4 * Copyright (C) 1991, 1992 Linus Torvalds
8 * 'fork.c' contains the help-routines for the 'fork' system call
9 * (see also entry.S and others).
10 * Fork is rather simple, once you get the hang of it, but the memory
11 * management can be a bitch. See 'mm/memory.c': 'copy_page_range()'
14 #include <linux/config.h>
15 #include <linux/slab.h>
16 #include <linux/init.h>
17 #include <linux/unistd.h>
18 #include <linux/smp_lock.h>
19 #include <linux/module.h>
20 #include <linux/vmalloc.h>
21 #include <linux/completion.h>
22 #include <linux/namespace.h>
23 #include <linux/personality.h>
24 #include <linux/mempolicy.h>
25 #include <linux/sem.h>
26 #include <linux/file.h>
27 #include <linux/key.h>
28 #include <linux/binfmts.h>
29 #include <linux/mman.h>
31 #include <linux/cpu.h>
32 #include <linux/security.h>
33 #include <linux/swap.h>
34 #include <linux/syscalls.h>
35 #include <linux/jiffies.h>
36 #include <linux/futex.h>
37 #include <linux/ptrace.h>
38 #include <linux/mount.h>
39 #include <linux/audit.h>
40 #include <linux/profile.h>
41 #include <linux/rmap.h>
42 #include <linux/ckrm.h>
43 #include <linux/ckrm_tsk.h>
44 #include <linux/ckrm_mem_inline.h>
45 #include <linux/vs_network.h>
46 #include <linux/vs_limit.h>
47 #include <linux/vs_memory.h>
49 #include <asm/pgtable.h>
50 #include <asm/pgalloc.h>
51 #include <asm/uaccess.h>
52 #include <asm/mmu_context.h>
53 #include <asm/cacheflush.h>
54 #include <asm/tlbflush.h>
56 /* The idle threads do not count..
57 * Protected by write_lock_irq(&tasklist_lock)
62 unsigned long total_forks; /* Handle normal Linux uptimes. */
64 DEFINE_PER_CPU(unsigned long, process_counts) = 0;
66 rwlock_t tasklist_lock __cacheline_aligned = RW_LOCK_UNLOCKED; /* outer */
68 EXPORT_SYMBOL(tasklist_lock);
70 int nr_processes(void)
75 for_each_online_cpu(cpu)
76 total += per_cpu(process_counts, cpu);
81 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
82 # define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL)
83 # define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk))
84 static kmem_cache_t *task_struct_cachep;
87 void free_task(struct task_struct *tsk)
89 free_thread_info(tsk->thread_info);
90 clr_vx_info(&tsk->vx_info);
91 clr_nx_info(&tsk->nx_info);
92 free_task_struct(tsk);
94 EXPORT_SYMBOL(free_task);
96 void __put_task_struct(struct task_struct *tsk)
98 WARN_ON(!(tsk->exit_state & (EXIT_DEAD | EXIT_ZOMBIE)));
99 WARN_ON(atomic_read(&tsk->usage));
100 WARN_ON(tsk == current);
102 if (unlikely(tsk->audit_context))
104 security_task_free(tsk);
106 put_group_info(tsk->group_info);
108 if (!profile_handoff_task(tsk))
112 void __init fork_init(unsigned long mempages)
114 #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
115 #ifndef ARCH_MIN_TASKALIGN
116 #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
118 /* create a slab on which task_structs can be allocated */
120 kmem_cache_create("task_struct", sizeof(struct task_struct),
121 ARCH_MIN_TASKALIGN, SLAB_PANIC, NULL, NULL);
125 * The default maximum number of threads is set to a safe
126 * value: the thread structures can take up at most half
129 max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
132 * we need to allow at least 20 threads to boot a system
137 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
138 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
141 static struct task_struct *dup_task_struct(struct task_struct *orig)
143 struct task_struct *tsk;
144 struct thread_info *ti;
146 prepare_to_copy(orig);
148 tsk = alloc_task_struct();
152 ti = alloc_thread_info(tsk);
154 free_task_struct(tsk);
158 *ti = *orig->thread_info;
