Re: unregister_netdevice: waiting for DEV to become free (2)

[David Ahern]

On 4/27/19 10:22 PM, Tetsuo Handa wrote:
> On 2019/04/28 8:52, Eric Dumazet wrote:
>> On 4/27/19 3:33 PM, Tetsuo Handa wrote:
>>>
>>> I'm waiting for davem why it is safe to move the dst entry from
>>> "a device to unregister" to "a loopback device in that namespace".
>>> I'm waiting for an explanation how the dst entry which was moved to
>>> "a loopback device in that namespace" is released (i.e. what the
>>> expected shutdown sequence is).
>>
>> The most probable explanation is that we make sure the loopback device
>> is the last one to be dismantled at netns deletion,
>> and this would obviously happen after all dst have been released.
>>
> 
> rt_flush_dev() becomes a no-op if "dev" == "a loopback device in that
> namespace". And according to debug printk(), rt_flush_dev() is called
> on "a loopback device in that namespace" itself.
> 
> If "a loopback device in that namespace" is the last "one" (== "a network
> device in that namespace" ?), which shutdown sequence should have called
> dev_put("a loopback device in that namespace") before unregistration of
> "a loopback device in that namespace" starts?
> 
> Since I'm not a netdev person, I appreciate if you can explain
> that shutdown sequence using a flow chart.
> 

The attached patch adds a tracepoint to notifier_call_chain. If you have
KALLSYMS enabled it will show the order of the function handlers:

perf record -e notifier:* -a -g &

ip netns del <NAME>
<wait a few seconds>

fg
<ctrl-c on perf-record>

perf script
From 5510c14e56f3e99e7dfd9208742143791e14a2e2 Mon Sep 17 00:00:00 2001
From: David Ahern <dsahern@xxxxxxxxx>
Date: Mon, 29 Apr 2019 11:22:07 -0700
Subject: [PATCH] notifier: add tracepoint to notifier_call_chain

Signed-off-by: David Ahern <dsahern@xxxxxxxxx>
---
 kernel/notifier.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/kernel/notifier.c b/kernel/notifier.c
index 6196af8a8223..9b65a9c56fd7 100644
--- a/kernel/notifier.c
+++ b/kernel/notifier.c
@@ -5,6 +5,8 @@
 #include <linux/rcupdate.h>
 #include <linux/vmalloc.h>
 #include <linux/reboot.h>
+#define CREATE_TRACE_POINTS
+#include <trace/events/notifier.h>
 
 /*
  *	Notifier list for kernel code which wants to be called
@@ -90,6 +92,7 @@ static int notifier_call_chain(struct notifier_block **nl,
 			continue;
 		}
 #endif
+		trace_notifier_call_chain(nb, val);
 		ret = nb->notifier_call(nb, val, v);
 
 		if (nr_calls)
-- 
2.11.0