasn't gotten to tracehook_report_clone() yet.  Now we
1450                  * clear it and set the child going.
1451                  */
1452                 p->flags &= ~PF_STARTING;
1453 
1454                 if (unlikely(clone_flags & CLONE_STOPPED)) {
1455                         /*
1456                          * We'll start up with an immediate SIGSTOP.
1457                          */
1458                         sigaddset(&p->pending.signal, SIGSTOP);
1459                         set_tsk_thread_flag(p, TIF_SIGPENDING);
1460                         __set_task_state(p, TASK_STOPPED);
1461                 } else {
1462                         wake_up_new_task(p, clone_flags);
1463                 }
1464 
1465                 tracehook_report_clone_complete(trace, regs,
1466                                                 clone_flags, nr, p);
1467 
1468                 if (clone_flags & CLONE_VFORK) {
1469                         freezer_do_not_count();
1470                         wait_for_completion(&vfork);
1471                         freezer_count();
1472                         tracehook_report_vfork_done(p, nr);
1473                 }
1474         } else {
1475                 nr = PTR_ERR(p);
1476         }
1477         return nr;
1478 }
1479 
1480 #ifndef ARCH_MIN_MMSTRUCT_ALIGN
1481 #define ARCH_MIN_MMSTRUCT_ALIGN 0
1482 #endif
1483 
1484 static void sighand_ctor(void *data)
1485 {
1486         struct sighand_struct *sighand = data;
1487 
1488         spin_lock_init(&sighand->siglock);
1489         init_waitqueue_head(&sighand->signalfd_wqh);
1490 }
1491 
1492 void __init proc_caches_init(void)
1493 {
1494         sighand_cachep = kmem_cache_create("sighand_cache",
1495                         sizeof(struct sighand_struct), 0,
1496                         SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
1497                         SLAB_NOTRACK, sighand_ctor);
1498         signal_cachep = kmem_cache_create("signal_cache",
1499                         sizeof(struct signal_struct), 0,
1500                         SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1501         files_cachep = kmem_cache_create("files_cache",
1502                         sizeof(struct files_struct), 0,
1503                         SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1504         fs_cachep = kmem_cache_create("fs_cache",
1505                         sizeof(struct fs_struct), 0,
1506                         SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1507         mm_cachep = kmem_cache_create("mm_struct",
1508                         sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
1509                         SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
1510         vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
1511         mmap_init();
1512 }
1513 
1514 /*
1515  * Check constraints on flags passed to the unshare system call and
1516  * force unsharing of additional process context as appropriate.
1517  */
1518 static void check_unshare_flags(unsigned long *flags_ptr)
1519 {
1520         /*
1521          * If unsharing a thread from a thread group, must also
1522          * unshare vm.
1523          */
1524         if (*flags_ptr & CLONE_THREAD)
1525                 *flags_ptr |= CLONE_VM;
1526 
1527         /*
1528          * If unsharing vm, must also unshare signal handlers.
1529          */
1530         if (*flags_ptr & CLONE_VM)
1531                 *flags_ptr |= CLONE_SIGHAND;
1532 
1533         /*
1534          * If unsharing signal handlers and the task was created
1535          * using CLONE_THREAD, then must unshare the thread
1536          */
1537         if ((*flags_ptr & CLONE_SIGHAND) &&
1538             (atomic_read(&current->signal->count) > 1))
1539                 *flags_ptr |= CLONE_THREAD;
1540 
1541         /*
1542          * If unsharing namespace, must also unshare filesystem information.
1543          */
1544         if (*flags_ptr & CLONE_NEWNS)
1545                 *flags_ptr |= CLONE_FS;
1546 }
1547 
1548 /*
1549  * Unsharing of tasks created with CLONE_THREAD is not supported yet
1550  */
1551 static int unshare_thread(unsigned long unshare_flags)
1552 {
1553         if (unshare_flags & CLONE_THREAD)
1554                 return -EINVAL;
1555 
1556         return 0;
1557 }
1558 
1559 /*
1560  * Unshare the filesystem structure if it is being shared
1561  */
1562 static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1563 {
1564         struct fs_struct *fs = current->fs;
1565 
1566         if (!(unshare_flags & CLONE_FS) || !fs)
1567                 return 0;
1568 
1569         /* don't need lock here; in the worst case we'll do useless copy */
1570         if (fs->users == 1)
1571                 return 0;
1572 
1573         *new_fsp = copy_fs_struct(fs);
1574         if (!*new_fsp)
1575                 return -ENOMEM;
1576 
1577         return 0;
1578 }
1579 
1580 /*
1581  * Unsharing of sighand is not supported yet
1582  */
1583 static int unshare_sighand(unsigned long unshare_flags, struct sighand_struct **new_sighp)
1584 {
1585         struct sighand_struct *sigh = current->sighand;
1586 
1587         if ((unshare_flags & CLONE_SIGHAND) && atomic_read(&sigh->count) > 1)
1588                 return -EINVAL;
1589         else
1590                 return 0;
1591 }
1592 
1593 /*
1594  * Unshare vm if it is being shared
1595  */
1596 static int unshare_vm(unsigned long unshare_flags, struct mm_struct **new_mmp)
1597 {
1598         struct mm_struct *mm = current->mm;
1599 
1600         if ((unshare_flags & CLONE_VM) &&
1601             (mm && atomic_read(&mm->mm_users) > 1)) {
1602                 return -EINVAL;
1603         }
1604 
1605         return 0;
1606 }
1607 
1608 /*
1609  * Unshare file descriptor table if it is being shared
1610  */
1611 static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1612 {
1613         struct files_struct *fd = current->files;
1614         int error = 0;
1615 
1616         if ((unshare_flags & CLONE_FILES) &&
1617             (fd && atomic_read(&fd->count) > 1)) {
1618                 *new_fdp = dup_fd(fd, &error);
1619                 if (!*new_fdp)
1620                         return error;
1621         }
1622 
1623         return 0;
1624 }
1625 
1626 /*
1627  * unshare allows a process to 'unshare' part of the process
1628  * context which was originally shared using clone.  copy_*
1629  * functions used by do_fork() cannot be used here directly
1630  * because they modify an inactive task_struct that is being
1631  * constructed. Here we are modifying the current, active,
1632  * task_struct.
1633  */
1634 SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
1635 {
1636         int err = 0;
1637         struct fs_struct *fs, *new_fs = NULL;
1638         struct sighand_struct *new_sigh = NULL;
1639         struct mm_struct *mm, *new_mm = NULL, *active_mm = NULL;
1640         struct files_struct *fd, *new_fd = NULL;
1641         struct nsproxy *new_nsproxy = NULL;
1642         int do_sysvsem = 0;
1643 
1644         check_unshare_flags(&unshare_flags);
1645 
1646         /* Return -EINVAL for all unsupported flags */
1647         err = -EINVAL;
1648         if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
1649                                 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
1650                                 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET))
1651                 goto bad_unshare_out;
1652 
1653         /*
1654          * CLONE_NEWIPC must also detach from the undolist: after switching
1655          * to a new ipc namespace, the semaphore arrays from the old
1656          * namespace are unreachable.
1657          */
1658         if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
1659                 do_sysvsem = 1;
1660         if ((err = unshare_thread(unshare_flags)))
1661                 goto bad_unshare_out;
1662         if ((err = unshare_fs(unshare_flags, &new_fs)))
1663                 goto bad_unshare_cleanup_thread;
1664         if ((err = unshare_sighand(unshare_flags, &new_sigh)))
1665                 goto bad_unshare_cleanup_fs;
1666         if ((err = unshare_vm(unshare_flags, &new_mm)))
1667                 goto bad_unshare_cleanup_sigh;
1668         if ((err = unshare_fd(unshare_flags, &new_fd)))
1669                 goto bad_unshare_cleanup_vm;
1670         if ((err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
1671                         new_fs)))
1672                 goto bad_unshare_cleanup_fd;
1673 
1674         if (new_fs ||  new_mm || new_fd || do_sysvsem || new_nsproxy) {
1675                 if (do_sysvsem) {
1676                         /*
1677                          * CLONE_SYSVSEM is equivalent to sys_exit().
1678                          */
1679                         exit_sem(current);
1680                 }
1681 
1682                 if (new_nsproxy) {
1683                         switch_task_namespaces(current, new_nsproxy);
1684                         new_nsproxy = NULL;
1685                 }
1686 
1687                 task_lock(current);
1688 
1689                 if (new_fs) {
1690                         fs = current->fs;
1691                         write_lock(&fs->lock);
1692                         current->fs = new_fs;
1693                         if (--fs->users)
1694                                 new_fs = NULL;
1695                         else
1696                                 new_fs = fs;
1697                         write_unlock(&fs->lock);
1698                 }
1699 
1700                 if (new_mm) {
1701                         mm = current->mm;
1702                         active_mm = current->active_mm;
1703                         current->mm = new_mm;
1704                         current->active_mm = new_mm;
1705                         activate_mm(active_mm, new_mm);
1706                         new_mm = mm;
1707                 }
1708 
1709                 if (new_fd) {
1710                         fd = current->files;
1711                         current->files = new_fd;
1712                         new_fd = fd;
1713                 }
1714 
1715                 task_unlock(current);
1716         }
1717 
1718         if (new_nsproxy)
1719                 put_nsproxy(new_nsproxy);
1720 
1721 bad_unshare_cleanup_fd:
1722         if (new_fd)
1723                 put_files_struct(new_fd);
1724 
1725 bad_unshare_cleanup_vm:
1726         if (new_mm)
1727                 mmput(new_mm);
1728 
1729 bad_unshare_cleanup_sigh:
1730         if (new_sigh)
1731                 if (atomic_dec_and_test(&new_sigh->count))
1732                         kmem_cache_free(sighand_cachep, new_sigh);
1733 
1734 bad_unshare_cleanup_fs:
1735         if (new_fs)
1736                 free_fs_struct(new_fs);
1737 
1738 bad_unshare_cleanup_thread:
1739 bad_unshare_out:
1740         return err;
1741 }
1742 
1743 /*
1744  *      Helper to unshare the files of the current task.
1745  *      We don't want to expose copy_files internals to
1746  *      the exec layer of the kernel.
1747  */
1748 
1749 int unshare_files(struct files_struct **displaced)
1750 {
1751         struct task_struct *task = current;
1752         struct files_struct *copy = NULL;
1753         int error;
1754 
1755         error = unshare_fd(CLONE_FILES, &copy);
1756         if (error || !copy) {
1757                 *displaced = NULL;
1758                 return error;
1759         }
1760         *displaced = task->files;
1761         task_lock(task);
1762         task->files = copy;
1763         task_unlock(task);
1764         return 0;
1765 }
1766 

~ [ 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