160 tsk->thread_info = ti;
163 ckrm_cb_newtask(tsk);
164 ckrm_task_mm_init(tsk);
165 /* One for us, one for whoever does the "release_task()" (usually parent) */
166 atomic_set(&tsk->usage,2);
171 static inline int dup_mmap(struct mm_struct * mm, struct mm_struct * oldmm)
173 struct vm_area_struct * mpnt, *tmp, **pprev;
174 struct rb_node **rb_link, *rb_parent;
176 unsigned long charge;
177 struct mempolicy *pol;
179 down_write(&oldmm->mmap_sem);
180 flush_cache_mm(current->mm);
183 mm->mmap_cache = NULL;
184 mm->free_area_cache = oldmm->mmap_base;
188 cpus_clear(mm->cpu_vm_mask);
190 rb_link = &mm->mm_rb.rb_node;
194 for (mpnt = current->mm->mmap ; mpnt ; mpnt = mpnt->vm_next) {
197 if (mpnt->vm_flags & VM_DONTCOPY) {
198 __vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
203 if (mpnt->vm_flags & VM_ACCOUNT) {
204 unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
205 if (security_vm_enough_memory(len))
209 tmp = kmem_cache_alloc(vm_area_cachep, SLAB_KERNEL);
213 pol = mpol_copy(vma_policy(mpnt));
214 retval = PTR_ERR(pol);
216 goto fail_nomem_policy;
217 vma_set_policy(tmp, pol);
218 tmp->vm_flags &= ~VM_LOCKED;
224 struct inode *inode = file->f_dentry->d_inode;
226 if (tmp->vm_flags & VM_DENYWRITE)
227 atomic_dec(&inode->i_writecount);
229 /* insert tmp into the share list, just after mpnt */
230 spin_lock(&file->f_mapping->i_mmap_lock);
231 flush_dcache_mmap_lock(file->f_mapping);
232 vma_prio_tree_add(tmp, mpnt);
233 flush_dcache_mmap_unlock(file->f_mapping);
234 spin_unlock(&file->f_mapping->i_mmap_lock);
238 * Link in the new vma and copy the page table entries:
239 * link in first so that swapoff can see swap entries,
240 * and try_to_unmap_one's find_vma find the new vma.
242 spin_lock(&mm->page_table_lock);
244 pprev = &tmp->vm_next;
246 __vma_link_rb(mm, tmp, rb_link, rb_parent);
247 rb_link = &tmp->vm_rb.rb_right;
248 rb_parent = &tmp->vm_rb;
251 retval = copy_page_range(mm, current->mm, tmp);
252 spin_unlock(&mm->page_table_lock);
254 if (tmp->vm_ops && tmp->vm_ops->open)
255 tmp->vm_ops->open(tmp);
263 flush_tlb_mm(current->mm);
264 up_write(&oldmm->mmap_sem);
267 kmem_cache_free(vm_area_cachep, tmp);
270 vm_unacct_memory(charge);
274 static inline int mm_alloc_pgd(struct mm_struct * mm)
276 mm->pgd = pgd_alloc(mm);
277 if (unlikely(!mm->pgd))
282 static inline void mm_free_pgd(struct mm_struct * mm)
287 #define dup_mmap(mm, oldmm) (0)
288 #define mm_alloc_pgd(mm) (0)
289 #define mm_free_pgd(mm)
290 #endif /* CONFIG_MMU */
292 spinlock_t mmlist_lock __cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED;
294 #define allocate_mm() (kmem_cache_alloc(mm_cachep, SLAB_KERNEL))
295 #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
297 #include <linux/init_task.h>
299 static struct mm_struct * mm_init(struct mm_struct * mm)
301 atomic_set(&mm->mm_users, 1);
302 atomic_set(&mm->mm_count, 1);
303 init_rwsem(&mm->mmap_sem);
304 INIT_LIST_HEAD(&mm->mmlist);
305 mm->core_waiters = 0;
307 spin_lock_init(&mm->page_table_lock);
308 rwlock_init(&mm->ioctx_list_lock);
309 mm->ioctx_list = NULL;
310 mm->default_kioctx = (struct kioctx)INIT_KIOCTX(mm->default_kioctx, *mm);
311 mm->free_area_cache = TASK_UNMAPPED_BASE;
314 if (likely(!mm_alloc_pgd(mm))) {
316 set_vx_info(&mm->mm_vx_info, current->vx_info);
324 * Allocate and initialize an mm_struct.
326 struct mm_struct * mm_alloc(void)
328 struct mm_struct * mm;
332 memset(mm, 0, sizeof(*mm));
334 ckrm_mm_setclass(mm, ckrm_get_mem_class(current));
340 * Called when the last reference to the mm
341 * is dropped: either by a lazy thread or by
342 * mmput. Free the page directory and the mm.
