diff options
Diffstat (limited to 'kernel/exit.c')
-rw-r--r-- | kernel/exit.c | 258 |
1 files changed, 76 insertions, 182 deletions
diff --git a/kernel/exit.c b/kernel/exit.c index f80dec3f1875..abf9cf3b95c6 100644 --- a/kernel/exit.c +++ b/kernel/exit.c | |||
@@ -46,6 +46,7 @@ | |||
46 | #include <linux/blkdev.h> | 46 | #include <linux/blkdev.h> |
47 | #include <linux/task_io_accounting_ops.h> | 47 | #include <linux/task_io_accounting_ops.h> |
48 | #include <linux/tracehook.h> | 48 | #include <linux/tracehook.h> |
49 | #include <linux/fs_struct.h> | ||
49 | #include <linux/init_task.h> | 50 | #include <linux/init_task.h> |
50 | #include <trace/sched.h> | 51 | #include <trace/sched.h> |
51 | 52 | ||
@@ -61,11 +62,6 @@ DEFINE_TRACE(sched_process_wait); | |||
61 | 62 | ||
62 | static void exit_mm(struct task_struct * tsk); | 63 | static void exit_mm(struct task_struct * tsk); |
63 | 64 | ||
64 | static inline int task_detached(struct task_struct *p) | ||
65 | { | ||
66 | return p->exit_signal == -1; | ||
67 | } | ||
68 | |||
69 | static void __unhash_process(struct task_struct *p) | 65 | static void __unhash_process(struct task_struct *p) |
70 | { | 66 | { |
71 | nr_threads--; | 67 | nr_threads--; |
@@ -118,6 +114,8 @@ static void __exit_signal(struct task_struct *tsk) | |||
118 | * We won't ever get here for the group leader, since it | 114 | * We won't ever get here for the group leader, since it |
119 | * will have been the last reference on the signal_struct. | 115 | * will have been the last reference on the signal_struct. |
120 | */ | 116 | */ |
117 | sig->utime = cputime_add(sig->utime, task_utime(tsk)); | ||
118 | sig->stime = cputime_add(sig->stime, task_stime(tsk)); | ||
121 | sig->gtime = cputime_add(sig->gtime, task_gtime(tsk)); | 119 | sig->gtime = cputime_add(sig->gtime, task_gtime(tsk)); |
122 | sig->min_flt += tsk->min_flt; | 120 | sig->min_flt += tsk->min_flt; |
123 | sig->maj_flt += tsk->maj_flt; | 121 | sig->maj_flt += tsk->maj_flt; |
@@ -126,6 +124,7 @@ static void __exit_signal(struct task_struct *tsk) | |||
126 | sig->inblock += task_io_get_inblock(tsk); | 124 | sig->inblock += task_io_get_inblock(tsk); |
127 | sig->oublock += task_io_get_oublock(tsk); | 125 | sig->oublock += task_io_get_oublock(tsk); |
128 | task_io_accounting_add(&sig->ioac, &tsk->ioac); | 126 | task_io_accounting_add(&sig->ioac, &tsk->ioac); |
127 | sig->sum_sched_runtime += tsk->se.sum_exec_runtime; | ||
129 | sig = NULL; /* Marker for below. */ | 128 | sig = NULL; /* Marker for below. */ |
130 | } | 129 | } |
131 | 130 | ||
@@ -359,16 +358,12 @@ static void reparent_to_kthreadd(void) | |||
359 | void __set_special_pids(struct pid *pid) | 358 | void __set_special_pids(struct pid *pid) |
360 | { | 359 | { |
361 | struct task_struct *curr = current->group_leader; | 360 | struct task_struct *curr = current->group_leader; |
362 | pid_t nr = pid_nr(pid); | ||
363 | 361 | ||
364 | if (task_session(curr) != pid) { | 362 | if (task_session(curr) != pid) |
365 | change_pid(curr, PIDTYPE_SID, pid); | 363 | change_pid(curr, PIDTYPE_SID, pid); |
366 | set_task_session(curr, nr); | 364 | |
