[Devel] Re: [PATCH 5/7][v8] zap_pid_ns_process() should use force_sig()

[Sukadev Bhattiprolu]

Oleg Nesterov [oleg at redhat.com] wrote:
| On 02/18, Sukadev Bhattiprolu wrote:
| >
| >  	read_lock(&tasklist_lock);
| >  	nr = next_pidmap(pid_ns, 1);
| >  	while (nr > 0) {
| > -		kill_proc_info(SIGKILL, SEND_SIG_PRIV, nr);
| > +		rcu_read_lock();
| > +
| > +		/*
| > +		 * Use force_sig() since it clears SIGNAL_UNKILLABLE ensuring
| > +		 * any nested-container's init processes don't ignore the
| > +		 * signal
| > +		 */
| > +		task = pid_task(find_vpid(nr), PIDTYPE_PID);
| > +		force_sig(SIGKILL, task);
| 
| Shouldn't we check task != NULL ?

Yes. Here is the updated patch.
---

From: Sukadev Bhattiprolu <sukadev at linux.vnet.ibm.com>
Date: Wed, 18 Feb 2009 15:12:30 -0800
Subject: [PATCH 5/7][v8] zap_pid_ns_process() should use force_sig()

send_signal() assumes that signals with SEND_SIG_PRIV are generated from
within the same namespace. So any nested container-init processes become
immune to the SIGKILL generated by kill_proc_info() in zap_pid_ns_processes().

Use force_sig() in zap_pid_ns_processes() instead - force_sig() clears the
SIGNAL_UNKILLABLE flag ensuring the signal is processed by container-inits.

Signed-off-by: Sukadev Bhattiprolu <sukadev at linux.vnet.ibm.com>
---
 kernel/pid_namespace.c |   15 ++++++++++++++-
 1 files changed, 14 insertions(+), 1 deletions(-)

diff --git a/kernel/pid_namespace.c b/kernel/pid_namespace.c
index fab8ea8..2d1001b 100644
--- a/kernel/pid_namespace.c
+++ b/kernel/pid_namespace.c
@@ -152,6 +152,7 @@ void zap_pid_ns_processes(struct pid_namespace *pid_ns)
 {
 	int nr;
 	int rc;
+	struct task_struct *task;
 
 	/*
 	 * The last thread in the cgroup-init thread group is terminating.
@@ -169,7 +170,19 @@ void zap_pid_ns_processes(struct pid_namespace *pid_ns)
 	read_lock(&tasklist_lock);
 	nr = next_pidmap(pid_ns, 1);
 	while (nr > 0) {
-		kill_proc_info(SIGKILL, SEND_SIG_PRIV, nr);
+		rcu_read_lock();
+
+		/*
+		 * Use force_sig() since it clears SIGNAL_UNKILLABLE ensuring
+		 * any nested-container's init processes don't ignore the
+		 * signal
+		 */
+		task = pid_task(find_vpid(nr), PIDTYPE_PID);
+		if (task)
+			force_sig(SIGKILL, task);
+
+		rcu_read_unlock();
+
 		nr = next_pidmap(pid_ns, nr);
 	}
 	read_unlock(&tasklist_lock);
-- 
1.5.2.5

_______________________________________________
Containers mailing list
Containers at lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers