[c-nsp] Sup32 TCAM limit

[marco at linuxgoeroe.dhs.org]

>> What I do not understand is what would happen to the routes being
>> inserted
>> above this limit. Would such routes be 'soft' routed ? Is there also a
>> treshold for
>
> Prefixes that cannot fit in the TCAM are punted to the MSFC, and thus
> software switched, probably in an interrupt based CEF path.

That's what supposed to happen, yes. But I ran into this recently (with a
SUP720 now upgraded to XL) and the box didn't punt the packets. It just
hardware switched them based on what was in the TCAM.

Result was that the correct route was in the routing table and CEF FIB,
but packets got dropped or shunted elsewhere without any sign except the
TCAM_FULL messages, and those didn't look right either (no route in the
message at all or junk chars where it should be).

This was the sup running hybrid, IOS 12.2(17d)SXB8 on the MSFC and CatOS
8.6(1) on the SP.