exit: Coredumps reach do_group_exit
authorEric W. Biederman <ebiederm@xmission.com>
Thu, 23 Dec 2021 16:05:19 +0000 (10:05 -0600)
committerEric W. Biederman <ebiederm@xmission.com>
Sat, 8 Jan 2022 18:43:57 +0000 (12:43 -0600)
The comment about coredumps not reaching do_group_exit and the
corresponding BUG_ON are bogus.

What happens and has happened for years is that get_signal calls
do_coredump (which sets SIGNAL_GROUP_EXIT and group_exit_code) and
then do_group_exit passing the signal number.  Then do_group_exit
ignores the exit_code it is passed and uses signal->group_exit_code
from the coredump.

The comment and BUG_ON were correct when they were added during the
2.5 development cycle, but became obsolete and incorrect when
get_signal was changed to fall through to do_group_exit after
do_coredump in 2.6.10-rc2.

So remove the stale comment and BUG_ON

Fixes: 63bd6144f191 ("[PATCH] Invalid BUG_ONs in signal.c")
History-Tree: https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git
Link: https://lkml.kernel.org/r/20220103213312.9144-2-ebiederm@xmission.com
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
kernel/exit.c

index 64e907b..db86307 100644 (file)
@@ -911,8 +911,6 @@ do_group_exit(int exit_code)
 {
        struct signal_struct *sig = current->signal;
 
-       BUG_ON(exit_code & 0x80); /* core dumps don't get here */
-
        if (sig->flags & SIGNAL_GROUP_EXIT)
                exit_code = sig->group_exit_code;
        else if (sig->group_exec_task)