[PATCH 1/3] utrace_control: fix utrace_reset(safe) usage when ->exit_state != 0

[Oleg Nesterov]

c36a311e8a6c1bbe6115dc1810ac88344782857c breaks utrace_control_dead() logic.

It is important that utrace_reset(safe) is called with "safe = true" when
we detach from the dead task. Otherwise, this engine will not be "unlinked"
until utrace_release_task(), and given that the dead task can stay in zombie
state unpredictably long we have problems: for example, utrace_barrier() will
hang.

Signed-off-by: Oleg Nesterov <o...@redhat.com>
---

 kernel/utrace.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

--- __UTRACE/kernel/utrace.c~1_FIX_DEAD_DETACH  2009-09-01 19:12:14.000000000 
+0200
+++ __UTRACE/kernel/utrace.c    2009-09-01 20:02:26.000000000 +0200
@@ -1037,7 +1037,7 @@ int utrace_control(struct task_struct *t
                   enum utrace_resume_action action)
 {
        struct utrace *utrace;
-       bool reset;
+       bool reset, safe = false;
        int ret;
 
        if (unlikely(action > UTRACE_DETACH))
@@ -1074,7 +1074,7 @@ int utrace_control(struct task_struct *t
                        spin_unlock(&utrace->lock);
                        return ret;
                }
-               reset = true;
+               reset = safe = true;
        }
 
        switch (action) {
@@ -1212,8 +1212,8 @@ int utrace_control(struct task_struct *t
         * there is nothing more we need to do.
         */
        if (reset)
-               utrace_reset(target, utrace,
-                            utrace->stopped || target == current);
+               utrace_reset(target, utrace, safe || utrace->stopped ||
+                                               target == current);
        else
                spin_unlock(&utrace->lock);