Re: ath0 timeout problem - again

On Sunday 31 December 2006 01:31, Sam Leffler wrote:
I really do not know what this event means (ether type 5e4), for my
understandings it is vague in the source, so I am lost here

Seems pretty clear: it's the type field extracted from the ethernet
header of the oversized packet.  A quick check of sys/net/ethernet.h
shows no such ETHERTYPE defined.  So something in your network is
transmitting packets that either being rx'd incorrectly or, more likely,
corrupted in transit.

good so far, point here is that we can not limit that someone tx this
kind of packets but should find a way that this traffic do not DoS the
AP.

ipfw accepts 0x5e4 but at the end it does not get this packages
changing mtu on any interface does not change a thing


when we track the oversized packages they we found they are coming from
nat servers where wingate and similar softwares are running, when we cut
them out the mentioned traffic stopps and the AP does not interrupt
service anymore

So the packet is real and it's being dropped at the 802.3 layer as it
should.  If the printf is the problem remove it or rate limit it.  I
think it should be rate-limited or just stuck under a debug flag but
I'll leave that to someone else.  Alternatively we can enforce the mtu
in the ath driver and drop it there.

the problem is that the ath card stops working rendering the machine useless

I do not get this problem with a wi prism card, when I switch to a wi card I
get some wi timeout events but the card does not stop as the ath does

when I run the server with ath card not as bridge but as a gateway it also
does not stop working and the mentioned packages do also exist on the network
but are not even mentioned/noticed by the AP

enforcing a mtu limit and dropping the larger packages sounds as a good start
for me. If you could be so kind and make such a patch to try it out I get the
results in a day or two since I have machines which hang several times a day.


--

João







A mensagem foi scaneada pelo sistema de e-mail e pode ser considerada segura.
Service fornecido pelo Datacenter Matik https://datacenter.matik.com.br
_______________________________________________
freebsd-stable@xxxxxxxxxxx mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "freebsd-stable-unsubscribe@xxxxxxxxxxx"

Relevant Pages

  • Re: Udp sending performance in Gbit Ethernet
    ... > saw a sudden drop of perfromance for certain packet sizes. ... > there can be just so many reasons like interrupts and DMA. ... > such as the maximum DMA block size or alike, which forces the driver to ... Try different versions of the driver for the card. ...
    (microsoft.public.development.device.drivers)
  • Re: Adding optional packages to Solaris 10
    ... Part of the problem is that the packages that make up a particular ... cluster tend to be scattered over all 4 discs. ... Started out with a GFX8P Raptor video card, because that was what was in the machine, but there were issues with colours once we got Firefox up and running + window resize and moves were very, very slow. ... The Alpha is fast and makes a great cross compile server, but it's essentially a dead end. ...
    (comp.sys.sun.admin)
  • Re: Udp sending performance in Gbit Ethernet
    ... >> saw a sudden drop of perfromance for certain packet sizes. ... >> such as the maximum DMA block size or alike, which forces the driver to ... >> split frames into several DMA transfers. ... Try different versions of the driver for the card. ...
    (microsoft.public.development.device.drivers)
  • Re: Etch installation media doesnt boot (neither NETINST CD, nor floppies), but other distros do...
    ... I don't know pfsense but my guess would be that it uses OpenBSD's pf ... Packets have to traverse the kernel. ... Being that I plan to add/remove experimental packages somewhat often, I heavily favor binary packages, Debian being the best at that, AND also considering that I expect this box to do a lot of things at once - performance benchmarks from Scalable Networking raise a concern or two. ... new cf card, power off, swap cards, power on, done. ...
    (Debian-User)
  • Re: vr speed issues
    ... I get about 30-50% packet loss. ... I've got a VIA Rhine III card that I can dig out and put in if the data would be of any use/interest etc - FWIW I seem to recall being able to get reasonably close to wire speed when I was using it. ... I've had the vr hosts going through three different switches, patch cables, the in-house cabling, and the issue remains the same. ... device = 'VT6105M/LOM Rhine III PCI Fast Ethernet Controller' ...
    (freebsd-stable)