[nsp] Breaking out components of IP Input process

[rpcbind at speakeasy.net]

12.2(8)T1 (ik9o3s). CEF appears to be working properly -- <300k packets total 
according to cef not-cef-switched. The only other thing about this, is there's 
.1q subints all around (tunnel endpoints are a .1q subint, with most traffic 
transiting them is input/output from a .1q subint).

I don't have the interfaces to move off the dot1q, but if it'd make a
difference, I could pretty easily justify a NM-2FE2W (assuming the 7400/VAM
would be the next significant step up).

On Thu, 19 Dec 2002, Siva Valliappan wrote:

> what version of code are you running?  there was an old bug that would
> cause us to process switch GRE-in-IPSec traffic.  you should actually
> be able to CEF / Fast-switch the traffic with the encryption offloaded to
> the AIM.
> 
> cheers
> .siva