Aie, seems I wrote the reply but got distracted and didn’t hit “send”.. sorry for that…
thanks Matt and Fan for clarification, indeed I missed the forest behind the lot of trees! Please feel free to do the cherrypicks for these 4 into stable/2202 within pre-RC2 window., i will merge them. Also seems seems like coverity found something to complain about in /src/plugins/wireguard/wireguard_output_tun.c: 285 in wg_add_to_async_frame() - could the SME take a glance at it ? --a > On 19 Jan 2022, at 17:47, Zhang, Roy Fan <roy.fan.zh...@intel.com> wrote: > > Thanks Matthew to point this out. > > To add a bit more detail here > Wireguard data-path encryption and decryption uses aad size of 0. Hence we > have to add new async AEAD op ID to represent this aad size. > > Regards, > Fan > > From: Matthew Smith <mgsm...@netgate.com> > Sent: Wednesday, January 19, 2022 4:36 PM > To: Andrew 👽 Yourtchenko <ayour...@gmail.com> > Cc: Zhang, Roy Fan <roy.fan.zh...@intel.com>; vpp-dev <vpp-dev@lists.fd.io>; > Andrew Yourtchenko (ayourtch) <ayour...@cisco.com>; Kinsella, Ray > <ray.kinse...@intel.com>; Oginski, GabrielX <gabrielx.ogin...@intel.com> > Subject: Re: [vpp-dev] Is it too late for Wireguard patches getting in for > VPP22.02? > Asdfasd > Hi Andrew, > > The change in crypto.h that you called out as being whitespace-only actually > changes more than whitespace... It also appends '_ (CHACHA20_POLY1305, > "chacha20-poly1305", 32, 16, 0)' to foreach_crypto_aead_async_alg. > > -Matt > > > On Wed, Jan 19, 2022 at 10:21 AM Andrew 👽 Yourtchenko <ayour...@gmail.com> > wrote: > Hi Fan, > > With my release manager hat on: > > the first three patches are solely contained (minus seemingly > whitespace change in 34660? can it be avoided ?) within wireguard > plugin, which has "experimental" status, with which I would be happy > to err on the side of keeping the velocity - so once the nit i pointed > out is taken care of, I would be happy to merge the cherry-picks into > the stable/2202, provided it is "soon" for some very proximate value > of "soon" :) > > the last patch to crypto-dev is an improvement, and looks fairly > straightforward addition, backwards compatible, and was +2'd by Damjan > back in December. So, from the "technical" standpoint it should have > been in, but there was new year break period, etc, etc. So I would > again would be fine with merging the cherry-pick into stable/2202, but > for this one" the value of "soon" should not exceed "end of this > week". > > In case no-one from the community objects to the above by this Friday > 12:00 UTC, let's get these as cherry-picks into stable/2202. > > (p.s.: the 22.02 RC1 is not tagged yet, but the v22.06-rc0 tag is > already on master, I would rather not mess with that). > > --a > > > On 1/19/22, Fan Zhang <roy.fan.zh...@intel.com> wrote: > > Hi, > > > > Sorry for the late notice, but we have a bunch of patches waiting to be > > reviewed/merged if possible for VPP22.02. > > > > The patches do 2 things > > > > * Optimizing wireguard performance by introducing burst processing of > > packets and chacha-poly encryption/decryption. > > * Adding async mode to wireguard so the crypto can be processed by > > sw-crypto-scheduler or QAT. > > > > They are > > https://gerrit.fd.io/r/c/vpp/+/34324 > > https://gerrit.fd.io/r/c/vpp/+/34660/2 > > https://gerrit.fd.io/r/c/vpp/+/34661/2 > > https://gerrit.fd.io/r/c/vpp/+/34662/4 > > > > We know the patches are not perfect - as Matthew already pointed out there > > are some improvements can be done (thank you very much Matthew!). > > The bottom line is > > > > * They passed the wireguard unit tests, and we will keep improving > > them. > > * The code change had >5% performance improvement with small packets. > > > > But are they too late for VPP22.02? Much appreciate for the help in > > advance! > > > > Regards, > > fan > > > > > >
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#20778): https://lists.fd.io/g/vpp-dev/message/20778 Mute This Topic: https://lists.fd.io/mt/88537247/21656 Group Owner: vpp-dev+ow...@lists.fd.io Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
