Messages in this thread | | | Subject | Re: unregister_netdevice: waiting for DEV to become free (2) | From | David Ahern <> | Date | Wed, 1 May 2019 08:52:37 -0600 |
| |
On 5/1/19 7:38 AM, Tetsuo Handa wrote: > On 2019/04/30 3:43, David Ahern wrote: >>> 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 >>> >> >> with the header file this time. >> > > What is the intent of your patch? I can see that many notifiers are called. But > how does this help identify which event is responsible for dropping the refcount? >
In a previous response you stated: "Since I'm not a netdev person, I appreciate if you can explain that shutdown sequence using a flow chart."
The notifier sequence tells you the order of cleanup handlers and what happens when a namespace is destroyed.
The dev_hold / dev_put tracepoint helps find the refcnt leak but requires some time analyzing the output to match up hold / put stack traces.
| |