Re: [PATCH] fix TASK_STOPPED vs TASK_NONINTERACTIVE interaction

[Roland McGrath <roland@redhat.com>]

I am dubious about this change.  I don't see a corresponding change to
fs/proc/array.c where it knows what all the bit values are.  I am not at
all sure there aren't other places that know these values and need a fixup
if you change them.

Any tests using < TASK_STOPPED or the like are left over from the time when
the TASK_ZOMBIE and TASK_DEAD bits were in the same word, and it served to
check for "stopped or dead".  I think this one in do_signal_stop is the
only such case.  It has been buggy ever since exit_state was separated.
Changing the bit values doesn't fix the bug that it isn't checking the
exit_state value.  This patch is probably the right fix for that, but I
have not tested it.

Signed-off-by: Roland McGrath <roland@redhat.com>

--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -1775,7 +1775,8 @@ do_signal_stop(int signr)
 				 * stop is always done with the siglock held,
 				 * so this check has no races.
 				 */
-				if (t->state < TASK_STOPPED) {
+				if (!t->exit_state &&
+				    !(t->state & (TASK_STOPPED|TASK_TRACED))) {
 					stop_count++;
 					signal_wake_up(t, 0);
 				}