I will not claim to understand the details here, or that I have actually
verified proper outgoing hashes - I have seen that the unit tests works,
that the memory operations seem safe (everything uses SIPHASH_KEY_SIZE,
the reference implementation does 2x 8 byte loads), that my t_server
test bed still happily talks to tls-crypt(-v2) clients.
"Being faster than SHA256" is the key here - the old implementation
exposes servers to a certain amount of "CPU exhaustion DoS" (due to
"work has to be done on each incoming packet to see if it's a reply
or not"). With the faster implementation, the bottleneck is now more
likely "the OS TCP/IP stack", and not something we can do anything about.
This is a master / 2.8-to-be feature, as lots of new infrastructure is not
in release/2.7.
Your patch has been applied to the master branch.
commit ffa0e712eadab73b7e7b94de637d88c539cf6bc6
Author: Arne Schwabe
Date: Thu Aug 13 19:02:05 2026 +0200
Replace SHA256 with SIPHASH24 in HMAC cookie approach
Signed-off-by: Arne Schwabe <[email protected]>
Acked-by: Frank Lichtenheld <[email protected]>
Gerrit URL: https://gerrit.openvpn.net/c/openvpn/+/1827
Message-Id: <[email protected]>
URL:
https://www.mail-archive.com/[email protected]/msg38316.html
Signed-off-by: Gert Doering <[email protected]>
--
kind regards,
Gert Doering
_______________________________________________
Openvpn-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/openvpn-devel
