Port-xen archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
Re: mbuf leak on 6.0.1?
On Mon, Jan 28, 2013 at 02:41:37AM +0100, Emmanuel Dreyfus wrote:
> Manuel Bouyer <bouyer%antioche.eu.org@localhost> wrote:
>
> > Please rebuild the kernel with MBUFTRACE and look at where the mbufs go with
> > netstat -mssv
>
> Here it is:
>
> ping: sendto: No buffer space available
> 13541 mbufs in use:
> 13077 mbufs allocated to data
> 464 mbufs allocated to packet headers
> 89397 calls to protocol drain routines
> small ext cluster
> lo0 inuse 0 0 0
> claims 47216 0 0
> releases 47216 0 0
> xennet0 rx inuse 0 0 0
> claims 3441889 3441889 0
> releases 3441889 3441889 0
> xennet0 tx inuse 716 27 27
This looks a bit high, but that's not what explains the problem.
> claims 10658569 5806650 5805497
> releases 10657853 5806623 5805470
> route inuse 0 0 0
> claims 29 0 0
> releases 29 0 0
> unix inuse 0 0 0
> claims 3677611 664185 664183
> releases 3677611 664185 664183
> tcp inuse 57 0 0
> claims 111934 0 0
> releases 111877 0 0
> tcp sock rx inuse 0 0 0
> claims 1878692 1868133 121
> releases 1878692 1868133 121
> tcp sock tx inuse 78304 1033 1033
> claims 7689179 5595451 5595451
> releases 7610875 5594418 5594418
> small ext cluster
So it looks like most mbufs are in a tcp socket send queue.
netstat -f inet should show which one.
It may not be a mbuf leack, just that the network stack is indeed using
a lot of ressources.
--
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
NetBSD: 26 ans d'experience feront toujours la difference
--
Home |
Main Index |
Thread Index |
Old Index