Re: [REGRESSION] NFS is creating a hidden port (left over from xs_bind() )

From: Trond Myklebust
Date: Thu Jun 18 2015 - 15:25:08 EST


On Wed, Jun 17, 2015 at 11:08 PM, Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
> On Fri, 12 Jun 2015 11:50:38 -0400
> Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
>
>> I reverted the following commits:
>>
>> c627d31ba0696cbd829437af2be2f2dee3546b1e
>> 9e2b9f37760e129cee053cc7b6e7288acc2a7134
>> caf4ccd4e88cf2795c927834bc488c8321437586
>>
>> And the issue goes away. That is, I watched the port go from
>> ESTABLISHED to TIME_WAIT, and then gone, and theirs no hidden port.
>>
>> In fact, I watched the port with my portlist.c module, and it
>> disappeared there too when it entered the TIME_WAIT state.
>>

I've scanned those commits again and again, and I'm not seeing how we
could be introducing a socket leak there. The only suspect I can see
would be the NFS swap bugs that Jeff fixed a few weeks ago. Are you
using NFS swap?

> I've been running v4.0.5 with the above commits reverted for 5 days
> now, and there's still no hidden port appearing.
>
> What's the status on this? Should those commits be reverted or is there
> another solution to this bug?
>

I'm trying to reproduce, but I've had no luck yet.

Cheers
Trond
--
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/