In article <44ECF520.3060908@lundman.net>,
Jorgen Lundman  <lundman@lundman.net> wrote:
>
>
>Christos Zoulas wrote:
>> In article <44EAEE0A.8010005@lundman.net>,
>> Jorgen Lundman  <lundman@lundman.net> wrote:
>> 
>> Compile a kernel with options MBUFTRACE and then use netstat to check
>> where the leak is.
>
>Done, (and thanks Pavel, nice pre-emptive).
>
>I should mention we have a rediculously large value for 
>"net.inet.tcp.send|recv-space", which certainly decreases the frequency of the 
>issue when set lower, albeit at a price in speed for those far away.
>
>But I put on the trace kernel just incase there was a rare bug triggered in the 
>interest of finding it. Normally it takes no more than a few hours for it to 
>happen, but very annoyingly, after booting the trace kernel, it has not
>happened 
>in 24 hours. Isn't that always the case.

So we see from the data that you have the same problem like I do. We are
leaking mbufs in the tcp recv. I've been reading the code, but I have not
found a problem...

christos