344 void fastcall __mmdrop(struct mm_struct *mm)
346 BUG_ON(mm == &init_mm);
349 ckrm_mm_clearclass(mm);
350 clr_vx_info(&mm->mm_vx_info);
355 * Decrement the use count and release all resources for an mm.
357 void mmput(struct mm_struct *mm)
359 if (atomic_dec_and_test(&mm->mm_users)) {
362 if (!list_empty(&mm->mmlist)) {
363 spin_lock(&mmlist_lock);
364 list_del(&mm->mmlist);
365 spin_unlock(&mmlist_lock);
371 EXPORT_SYMBOL_GPL(mmput);
374 * get_task_mm - acquire a reference to the task's mm
376 * Returns %NULL if the task has no mm. Checks PF_BORROWED_MM (meaning
377 * this kernel workthread has transiently adopted a user mm with use_mm,
378 * to do its AIO) is not set and if so returns a reference to it, after
379 * bumping up the use count. User must release the mm via mmput()
380 * after use. Typically used by /proc and ptrace.
382 struct mm_struct *get_task_mm(struct task_struct *task)
384 struct mm_struct *mm;
389 if (task->flags & PF_BORROWED_MM)
392 atomic_inc(&mm->mm_users);
397 EXPORT_SYMBOL_GPL(get_task_mm);
399 /* Please note the differences between mmput and mm_release.
400 * mmput is called whenever we stop holding onto a mm_struct,
401 * error success whatever.
403 * mm_release is called after a mm_struct has been removed
404 * from the current process.
406 * This difference is important for error handling, when we
407 * only half set up a mm_struct for a new process and need to restore
408 * the old one. Because we mmput the new mm_struct before
409 * restoring the old one. . .
410 * Eric Biederman 10 January 1998
412 void mm_release(struct task_struct *tsk, struct mm_struct *mm)
414 struct completion *vfork_done = tsk->vfork_done;
416 /* Get rid of any cached register state */
417 deactivate_mm(tsk, mm);
419 /* notify parent sleeping on vfork() */
421 tsk->vfork_done = NULL;
422 complete(vfork_done);
424 if (tsk->clear_child_tid && atomic_read(&mm->mm_users) > 1) {
425 u32 __user * tidptr = tsk->clear_child_tid;
426 tsk->clear_child_tid = NULL;
429 * We don't check the error code - if userspace has
430 * not set up a proper pointer then tough luck.
433 sys_futex(tidptr, FUTEX_WAKE, 1, NULL, NULL, 0);
437 static int copy_mm(unsigned long clone_flags, struct task_struct * tsk)
439 struct mm_struct * mm, *oldmm;
442 tsk->min_flt = tsk->maj_flt = 0;
443 tsk->nvcsw = tsk->nivcsw = 0;
446 tsk->active_mm = NULL;
449 * Are we cloning a kernel thread?
451 * We need to steal a active VM for that..
457 if (clone_flags & CLONE_VM) {
458 atomic_inc(&oldmm->mm_users);
461 * There are cases where the PTL is held to ensure no
462 * new threads start up in user mode using an mm, which
463 * allows optimizing out ipis; the tlb_gather_mmu code
466 spin_unlock_wait(&oldmm->page_table_lock);
475 /* Copy the current MM stuff.. */
476 memcpy(mm, oldmm, sizeof(*mm));
477 mm->mm_vx_info = NULL;
481 if (init_new_context(tsk,mm))
484 retval = dup_mmap(mm, oldmm);
489 ckrm_mm_setclass(mm, oldmm->memclass);
492 ckrm_init_mm_to_task(mm, tsk);
502 * If init_new_context() failed, we cannot use mmput() to free the mm
503 * because it calls destroy_context()
510 static inline struct fs_struct *__copy_fs_struct(struct fs_struct *old)
512 struct fs_struct *fs = kmem_cache_alloc(fs_cachep, GFP_KERNEL);
513 /* We don't need to lock fs - think why ;-) */
515 atomic_set(&fs->count, 1);
516 rwlock_init(&fs->lock);
517 fs->umask = old->umask;
518 read_lock(&old->lock);
519 fs->rootmnt = mntget(old->rootmnt);
520 fs->root = dget(old->root);
521 fs->pwdmnt = mntget(old->pwdmnt);
522 fs->pwd = dget(old->pwd);
524 fs->altrootmnt = mntget(old->altrootmnt);
525 fs->altroot = dget(old->altroot);
527 fs->altrootmnt = NULL;
530 read_unlock(&old->lock);
535 struct fs_struct *copy_fs_struct(struct fs_struct *old)
537 return __copy_fs_struct(old);
540 EXPORT_SYMBOL_GPL(copy_fs_struct);
542 static inline int copy_fs(unsigned long clone_flags, struct task_struct * tsk)
544 if (clone_flags & CLONE_FS) {
545 atomic_inc(¤t->fs->count);
548 tsk->fs = __copy_fs_struct(current->fs);
554 static int count_open_files(struct files_struct *files, int size)
558 /* Find the last open fd */
559 for (i = size/(8*sizeof(long)); i > 0; ) {
560 if (files->open_fds->fds_bits[--i])
563 i = (i+1) * 8 * sizeof(long);
567 static int copy_files(unsigned long clone_flags, struct task_struct * tsk)
569 struct files_struct *oldf, *newf;
570 struct file **old_fds, **new_fds;
571 int open_files, nfds, size, i, error = 0;
574 * A background process may not have any files ...
