Re: 5.8-rc*: kernel BUG at kernel/signal.c:1917

[Oleg Nesterov <oleg@redhat.com>]

On 07/17, Jiri Slaby wrote:
>
> On 17. 07. 20, 12:45, Jiri Slaby wrote:
> > Hi,
> >
> > the strace testsuite triggers this on 5.8-rc4 and -rc5 both on x86_64
> > and i586:
>
> make check needs -jsomething, running is sequentially (-j1) doesn't
> trigger it. After the error, I cannot run anything. Like ps to find out
> what test caused the crash...

Strange... I'll try to reproduce but I can't do this till Monday.

Meanwhile, could you try the patch below? It needs CONFIG_DEBUG_ATOMIC_SLEEP.

Oleg.

diff --git a/kernel/signal.c b/kernel/signal.c
index ee22ec78fd6d..84782843eb1c 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -1914,7 +1914,9 @@ bool do_notify_parent(struct task_struct *tsk, int sig)
 	BUG_ON(sig == -1);
 
  	/* do_notify_parent_cldstop should have been called instead.  */
- 	BUG_ON(task_is_stopped_or_traced(tsk));
+ 	WARN(task_is_stopped_or_traced(tsk),
+ 		"exist with state=%ld set at %pS",
+ 		tsk->state, (void*)tsk->task_state_change);
 
 	BUG_ON(!tsk->ptrace &&
 	       (tsk->group_leader != tsk || !thread_group_empty(tsk)));
@@ -2214,6 +2216,7 @@ static void ptrace_stop(int exit_code, int why, int clear_code, kernel_siginfo_t
 		preempt_enable_no_resched();
 		freezable_schedule();
 		cgroup_leave_frozen(true);
+		WARN_ON(current->state);
 	} else {
 		/*
 		 * By the time we got the lock, our tracer went away.
@@ -2393,6 +2396,7 @@ static bool do_signal_stop(int signr)
 		/* Now we don't run again until woken by SIGCONT or SIGKILL */
 		cgroup_enter_frozen();
 		freezable_schedule();
+		WARN_ON(current->state);
 		return true;
 	} else {
 		/*