Kernel memory leak in ipfw internal NAT 10-STABLE?

[Karl Denninger]

It certainly looks like there is one; with the internal NAT capability enabled (which was fine in 9.2-STABLE) I have a very busy machine that consumed all (~20+GB!) of its RAM into the "inact" bucket and was threatening to deadlock -- it got rid of all the free RAM and slowly started forcing the working set out onto swap as well.

Kernel rev is FreeBSD 10.0-STABLE #13 r263037M

--
-- Karl
k...@denninger.net

None of the user processes showed a problem and killing them all did not drop the allocation. Nor did unloading everything I could unload from the kernel either. I reverted to running natd, and the problem has disappeared. A quick look through the commits doesn't show anything suspicious -- and this was working fine on 9.2-STABLE.

smime.p7s
Description: S/MIME Cryptographic Signature