576 oldf = current->files;
580 if (clone_flags & CLONE_FILES) {
581 atomic_inc(&oldf->count);
586 * Note: we may be using current for both targets (See exec.c)
587 * This works because we cache current->files (old) as oldf. Don't
592 newf = kmem_cache_alloc(files_cachep, SLAB_KERNEL);
596 atomic_set(&newf->count, 1);
598 spin_lock_init(&newf->file_lock);
600 newf->max_fds = NR_OPEN_DEFAULT;
601 newf->max_fdset = __FD_SETSIZE;
602 newf->close_on_exec = &newf->close_on_exec_init;
603 newf->open_fds = &newf->open_fds_init;
604 newf->fd = &newf->fd_array[0];
606 /* We don't yet have the oldf readlock, but even if the old
607 fdset gets grown now, we'll only copy up to "size" fds */
608 size = oldf->max_fdset;
609 if (size > __FD_SETSIZE) {
611 spin_lock(&newf->file_lock);
612 error = expand_fdset(newf, size-1);
613 spin_unlock(&newf->file_lock);
617 spin_lock(&oldf->file_lock);
619 open_files = count_open_files(oldf, size);
622 * Check whether we need to allocate a larger fd array.
623 * Note: we're not a clone task, so the open count won't
626 nfds = NR_OPEN_DEFAULT;
627 if (open_files > nfds) {
628 spin_unlock(&oldf->file_lock);
630 spin_lock(&newf->file_lock);
631 error = expand_fd_array(newf, open_files-1);
632 spin_unlock(&newf->file_lock);
635 nfds = newf->max_fds;
636 spin_lock(&oldf->file_lock);
642 memcpy(newf->open_fds->fds_bits, oldf->open_fds->fds_bits, open_files/8);
643 memcpy(newf->close_on_exec->fds_bits, oldf->close_on_exec->fds_bits, open_files/8);
645 for (i = open_files; i != 0; i--) {
646 struct file *f = *old_fds++;
651 * The fd may be claimed in the fd bitmap but not yet
652 * instantiated in the files array if a sibling thread
653 * is partway through open(). So make sure that this
654 * fd is available to the new process.
656 FD_CLR(open_files - i, newf->open_fds);
660 spin_unlock(&oldf->file_lock);
662 /* compute the remainder to be cleared */
663 size = (newf->max_fds - open_files) * sizeof(struct file *);
665 /* This is long word aligned thus could use a optimized version */
666 memset(new_fds, 0, size);
668 if (newf->max_fdset > open_files) {
669 int left = (newf->max_fdset-open_files)/8;
670 int start = open_files / (8 * sizeof(unsigned long));
672 memset(&newf->open_fds->fds_bits[start], 0, left);
673 memset(&newf->close_on_exec->fds_bits[start], 0, left);
682 free_fdset (newf->close_on_exec, newf->max_fdset);
683 free_fdset (newf->open_fds, newf->max_fdset);
684 kmem_cache_free(files_cachep, newf);
689 * Helper to unshare the files of the current task.
690 * We don't want to expose copy_files internals to
691 * the exec layer of the kernel.
