Correct accounting when dropping fragment queues

[thj]

Summary

When dropping a fragment queue count the number of frags in the queue.

Test Plan

Scapy test program here:
http://people.freebsd.org/~thj/tests/sendpartialchain.py

The test program sends the final two fragments of a three fragment UDP chain,
but does not send the final fragment. When run before this change the timeout
counter will increase by 1 after frag timeout (60 seconds), with the change it
will increase by 2.

[jtl]

sys/netinet6/frag6.c

864

My sense is that this should increment timeout by one (since one reassembly queue timed out), but also increment dropped by the number of fragments. What do you think about that approach?

(I'm not sure there is a "right" approach, as long as we document the approach we took?)

883

Same thing here.

901

Same thing here.

[bz]

Not sure where the documentation will go but netstat -s -p ip6 says "fragments" not "packets" so this change sees correct.

[bz]

Hit enter too early; you might want to be slightly more verbose in the commit message as to why this change is done.

[jtl]

Please commit with "Approved by: jtl"

[thj]

typo in the commit message stopped this being closed automatically