Kernel memory corruption(?) with age(4)

[YongHyeon PYUN]

On Sun, Jun 26, 2011 at 06:53:57AM +0200, Michael Schuh wrote:
> Hi,
> 
> i am not on the list atm.
> 
> i found this entry after digging in google during problems with a normally
> rock stable
> system ( under linux).
> 
> I can also confirm the behaviour from the first post.
> After a long hard night digging in the system and changing some options
> i found a repeatable way to crash the system ultrafast.
> 
> at the initial crashes the system crashed after copying around 800GB.
> the second and third and further attempts/impacts got issued everytime
> after an  ~ hour from reboot on. leaving the system withouth any power,
> changing some hardware
> forth and back didn't changed anything, except the network card.
> 
> a happening beside that: immediatly after the initial boot will get the
> ntpdate and ntpd started.
> by this, the ntpd complains about unsresolveable hostnames from ntp.conf.
> if you log in and restart the ntpd by hand, everything is fine, no
> complains,
> no unresolveable hostnames.
> the hostnames are well known and tested and the DNS works still fine.
> this happening will also leave if you change the Ethernet-adapter to an
> Intel Card.
> 
> My system crashed so fast that i didn't got any log entry nor a crashdump.
> 
> my last try's crashed my system after 100 - 120 seconds.
> 
> HW: AMD P5K with Atheros age-Interface enabled, 6GB Corsair DDR2 800MHZ
> (pc6400) RAM ( latency 5-5-5-5-18)
> Ati Radeon HD4350 with 512MB Onboard Memory, 3 disks 2 PATA, 1 SATA.
> 
> OS: FreeBSD/amd64 8.2-RELEASE ( unmodified original RELEASE install )
> 
> REQUIREMENTS to repeat the issue:
> box A: a stable FreeBSD or Linux Box with GBitEthernet interface and no
> essential networkload
> box B: the affected FreeBSD Host with the age interface.
> 
> the age0 interface is configured with default options.
> only the ip-configuration got applied through rc.conf
> 
> ifconfig_age0="192.168.1.3 netmask 255.255.255.0"
> 
> PROCEDURE:
> On Box A ( p.e IP: 192.168.1.2 )  issue the command:
> 
> # dd if=/dev/zero |nc -l -p 1666
> 
> On Box B issue a complementary command like:
> 
> # nc 192.168.1.2 1666| dd of=/dev/null
> 
> I am sorry i cannot test a patch at this time cause this is a backup server,
> that has to get asap into production.
> After disabling the age-Interface in the bios settings and inserting a Intel
> GBit NIC the System runs stable again.
> 
> i hope this informations will help a bit.
> if you need to message me, please message me directly cause i am not
> registered for this list.
> 

I think the issue was already fixed(r220249, r220252). Apply the
patch and rebuild your kernel after downloading it from the
following URL.
http://svnweb.freebsd.org/base/head/sys/dev/age/if_age.c?r1=219902&r2=220252&view=patch
If you see the same issue again with this patch please let me know.