dst cache overflow in 2.6.8

["Chris Bennett" <chris@xxxxxxxxxx>]

There appears to be a pretty serious router bug in kernel 2.6.8.  One 
reference to it is here: 
http://www.debiantalk.com/_Bug279666_kernel-image-2_6_8-1-k7_Runs_out_of_network_buffers-10116882-5788-a.html 
and a followup that it may now be fixed in later kernels here: 
http://lists.debian.org/debian-kernel/2004/12/msg00233.html.

This is my personal experience with it....

My router fails few days now and when it fails the syslog fills up with 
messages saying "kernel: dst cache overflow".

I have done a lot of googling tonight to research this.  Apparently this has 
happened to other people in the past, but not very often for a long stretch 
until recently when I can find mention of this problem once again popping up 
again starting around September.

> From what I gather this error means that the routing cache filled up.  The 
maximum size of the routing cache can be set by changing 
/proc/sys/net/ipv4/route/max_size.  The default on my mandrake install is 
32768.

The current size of the routing cache can apparently be determined by "grep 
ip_dst_cache /proc/slabinfo".  On my router this value is slowly, but 
constantly, trending upwards at all times.  However, if I list the cache 
with "route -Cn", or "cat /proc/net/rt_cache", I see significantly fewer 
entries than what slabinfo reports.

Two things have changed since the time when my router ran smoothly without 
any problems:

1) I upgraded from Mandrake 9.2.1 to Mandrake 10.1
2) I now have two WAN connections coming into the router that map to 
different halves of the same subnet on my LAN.

I use HTB for bandwidth shaping, and I also map my public IP addresses to 
private address space, but this part of my config has not really changed.

At this point I'm just throwing this out as a heads-up  At the moment I'm 
just monitoring the ip_dst_cache value from /proc/slabinfo, and when it 
starts to get close to the vavlue in /proc/sys/net/ipv4/route/max_size I 
just reboot the router... not the most elegant situation, but at least I no 
longer have random crashes.  I'm going to look into trying a different 
kernel in the next day or so to see if that resolves the issue.

Chris 



--
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.298 / Virus Database: 265.6.4 - Release Date: 12/22/2004

_______________________________________________
LARTC mailing list / LARTC@xxxxxxxxxxxxxxx
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/