694 int unshare_files(void)
696 struct files_struct *files = current->files;
702 /* This can race but the race causes us to copy when we don't
703 need to and drop the copy */
704 if(atomic_read(&files->count) == 1)
706 atomic_inc(&files->count);
709 rc = copy_files(0, current);
711 current->files = files;
715 EXPORT_SYMBOL(unshare_files);
717 static inline int copy_sighand(unsigned long clone_flags, struct task_struct * tsk)
719 struct sighand_struct *sig;
721 if (clone_flags & (CLONE_SIGHAND | CLONE_THREAD)) {
722 atomic_inc(¤t->sighand->count);
725 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
729 spin_lock_init(&sig->siglock);
730 atomic_set(&sig->count, 1);
731 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
735 static inline int copy_signal(unsigned long clone_flags, struct task_struct * tsk)
737 struct signal_struct *sig;
739 if (clone_flags & CLONE_THREAD) {
740 atomic_inc(¤t->signal->count);
741 atomic_inc(¤t->signal->live);
744 sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL);
748 atomic_set(&sig->count, 1);
749 atomic_set(&sig->live, 1);
751 sig->group_exit_code = 0;
752 sig->group_exit_task = NULL;
753 sig->group_stop_count = 0;
755 sig->curr_target = NULL;
756 init_sigpending(&sig->shared_pending);
757 INIT_LIST_HEAD(&sig->posix_timers);
759 sig->tty = current->signal->tty;
760 sig->pgrp = process_group(current);
761 sig->session = current->signal->session;
762 sig->leader = 0; /* session leadership doesn't inherit */
763 sig->tty_old_pgrp = 0;
765 sig->utime = sig->stime = sig->cutime = sig->cstime = 0;
766 sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0;
767 sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0;
769 task_lock(current->group_leader);
770 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
771 task_unlock(current->group_leader);
776 static inline void copy_flags(unsigned long clone_flags, struct task_struct *p)
778 unsigned long new_flags = p->flags;
780 new_flags &= ~PF_SUPERPRIV;
781 new_flags |= PF_FORKNOEXEC;
782 if (!(clone_flags & CLONE_PTRACE))
784 p->flags = new_flags;
787 asmlinkage long sys_set_tid_address(int __user *tidptr)
789 current->clear_child_tid = tidptr;
795 * This creates a new process as a copy of the old one,
796 * but does not actually start it yet.
798 * It copies the registers, and all the appropriate
799 * parts of the process environment (as per the clone
800 * flags). The actual kick-off is left to the caller.
802 static task_t *copy_process(unsigned long clone_flags,
803 unsigned long stack_start,
804 struct pt_regs *regs,
805 unsigned long stack_size,
806 int __user *parent_tidptr,
807 int __user *child_tidptr,
811 struct task_struct *p = NULL;
814 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
815 return ERR_PTR(-EINVAL);
818 * Thread groups must share signals as well, and detached threads
819 * can only be started up within the thread group.
821 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
822 return ERR_PTR(-EINVAL);
825 * Shared signal handlers imply shared VM. By way of the above,
826 * thread groups also imply shared VM. Blocking this case allows
827 * for various simplifications in other code.
829 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
830 return ERR_PTR(-EINVAL);
832 retval = security_task_create(clone_flags);
837 p = dup_task_struct(current);
843 set_vx_info(&p->vx_info, current->vx_info);
845 set_nx_info(&p->nx_info, current->nx_info);
847 /* check vserver memory */
848 if (p->mm && !(clone_flags & CLONE_VM)) {
849 if (vx_vmpages_avail(p->mm, p->mm->total_vm))
850 vx_pages_add(p->mm->mm_vx_info, RLIMIT_AS, p->mm->total_vm);
854 if (p->mm && vx_flags(VXF_FORK_RSS, 0)) {
855 if (!vx_rsspages_avail(p->mm, p->mm->rss))
856 goto bad_fork_cleanup_vm;
860 set_vx_info(&p->vx_info, current->vx_info);
862 set_nx_info(&p->nx_info, current->nx_info);
864 /* check vserver memory */
865 if (p->mm && !(clone_flags & CLONE_VM)) {
866 if (vx_vmpages_avail(p->mm, p->mm->total_vm))
867 vx_pages_add(p->mm->mm_vx_info, RLIMIT_AS, p->mm->total_vm);
871 if (p->mm && vx_flags(VXF_FORK_RSS, 0)) {
872 if (!vx_rsspages_avail(p->mm, p->mm->rss))
873 goto bad_fork_cleanup_vm;
877 if (!vx_nproc_avail(1))
878 goto bad_fork_cleanup_vm;
880 if (atomic_read(&p->user->processes) >=
881 p->signal->rlim[RLIMIT_NPROC].rlim_cur) {
882 if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
883 p->user != &root_user)
884 goto bad_fork_cleanup_vm;
887 atomic_inc(&p->user->__count);
888 atomic_inc(&p->user->processes);
889 get_group_info(p->group_info);
892 * If multiple threads are within copy_process(), then this check
893 * triggers too late. This doesn't hurt, the check is only there
894 * to stop root fork bombs.
