8.0-RELEASE-p3: 4k jumbo mbuf cluster exhaustion

[Andre Oppermann]

On 24.08.2010 14:37, Adrian Chadd wrote:
> On 23 August 2010 18:18, Andre Oppermann<andre at freebsd.org>  wrote:
>> It seems the 4k clusters do not get freed back to the pool after they've
>> been sent by the NIC and dropped from the socket buffer after the ACK has
>> arrived.  The leak must occur in one of these two places.  The socket
>> buffer is unlikely as it would affect not just you but everyone else too.
>> Thus the mbuf freeing after DMA/tx in the bce(4) driver is the prime
>> suspect.
>
> They don't stay leaked though. Killing the offending process sees
> mbuf's eventually returned.
> It isn't immediate though. It may be related to timing out existing
> socket connections or something?

Try "netstat -n -p tcp -x" to see whether one socket is holding on to
too much data.

> I haven't yet brought up the second box enough to start passing test
> traffic, so I can't provide any further details than this.

Testing with a different network card would help to narrow down the
area to look for the bug as well.

Can you describe your connection capturing setup some more?  Do you
use "ipfw fwd" or some form of NAT?

-- 
Andre