[Openswan Users] kernel memory leak 2.4.7, 2.4.33 - misconfiguration or bug? (fwd)

[Brad Langhorst]

On Sun, 2007-05-13 at 08:54 -0400, Michael Richardson wrote:

>   Yes, those are in units of the number of objects.
>   Probably this is the case.  The question is why are you loosing
> skbuffs...  This requires some instrumentation of the code.
>   
>  
> This is an updated shell script version of "slabwatch"
> 
> #!/bin/sh
> 
> while true
> do
>   d=`date +%s`
>   awk '{ printf "echo >>%s '$d' %s %s\n", $1, $2, $3;}' /proc/slabinfo | tr -d '()'
>   sleep 1
> done | sh

how about this

while true
do
d=`date +%s`
awk '{ printf "%s '$d' %sMB\n", $1, $2* $4/1024/1024;}' /proc/slabinfo |
tr -d '()'
done 

i'm not sure what that echo and pipe to sh are supposed to do...

When I turn off natt, the leak is much slower, but still present.
only 1M of skbuff after 100+ M downloaded.

as I watch it, it does not increase steadily, it goes in jumps.


brad

-- 
Brad Langhorst
CTO - CoopMetrics