From: Kirill Korotaev <dev@sw.ru>
Note, that during exit process there can be a thread in the system with
tsk->sighand == NULL, since the following call trace:
release_task()
{
....
__exit_sighand() <<< makes tsk->sighand == NULL;
__unhash_process() <<< unhashes thread
....
}
next, we see that next_thread checks for tsk->sighand != NULL:
task_t fastcall *next_thread(const task_t *p)
{
#ifdef CONFIG_SMP
if (!p->sighand)
BUG(); <<< BUG happened here!!!
if (!spin_is_locked(&p->sighand->siglock) &&
!rwlock_is_locked(&tasklist_lock))
....
}
So the question is why next_thread() should check for (p->sighand != NULL)
&& spin_is_locked(&p->sighand->siglock)?
I think these checks are invalid. For example do_task_stat() (which called
next_thread() in this BUG) checks for tsk->sighand != NULL explicitly.
And moreover, next_thread() DOES always works correctly, whether there are
threads or none.
This patch removes sighand checks from the next_thread(), since they are
incorrect and has nothing to do with the next_thread() function. So they
could trigger BUG() when there were no actually bug at all.
Signed-Off-By: Kirill Korotaev <dev@sw.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---
25-akpm/kernel/exit.c | 5 +----
1 files changed, 1 insertion(+), 4 deletions(-)
diff -puN kernel/exit.c~next_thread-bug-fixes kernel/exit.c
--- 25/kernel/exit.c~next_thread-bug-fixes 2004-09-13 01:01:22.227228320 -0700
+++ 25-akpm/kernel/exit.c 2004-09-13 01:01:22.231227712 -0700
@@ -857,10 +857,7 @@ asmlinkage long sys_exit(int error_code)
task_t fastcall *next_thread(const task_t *p)
{
#ifdef CONFIG_SMP
- if (!p->sighand)
- BUG();
- if (!spin_is_locked(&p->sighand->siglock) &&
- !rwlock_is_locked(&tasklist_lock))
+ if (!rwlock_is_locked(&tasklist_lock))
BUG();
#endif
return pid_task(p->pids[PIDTYPE_TGID].pid_list.next, PIDTYPE_TGID);
_