367 | } | 365 | if (task_pgrp(curr) != pid) |
368 | if (task_pgrp(curr) != pid) { | ||
369 | change_pid(curr, PIDTYPE_PGID, pid); | 366 | change_pid(curr, PIDTYPE_PGID, pid); |
370 | set_task_pgrp(curr, nr); | ||
371 | } | ||
372 | } | 367 | } |
373 | 368 | ||
374 | static void set_special_pids(struct pid *pid) | 369 | static void set_special_pids(struct pid *pid) |
@@ -426,7 +421,6 @@ EXPORT_SYMBOL(disallow_signal); | |||
426 | void daemonize(const char *name, ...) | 421 | void daemonize(const char *name, ...) |
427 | { | 422 | { |
428 | va_list args; | 423 | va_list args; |
429 | struct fs_struct *fs; | ||
430 | sigset_t blocked; | 424 | sigset_t blocked; |
431 | 425 | ||
432 | va_start(args, name); | 426 | va_start(args, name); |
@@ -459,11 +453,7 @@ void daemonize(const char *name, ...) | |||
459 | 453 | ||
460 | /* Become as one with the init task */ | 454 | /* Become as one with the init task */ |
461 | 455 | ||
462 | exit_fs(current); /* current->fs->count--; */ | 456 | daemonize_fs_struct(); |
463 | fs = init_task.fs; | ||
464 | current->fs = fs; | ||
465 | atomic_inc(&fs->count); | ||
466 | |||
467 | exit_files(current); | 457 | exit_files(current); |
468 | current->files = init_task.files; | 458 | current->files = init_task.files; |
469 | atomic_inc(¤t->files->count); | 459 | atomic_inc(¤t->files->count); |
@@ -562,30 +552,6 @@ void exit_files(struct task_struct *tsk) | |||
562 | } | 552 | } |
563 | } | 553 | } |
564 | 554 | ||
565 | void put_fs_struct(struct fs_struct *fs) | ||
566 | { | ||
567 | /* No need to hold fs->lock if we are killing it */ | ||
568 | if (atomic_dec_and_test(&fs->count)) { | ||
569 | path_put(&fs->root); | ||
570 | path_put(&fs->pwd); | ||
571 | kmem_cache_free(fs_cachep, fs); | ||
572 | } | ||
573 | } | ||
574 | |||
575 | void exit_fs(struct task_struct *tsk) | ||
576 | { | ||
577 | struct fs_struct * fs = tsk->fs; | ||
578 | |||
579 | if (fs) { | ||
580 | task_lock(tsk); | ||
581 | tsk->fs = NULL; | ||
582 | task_unlock(tsk); | ||
583 | put_fs_struct(fs); | ||
584 | } | ||
585 | } | ||
586 | |||
587 | EXPORT_SYMBOL_GPL(exit_fs); | ||
588 | |||
589 | #ifdef CONFIG_MM_OWNER | 555 | #ifdef CONFIG_MM_OWNER |
590 | /* | 556 | /* |
591 | * Task p is exiting and it owned mm, lets find a new owner for it | 557 | * Task p is exiting and it owned mm, lets find a new owner for it |
@@ -729,119 +695,6 @@ static void exit_mm(struct task_struct * tsk) | |||
729 | } | 695 | } |
730 | 696 | ||
731 | /* | 697 | /* |
732 | * Return nonzero if @parent's children should reap themselves. | ||
733 | * | ||
734 | * Called with write_lock_irq(&tasklist_lock) held. | ||
735 | */ | ||
736 | static int ignoring_children(struct task_struct *parent) | ||
737 | { | ||
738 | int ret; | ||
739 | struct sighand_struct *psig = parent->sighand; | ||
740 | unsigned long flags; | ||
741 | spin_lock_irqsave(&psig->siglock, flags); | ||
742 | ret = (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN || | ||
743 | (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT)); | ||
744 | spin_unlock_irqrestore(&psig->siglock, flags); | ||
745 | return ret; | ||
746 | } | ||
747 | |||