896 if (nr_threads >= max_threads)
897 goto bad_fork_cleanup_count;
899 if (!try_module_get(p->thread_info->exec_domain->module))
900 goto bad_fork_cleanup_count;
902 if (p->binfmt && !try_module_get(p->binfmt->module))
903 goto bad_fork_cleanup_put_domain;
907 copy_flags(clone_flags, p);
910 if (clone_flags & CLONE_PARENT_SETTID)
911 if (put_user(p->pid, parent_tidptr))
912 goto bad_fork_cleanup;
914 p->proc_dentry = NULL;
916 INIT_LIST_HEAD(&p->children);
917 INIT_LIST_HEAD(&p->sibling);
918 init_waitqueue_head(&p->wait_chldexit);
919 p->vfork_done = NULL;
920 spin_lock_init(&p->alloc_lock);
921 spin_lock_init(&p->proc_lock);
923 clear_tsk_thread_flag(p, TIF_SIGPENDING);
924 init_sigpending(&p->pending);
926 p->it_real_value = p->it_virt_value = p->it_prof_value = 0;
927 p->it_real_incr = p->it_virt_incr = p->it_prof_incr = 0;
928 init_timer(&p->real_timer);
929 p->real_timer.data = (unsigned long) p;
931 p->utime = p->stime = 0;
932 p->lock_depth = -1; /* -1 = no lock */
933 do_posix_clock_monotonic_gettime(&p->start_time);
935 p->io_context = NULL;
937 p->audit_context = NULL;
939 p->mempolicy = mpol_copy(p->mempolicy);
940 if (IS_ERR(p->mempolicy)) {
941 retval = PTR_ERR(p->mempolicy);
943 goto bad_fork_cleanup;
948 if (clone_flags & CLONE_THREAD)
949 p->tgid = current->tgid;
951 if ((retval = security_task_alloc(p)))
952 goto bad_fork_cleanup_policy;
953 if ((retval = audit_alloc(p)))
954 goto bad_fork_cleanup_security;
955 /* copy all the process information */
956 if ((retval = copy_semundo(clone_flags, p)))
957 goto bad_fork_cleanup_audit;
958 if ((retval = copy_files(clone_flags, p)))
959 goto bad_fork_cleanup_semundo;
960 if ((retval = copy_fs(clone_flags, p)))
961 goto bad_fork_cleanup_files;
962 if ((retval = copy_sighand(clone_flags, p)))
963 goto bad_fork_cleanup_fs;
964 if ((retval = copy_signal(clone_flags, p)))
965 goto bad_fork_cleanup_sighand;
966 if ((retval = copy_mm(clone_flags, p)))
967 goto bad_fork_cleanup_signal;
968 if ((retval = copy_keys(clone_flags, p)))
969 goto bad_fork_cleanup_mm;
970 if ((retval = copy_namespace(clone_flags, p)))
971 goto bad_fork_cleanup_keys;
972 retval = copy_thread(0, clone_flags, stack_start, stack_size, p, regs);
974 goto bad_fork_cleanup_namespace;
976 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
978 * Clear TID on mm_release()?
980 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL;
983 * Syscall tracing should be turned off in the child regardless
986 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
988 /* Our parent execution domain becomes current domain
989 These must match for thread signalling to apply */
991 p->parent_exec_id = p->self_exec_id;
993 /* ok, now we should be set up.. */
994 p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
995 p->pdeath_signal = 0;
998 /* Perform scheduler related setup */
1002 * Ok, make it visible to the rest of the system.
1003 * We dont wake it up yet.
