kernel panic - ip_vs_conn_hash(): request for already hashed

[Roberto Nibali]

G'day,

BTW, congratulations to all Australians for yesterday's (today for you 
guys) soccer game, it was simply awesome! I'm still a bit hung-over :). 
Too bad they play Italy now, I don't know which side I should be on ...

>> How frequent do does 'oops' show up?
> 
> On average, about once every 2 weeks.

Could you send us along your slabtop output and /proc/interrupts, 
please? It's probably not related to your oops, but better safe than 
sorry :).

> .config

Massive! :)

However, thanks, comments are below:

> CONFIG_HIGHMEM64G=y
 > CONFIG_HIGHMEM=y

Do you need high memory support? This normally degrades performance on 
32bit boxes, especially with regard to DMA writes.

> CONFIG_HIGHIO=y

This can give weird results in VFS, we've fixes some issues there in 
2.4.33-RC1. Should not be related to your problem though.

> CONFIG_NR_CPUS=8
> CONFIG_X86_NUMA=y
> CONFIG_X86_SUMMIT=y
> CONFIG_X86_CLUSTERED_APIC=y

What kind of machine do you have? dmidecode output, please.

> CONFIG_X86_IO_APIC=y
> CONFIG_X86_LOCAL_APIC=y

Probably not needed for your machine and can cause the problem indicated 
by another poster which asked you to instrument irqbalance.

> CONFIG_IP_VS=m
> CONFIG_IP_VS_TAB_BITS=16
> CONFIG_IP_VS_RR=m
> CONFIG_IP_VS_WRR=m
> CONFIG_IP_VS_LC=m
> CONFIG_IP_VS_WLC=m
> CONFIG_IP_VS_LBLC=m
> CONFIG_IP_VS_LBLCR=m
> CONFIG_IP_VS_DH=m
> CONFIG_IP_VS_SH=m
> CONFIG_IP_VS_SED=m
> CONFIG_IP_VS_NQ=m
> CONFIG_IP_VS_FTP=m

Ok.

> CONFIG_DEBUG_KERNEL=y
> CONFIG_DEBUG_STACKOVERFLOW=y
> CONFIG_DEBUG_HIGHMEM=y
> CONFIG_MAGIC_SYSRQ=y

Please make sure, you've also set those (at least the frame pointer one) 
to get reliable stack traces:

CONFIG_DEBUG_SLAB=y
CONFIG_FRAME_POINTER=y

Thanks and best regards,
Roberto Nibali, ratz
-- 
echo 
'[q]sa[ln0=aln256%Pln256/snlbx]sb3135071790101768542287578439snlbxq' | dc