This is the mail archive of the ecos-discuss@sources.redhat.com mailing list for the eCos project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: MBUF's in stack TCP/IP


> I have run tests repeatedly opening and closing sockets to do ping
> operations, with hosts pinging and flood-pinging the target, and also
> performing continuous SNMP interrogations of the target (ie. connect to a
> UDP server, do lots of traffic, disconnect).  These run OK for days at a
> time, honest!

Even with all these tests we have found one case it can leak. I
_think_ in eth_drv_recv() it can leak an mbuf if the MCLGET() fails. I
don't have the patch to hand and i don't remember the details.

When you have a problem, do you run out of clusters or mbufs, or both?
If you run out of clusters first and then mbufs, then it could be
this. We have only seen this happen when pushing the CPU to 100% load
with hugh amounts of network traffic. A plain TCP connection cannot
cause it to happen since it does flow control and the load drops to
below 100%. If you just run out of mbufs you probably have another
problem.

I plan to submit a set of net patches to RedHat in the next couple of
days which will include a fix for this.

        Andrew



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]