1005 p->group_leader = p;
1006 INIT_LIST_HEAD(&p->ptrace_children);
1007 INIT_LIST_HEAD(&p->ptrace_list);
1009 /* Need tasklist lock for parent etc handling! */
1010 write_lock_irq(&tasklist_lock);
1013 * The task hasn't been attached yet, so cpus_allowed mask cannot
1014 * have changed. The cpus_allowed mask of the parent may have
1015 * changed after it was copied first time, and it may then move to
1016 * another CPU - so we re-copy it here and set the child's CPU to
1017 * the parent's CPU. This avoids alot of nasty races.
1019 p->cpus_allowed = current->cpus_allowed;
1020 set_task_cpu(p, smp_processor_id());
1023 * Check for pending SIGKILL! The new thread should not be allowed
1024 * to slip out of an OOM kill. (or normal SIGKILL.)
1026 if (sigismember(¤t->pending.signal, SIGKILL)) {
1027 write_unlock_irq(&tasklist_lock);
1029 goto bad_fork_cleanup_namespace;
1032 /* CLONE_PARENT re-uses the old parent */
1033 if (clone_flags & (CLONE_PARENT|CLONE_THREAD))
1034 p->real_parent = current->real_parent;
1036 p->real_parent = current;
1037 p->parent = p->real_parent;
1039 if (clone_flags & CLONE_THREAD) {
1040 spin_lock(¤t->sighand->siglock);
1042 * Important: if an exit-all has been started then
1043 * do not create this new thread - the whole thread
1044 * group is supposed to exit anyway.
1046 if (current->signal->group_exit) {
1047 spin_unlock(¤t->sighand->siglock);
1048 write_unlock_irq(&tasklist_lock);
1050 goto bad_fork_cleanup_namespace;
1052 p->group_leader = current->group_leader;
1054 if (current->signal->group_stop_count > 0) {
1056 * There is an all-stop in progress for the group.
1057 * We ourselves will stop as soon as we check signals.
1058 * Make the new thread part of that group stop too.
1060 current->signal->group_stop_count++;
1061 set_tsk_thread_flag(p, TIF_SIGPENDING);
1064 spin_unlock(¤t->sighand->siglock);
1068 if (unlikely(p->ptrace & PT_PTRACED))
1069 __ptrace_link(p, current->parent);
1071 attach_pid(p, PIDTYPE_PID, p->pid);
1072 attach_pid(p, PIDTYPE_TGID, p->tgid);
1073 if (thread_group_leader(p)) {
1074 attach_pid(p, PIDTYPE_PGID, process_group(p));
1075 attach_pid(p, PIDTYPE_SID, p->signal->session);
1077 __get_cpu_var(process_counts)++;
1080 p->ioprio = current->ioprio;
1082 /* p is copy of current */
1085 atomic_inc(&vxi->cvirt.nr_threads);
1088 write_unlock_irq(&tasklist_lock);
1093 return ERR_PTR(retval);
1096 bad_fork_cleanup_namespace:
1098 bad_fork_cleanup_keys:
1100 bad_fork_cleanup_mm:
1103 bad_fork_cleanup_signal:
1105 bad_fork_cleanup_sighand:
1107 bad_fork_cleanup_fs:
1108 exit_fs(p); /* blocking */
1109 bad_fork_cleanup_files:
1110 exit_files(p); /* blocking */
1111 bad_fork_cleanup_semundo:
1113 bad_fork_cleanup_audit:
1115 bad_fork_cleanup_security:
1116 security_task_free(p);
1117 bad_fork_cleanup_policy:
1119 mpol_free(p->mempolicy);
1123 module_put(p->binfmt->module);
1124 bad_fork_cleanup_put_domain:
1125 module_put(p->thread_info->exec_domain->module);
1126 bad_fork_cleanup_count:
1127 put_group_info(p->group_info);
1128 atomic_dec(&p->user->processes);
1130 bad_fork_cleanup_vm:
1131 if (p->mm && !(clone_flags & CLONE_VM))
1132 vx_pages_sub(p->mm->mm_vx_info, RLIMIT_AS, p->mm->total_vm);
1138 struct pt_regs * __devinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
1140 memset(regs, 0, sizeof(struct pt_regs));
1144 task_t * __devinit fork_idle(int cpu)
1147 struct pt_regs regs;
1149 task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL, NULL, 0);
1151 return ERR_PTR(-ENOMEM);
1152 init_idle(task, cpu);
1153 unhash_process(task);
1157 static inline int fork_traceflag (unsigned clone_flags)
1159 if (clone_flags & CLONE_UNTRACED)
1161 else if (clone_flags & CLONE_VFORK) {
1162 if (current->ptrace & PT_TRACE_VFORK)
1163 return PTRACE_EVENT_VFORK;
1164 } else if ((clone_flags & CSIGNAL) != SIGCHLD) {
1165 if (current->ptrace & PT_TRACE_CLONE)
1166 return PTRACE_EVENT_CLONE;
1167 } else if (current->ptrace & PT_TRACE_FORK)
1168 return PTRACE_EVENT_FORK;
1174 * Ok, this is the main fork-routine.
