Re: Zombie stuck in zap_pid_ns_processes()

[Eric W. Biederman]

Oleg Nesterov <oleg@xxxxxxxxxx> writes:

> On 03/24, Eric W. Biederman wrote:
>>
>> --- a/kernel/pid_namespace.c
>> +++ b/kernel/pid_namespace.c
>> @@ -181,6 +181,7 @@ void zap_pid_ns_processes(struct pid_namespace *pid_ns)
>>  	int nr;
>>  	int rc;
>>  	struct task_struct *task, *me = current;
>> +	int init_pids = task_pid_vnr(me) == 1 ? 1 : 2;
>                         ^^^^^^^^^^^^^^^^^^^^^
>
> Ah, mt-init.
>
> perhaps thread_group_leader(me) would be more clean/simple,
> but I won't insist of course.

It is.  Thanks, I was going to follow the reaper but that changes when the
initial thread exits...

Here is my final version of the patch, that I will push shortly.

Eric

From: "Eric W. Biederman" <ebiederm@xxxxxxxxxxxx>
Date: Tue, 26 Mar 2013 02:27:11 -0700
Subject: [PATCH] pid: Handle the exit of a multi-threaded init.

When a multi-threaded init exits and the initial thread is not the
last thread to exit the initial thread hangs around as a zombie
until the last thread exits.  In that case zap_pid_ns_processes
needs to wait until there are only 2 hashed pids in the pid
namespace not one.

v2. Replace thread_pid_vnr(me) == 1 with the test thread_group_leader(me)
    as suggested by Oleg.

Cc: stable@xxxxxxxxxxxxxxx
Cc: Oleg Nesterov <oleg@xxxxxxxxxx>
Reported-by: Caj Larsson <caj@xxxxxxxxxxxxx>
Signed-off-by: "Eric W. Biederman" <ebiederm@xxxxxxxxxxxx>
---
 kernel/pid_namespace.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/kernel/pid_namespace.c b/kernel/pid_namespace.c
index c1c3dc1..bea15bd 100644
--- a/kernel/pid_namespace.c
+++ b/kernel/pid_namespace.c
@@ -181,6 +181,7 @@ void zap_pid_ns_processes(struct pid_namespace *pid_ns)
 	int nr;
 	int rc;
 	struct task_struct *task, *me = current;
+	int init_pids = thread_group_leader(me) ? 1 : 2;
 
 	/* Don't allow any more processes into the pid namespace */
 	disable_pid_allocation(pid_ns);
@@ -230,7 +231,7 @@ void zap_pid_ns_processes(struct pid_namespace *pid_ns)
 	 */
 	for (;;) {
 		set_current_state(TASK_UNINTERRUPTIBLE);
-		if (pid_ns->nr_hashed == 1)
+		if (pid_ns->nr_hashed == init_pids)
 			break;
 		schedule();
 	}
-- 
1.7.5.4


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/