748 | /* | ||
749 | * Detach all tasks we were using ptrace on. | ||
750 | * Any that need to be release_task'd are put on the @dead list. | ||
751 | * | ||
752 | * Called with write_lock(&tasklist_lock) held. | ||
753 | */ | ||
754 | static void ptrace_exit(struct task_struct *parent, struct list_head *dead) | ||
755 | { | ||
756 | struct task_struct *p, *n; | ||
757 | int ign = -1; | ||
758 | |||
759 | list_for_each_entry_safe(p, n, &parent->ptraced, ptrace_entry) { | ||
760 | __ptrace_unlink(p); | ||
761 | |||
762 | if (p->exit_state != EXIT_ZOMBIE) | ||
763 | continue; | ||
764 | |||
765 | /* | ||
766 | * If it's a zombie, our attachedness prevented normal | ||
767 | * parent notification or self-reaping. Do notification | ||
768 | * now if it would have happened earlier. If it should | ||
769 | * reap itself, add it to the @dead list. We can't call | ||
770 | * release_task() here because we already hold tasklist_lock. | ||
771 | * | ||
772 | * If it's our own child, there is no notification to do. | ||
773 | * But if our normal children self-reap, then this child | ||
774 | * was prevented by ptrace and we must reap it now. | ||
775 | */ | ||
776 | if (!task_detached(p) && thread_group_empty(p)) { | ||
777 | if (!same_thread_group(p->real_parent, parent)) | ||
778 | do_notify_parent(p, p->exit_signal); | ||
779 | else { | ||
780 | if (ign < 0) | ||
781 | ign = ignoring_children(parent); | ||
782 | if (ign) | ||
783 | p->exit_signal = -1; | ||
784 | } | ||
785 | } | ||
786 | |||
787 | if (task_detached(p)) { | ||
788 | /* | ||
789 | * Mark it as in the process of being reaped. | ||
790 | */ | ||
791 | p->exit_state = EXIT_DEAD; | ||
792 | list_add(&p->ptrace_entry, dead); | ||
793 | } | ||
794 | } | ||
795 | } | ||
796 | |||
797 | /* | ||
798 | * Finish up exit-time ptrace cleanup. | ||
799 | * | ||
800 | * Called without locks. | ||
801 | */ | ||
802 | static void ptrace_exit_finish(struct task_struct *parent, | ||
803 | struct list_head *dead) | ||
804 | { | ||
805 | struct task_struct *p, *n; | ||
806 | |||
807 | BUG_ON(!list_empty(&parent->ptraced)); | ||
808 | |||
809 | list_for_each_entry_safe(p, n, dead, ptrace_entry) { | ||
810 | list_del_init(&p->ptrace_entry); | ||
811 | release_task(p); | ||
812 | } | ||
813 | } | ||
814 | |||
815 | static void reparent_thread(struct task_struct *p, struct task_struct *father) | ||
816 | { | ||
817 | if (p->pdeath_signal) | ||
818 | /* We already hold the tasklist_lock here. */ | ||
819 | group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p); | ||
820 | |||
821 | list_move_tail(&p->sibling, &p->real_parent->children); | ||
822 | |||
823 | /* If this is a threaded reparent there is no need to | ||
824 | * notify anyone anything has happened. | ||
825 | */ | ||
826 | if (same_thread_group(p->real_parent, father)) | ||
827 | return; | ||
828 | |||
829 | /* We don't want people slaying init. */ | ||
830 | if (!task_detached(p)) | ||
831 | p->exit_signal = SIGCHLD; | ||
832 | |||
833 | /* If we'd notified the old parent about this child's death, | ||
834 | * also notify the new parent. | ||
835 | */ | ||
836 | if (!ptrace_reparented(p) && | ||