1176 * It copies the process, and if successful kick-starts
1177 * it and waits for it to finish using the VM if required.
1179 long do_fork(unsigned long clone_flags,
1180 unsigned long stack_start,
1181 struct pt_regs *regs,
1182 unsigned long stack_size,
1183 int __user *parent_tidptr,
1184 int __user *child_tidptr)
1186 struct task_struct *p;
1188 long pid = alloc_pidmap();
1192 if (unlikely(current->ptrace)) {
1193 trace = fork_traceflag (clone_flags);
1195 clone_flags |= CLONE_PTRACE;
1198 #ifdef CONFIG_CKRM_TYPE_TASKCLASS
1199 if (numtasks_get_ref(current->taskclass, 0) == 0) {
1204 p = copy_process(clone_flags, stack_start, regs, stack_size, parent_tidptr, child_tidptr, pid);
1206 * Do this prior waking up the new thread - the thread pointer
1207 * might get invalid after that point, if the thread exits quickly.
1210 struct completion vfork;
1214 if (clone_flags & CLONE_VFORK) {
1215 p->vfork_done = &vfork;
1216 init_completion(&vfork);
1219 if ((p->ptrace & PT_PTRACED) || (clone_flags & CLONE_STOPPED)) {
1221 * We'll start up with an immediate SIGSTOP.
1223 sigaddset(&p->pending.signal, SIGSTOP);
1224 set_tsk_thread_flag(p, TIF_SIGPENDING);
1227 if (!(clone_flags & CLONE_STOPPED))
1228 wake_up_new_task(p, clone_flags);
1230 p->state = TASK_STOPPED;
1233 if (unlikely (trace)) {
1234 current->ptrace_message = pid;
1235 ptrace_notify ((trace << 8) | SIGTRAP);
1238 if (clone_flags & CLONE_VFORK) {
1239 wait_for_completion(&vfork);
1240 if (unlikely (current->ptrace & PT_TRACE_VFORK_DONE))
1241 ptrace_notify ((PTRACE_EVENT_VFORK_DONE << 8) | SIGTRAP);
1244 #ifdef CONFIG_CKRM_TYPE_TASKCLASS
1245 numtasks_put_ref(current->taskclass);
1253 /* SLAB cache for signal_struct structures (tsk->signal) */
1254 kmem_cache_t *signal_cachep;
1256 /* SLAB cache for sighand_struct structures (tsk->sighand) */
1257 kmem_cache_t *sighand_cachep;
1259 /* SLAB cache for files_struct structures (tsk->files) */
1260 kmem_cache_t *files_cachep;
1262 /* SLAB cache for fs_struct structures (tsk->fs) */
1263 kmem_cache_t *fs_cachep;
1265 /* SLAB cache for vm_area_struct structures */
1266 kmem_cache_t *vm_area_cachep;
1268 /* SLAB cache for mm_struct structures (tsk->mm) */
1269 kmem_cache_t *mm_cachep;
1271 void __init proc_caches_init(void)
1273 sighand_cachep = kmem_cache_create("sighand_cache",
1274 sizeof(struct sighand_struct), 0,
1275 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
1276 signal_cachep = kmem_cache_create("signal_cache",
1277 sizeof(struct signal_struct), 0,
1278 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
1279 files_cachep = kmem_cache_create("files_cache",
1280 sizeof(struct files_struct), 0,
1281 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
1282 fs_cachep = kmem_cache_create("fs_cache",
1283 sizeof(struct fs_struct), 0,
1284 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
1285 vm_area_cachep = kmem_cache_create("vm_area_struct",
1286 sizeof(struct vm_area_struct), 0,
1287 SLAB_PANIC, NULL, NULL);
1288 mm_cachep = kmem_cache_create("mm_struct",
1289 sizeof(struct mm_struct), 0,
1290 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);