For monitoring gated, would something on the order of
/proc/net/route_log work better? I could work out the
code to do logging to /proc. Would notification of
routing changes work better than scanning the kernel
routing tables? I'm not sure of what exactly your gated
monitor does....
I just think there might be another way of looking at the
problem.
>Its not even neccessarily compatibility thats an issue. We just need
>a way to say 'Dont cache me Im potentially big', or for the code to
>get a 32K block and then say "Stuff this Im not caching any more".
This should be easy to do, but does not solve the any
problems. Might as well stay with the current 4K buffer.
We might also try removing the need for huge proc files.
Are there any other potentially huge (>1MB) /proc/net
files?
Rob
(rriggs@tesser.com)