837 | p->exit_state == EXIT_ZOMBIE && | ||
838 | !task_detached(p) && thread_group_empty(p)) | ||
839 | do_notify_parent(p, p->exit_signal); | ||
840 | |||
841 | kill_orphaned_pgrp(p, father); | ||
842 | } | ||
843 | |||
844 | /* | ||
845 | * When we die, we re-parent all our children. | 698 | * When we die, we re-parent all our children. |
846 | * Try to give them to another thread in our thread | 699 | * Try to give them to another thread in our thread |
847 | * group, and if no such member exists, give it to | 700 | * group, and if no such member exists, give it to |
@@ -880,17 +733,51 @@ static struct task_struct *find_new_reaper(struct task_struct *father) | |||
880 | return pid_ns->child_reaper; | 733 | return pid_ns->child_reaper; |
881 | } | 734 | } |
882 | 735 | ||
736 | /* | ||
737 | * Any that need to be release_task'd are put on the @dead list. | ||
738 | */ | ||
739 | static void reparent_thread(struct task_struct *father, struct task_struct *p, | ||
740 | struct list_head *dead) | ||
741 | { | ||
742 | if (p->pdeath_signal) | ||
743 | group_send_sig_info(p->pdeath_signal, SEND_SIG_NOINFO, p); | ||
744 | |||
745 | list_move_tail(&p->sibling, &p->real_parent->children); | ||
746 | |||
747 | if (task_detached(p)) | ||
748 | return; | ||
749 | /* | ||
750 | * If this is a threaded reparent there is no need to | ||
751 | * notify anyone anything has happened. | ||
752 | */ | ||
753 | if (same_thread_group(p->real_parent, father)) | ||
754 | return; | ||
755 | |||
756 | /* We don't want people slaying init. */ | ||
757 | p->exit_signal = SIGCHLD; | ||
758 | |||
759 | /* If it has exited notify the new parent about this child's death. */ | ||
760 | if (!p->ptrace && | ||
761 | p->exit_state == EXIT_ZOMBIE && thread_group_empty(p)) { | ||
762 | do_notify_parent(p, p->exit_signal); | ||
763 | if (task_detached(p)) { | ||
764 | p->exit_state = EXIT_DEAD; | ||
765 | list_move_tail(&p->sibling, dead); | ||
766 | } | ||
767 | } | ||
768 | |||
769 | kill_orphaned_pgrp(p, father); | ||
770 | } | ||
771 | |||
883 | static void forget_original_parent(struct task_struct *father) | 772 | static void forget_original_parent(struct task_struct *father) |
884 | { | 773 | { |
885 | struct task_struct *p, *n, *reaper; | 774 | struct task_struct *p, *n, *reaper; |
886 | LIST_HEAD(ptrace_dead); | 775 | LIST_HEAD(dead_children); |
776 | |||
777 | exit_ptrace(father); | ||
887 | 778 | ||
888 | write_lock_irq(&tasklist_lock); | 779 | write_lock_irq(&tasklist_lock); |
889 | reaper = find_new_reaper(father); | 780 | reaper = find_new_reaper(father); |
890 | /* | ||
891 | * First clean up ptrace if we were using it. | ||
892 | */ | ||
893 | ptrace_exit(father, &ptrace_dead); | ||
894 | 781 | ||
895 | list_for_each_entry_safe(p, n, &father->children, sibling) { | 782 | list_for_each_entry_safe(p, n, &father->children, sibling) { |
896 | p->real_parent = reaper; | 783 | p->real_parent = reaper; |
@@ -898,13 +785,16 @@ static void forget_original_parent(struct task_struct *father) | |||
898 | BUG_ON(p->ptrace); | 785 | BUG_ON(p->ptrace); |
899 | p->parent = p->real_parent; | 786 | p->parent = p->real_parent; |
