ptrace_init_task: initialize child->jobctl explicitly
authorOleg Nesterov <oleg@redhat.com>
Fri, 8 Jul 2011 17:13:39 +0000 (19:13 +0200)
committerOleg Nesterov <oleg@redhat.com>
Sun, 17 Jul 2011 18:23:51 +0000 (20:23 +0200)
new_child->jobctl is not initialized during the fork, it is copied
from parent->jobctl. Currently this is harmless, the forking task
is running and copy_process() can't succeed if signal_pending() is
true, so only JOBCTL_STOP_DEQUEUED can be copied. Still this is a
bit fragile, it would be more clean to set ->jobctl = 0 explicitly.

Also, check ->ptrace != 0 instead of PT_PTRACED, move the
CONFIG_HAVE_HW_BREAKPOINT code up.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
include/linux/ptrace.h

index eae381d..fd8669f 100644 (file)
@@ -217,16 +217,17 @@ static inline void ptrace_init_task(struct task_struct *child, bool ptrace)
 {
        INIT_LIST_HEAD(&child->ptrace_entry);
        INIT_LIST_HEAD(&child->ptraced);
-       child->parent = child->real_parent;
+#ifdef CONFIG_HAVE_HW_BREAKPOINT
+       atomic_set(&child->ptrace_bp_refcnt, 1);
+#endif
+       child->jobctl = 0;
        child->ptrace = 0;
-       if (unlikely(ptrace) && (current->ptrace & PT_PTRACED)) {
+       child->parent = child->real_parent;
+
+       if (unlikely(ptrace) && current->ptrace) {
                child->ptrace = current->ptrace;
                __ptrace_link(child, current->parent);
        }
-
-#ifdef CONFIG_HAVE_HW_BREAKPOINT
-       atomic_set(&child->ptrace_bp_refcnt, 1);
-#endif
 }
 
 /**