900 | } | 787 | } |
901 | reparent_thread(p, father); | 788 | reparent_thread(father, p, &dead_children); |
902 | } | 789 | } |
903 | |||
904 | write_unlock_irq(&tasklist_lock); | 790 | write_unlock_irq(&tasklist_lock); |
791 | |||
905 | BUG_ON(!list_empty(&father->children)); | 792 | BUG_ON(!list_empty(&father->children)); |
906 | 793 | ||
907 | ptrace_exit_finish(father, &ptrace_dead); | 794 | list_for_each_entry_safe(p, n, &dead_children, sibling) { |
795 | list_del_init(&p->sibling); | ||
796 | release_task(p); | ||
797 | } | ||
908 | } | 798 | } |
909 | 799 | ||
910 | /* | 800 | /* |
@@ -947,8 +837,7 @@ static void exit_notify(struct task_struct *tsk, int group_dead) | |||
947 | */ | 837 | */ |
948 | if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) && | 838 | if (tsk->exit_signal != SIGCHLD && !task_detached(tsk) && |
949 | (tsk->parent_exec_id != tsk->real_parent->self_exec_id || | 839 | (tsk->parent_exec_id != tsk->real_parent->self_exec_id || |
950 | tsk->self_exec_id != tsk->parent_exec_id) && | 840 | tsk->self_exec_id != tsk->parent_exec_id)) |
951 | !capable(CAP_KILL)) | ||
952 | tsk->exit_signal = SIGCHLD; | 841 | tsk->exit_signal = SIGCHLD; |
953 | 842 | ||
954 | signal = tracehook_notify_death(tsk, &cookie, group_dead); | 843 | signal = tracehook_notify_death(tsk, &cookie, group_dead); |
@@ -977,12 +866,9 @@ static void check_stack_usage(void) | |||
977 | { | 866 | { |
978 | static DEFINE_SPINLOCK(low_water_lock); | 867 | static DEFINE_SPINLOCK(low_water_lock); |
979 | static int lowest_to_date = THREAD_SIZE; | 868 | static int lowest_to_date = THREAD_SIZE; |
980 | unsigned long *n = end_of_stack(current); | ||
981 | unsigned long free; | 869 | unsigned long free; |
982 | 870 | ||
983 | while (*n == 0) | 871 | free = stack_not_used(current); |
984 | n++; | ||
985 | free = (unsigned long)n - (unsigned long)end_of_stack(current); | ||
986 | 872 | ||
987 | if (free >= lowest_to_date) | 873 | if (free >= lowest_to_date) |
988 | return; | 874 | return; |
@@ -1037,6 +923,8 @@ NORET_TYPE void do_exit(long code) | |||
1037 | schedule(); | 923 | schedule(); |
1038 | } | 924 | } |
1039 | 925 | ||
926 | exit_irq_thread(); | ||
927 | |||
1040 | exit_signals(tsk); /* sets PF_EXITING */ | 928 | exit_signals(tsk); /* sets PF_EXITING */ |
1041 | /* | 929 | /* |
1042 | * tsk->flags are checked in the futex code to protect against | 930 | * tsk->flags are checked in the futex code to protect against |
@@ -1417,6 +1305,18 @@ static int wait_task_zombie(struct task_struct *p, int options, | |||
1417 | return retval; | 1305 | return retval; |
1418 | } | 1306 | } |
1419 | 1307 | ||
1308 | static int *task_stopped_code(struct task_struct *p, bool ptrace) | ||
1309 | { | ||
1310 | if (ptrace) { | ||
1311 | if (task_is_stopped_or_traced(p)) | ||
1312 | return &p->exit_code; | ||
1313 | } else { | ||
1314 | if (p->signal->flags & SIGNAL_STOP_STOPPED) | ||
1315 | return &p->signal->group_exit_code; | ||
1316 | } | ||
1317 | return NULL; | ||
1318 | } | ||
1319 | |||
1420 | /* | 1320 | /* |
1421 | * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold | 1321 | * Handle sys_wait4 work for one task in state TASK_STOPPED. We hold |
1422 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold | 1322 | * read_lock(&tasklist_lock) on entry. If we return zero, we still hold |
@@ -1427,7 +1327,7 @@ static int wait_task_stopped(int ptrace, struct task_struct *p, | |||
1427 | int options, struct siginfo __user *infop, | 1327 | int options, struct siginfo __user *infop, |
1428 | int __user *stat_addr, struct rusage __user *ru) | 1328 | int __user *stat_addr, struct rusage __user *ru) |
1429 | { | 1329 | { |
1430 | int retval, exit_code, why; | 1330 | int retval, exit_code, *p_code, why; |
1431 | uid_t uid = 0; /* unneeded, required by compiler */ | 1331 | uid_t uid = 0; /* unneeded, required by compiler */ |
1432 | pid_t pid; | 1332 | pid_t pid; |
1433 | 1333 | ||
@@ -1437,22 +1337,16 @@ static int wait_task_stopped(int ptrace, struct task_struct *p, | |||
1437 | exit_code = 0; | 1337 | exit_code = 0; |
1438 | spin_lock_irq(&p->sighand->siglock); | 1338 | spin_lock_irq(&p->sighand->siglock); |
1439 | 1339 | ||
1440 | if (unlikely(!task_is_stopped_or_traced(p))) | 1340 | p_code = task_stopped_code(p, ptrace); |
1441 | goto unlock_sig; | 1341 | if (unlikely(!p_code)) |
1442 | |||
1443 | if (!ptrace && p->signal->group_stop_count > 0) | ||
1444 | /* | ||
1445 | * A group stop is in progress and this is the group leader. | ||
1446 | * We won't report until all threads have stopped. | ||
1447 | */ | ||
1448 | goto unlock_sig; | 1342 | goto unlock_sig; |
1449 | 1343 | ||
1450 | exit_code = p->exit_code; | 1344 | exit_code = *p_code; |
1451 | if (!exit_code) | 1345 | if (!exit_code) |
1452 | goto unlock_sig; | 1346 | goto unlock_sig; |
1453 | 1347 | ||
1454 | if (!unlikely(options & WNOWAIT)) | 1348 | if (!unlikely(options & WNOWAIT)) |
1455 | p->exit_code = 0; | 1349 | *p_code = 0; |
1456 | 1350 | ||
1457 | /* don't need the RCU readlock here as we're holding a spinlock */ | 1351 | /* don't need the RCU readlock here as we're holding a spinlock */ |
1458 | uid = __task_cred(p)->uid; | 1352 | uid = __task_cred(p)->uid; |
@@ -1608,7 +1502,7 @@ static int wait_consider_task(struct task_struct *parent, int ptrace, | |||
1608 | */ | 1502 | */ |
1609 | *notask_error = 0; | 1503 | *notask_error = 0; |
1610 | 1504 | ||
1611 | if (task_is_stopped_or_traced(p)) | 1505 | if (task_stopped_code(p, ptrace)) |
1612 | return wait_task_stopped(ptrace, p, options, | 1506 | return wait_task_stopped(ptrace, p, options, |
1613 | infop, stat_addr, ru); | 1507 | infop, stat_addr, ru); |
1614 | 1508 | ||
@@ -1812,7 +1706,7 @@ SYSCALL_DEFINE4(wait4, pid_t, upid, int __user *, stat_addr, | |||
1812 | pid = find_get_pid(-upid); | 1706 | pid = find_get_pid(-upid); |
1813 | } else if (upid == 0) { | 1707 | } else if (upid == 0) { |
1814 | type = PIDTYPE_PGID; | 1708 | type = PIDTYPE_PGID; |
1815 | pid = get_pid(task_pgrp(current)); | 1709 | pid = get_task_pid(current, PIDTYPE_PGID); |
1816 | } else /* upid > 0 */ { | 1710 | } else /* upid > 0 */ { |
1817 | type = PIDTYPE_PID; | 1711 | type = PIDTYPE_PID; |
1818 | pid = find_get_pid(upid); | 1712 | pid = find_get_pid(upid); |