Dear John and Selva, I've tried almost all advises I was able to find in Internet but still no luck. Also, I've tried to test this hacky tap-adapter from jkunkee. I was able to build the tapinstall itself, but failed to add new adapter with it.
Are you able to recreate my scenario and confirm the problem as I have on your side? Do you have any suggestions I can try? On Mon, Oct 8, 2018 at 10:43 PM Jon Kunkee <jkun...@microsoft.com> wrote: > > (Side note: I do work for Microsoft, but I don't speak for Microsoft, > don't know the networking stack well, and am not an NDIS dev.) > > >>>> I don't know what causes this (the TAP driver?) but hopefully others > may chime in with some clue. > > >>> I wonder if this will be addressed by adding the missing > buffer-remaining query OIDs. Unfortunately I don’t know enough to say for > sure, but they are used for protocol behavior hinting. > > Please note that this is just one theory. I don't currently have the > networking expertise to determine what the problem is given the information > in this thread, so it's possible something else entirely is the problem. > > >> That was what I was hinting at too, but in the same boat here: not > knowing enough about NDIS architecture I'm not in a position to make a > patch for the missing OIDs. > >> > >> That said, as server 2016 uses the same kernel as Windows 10 (doesn't > it?) such issues should show up on that platform too. Or are there some > network protocol improvements that are enabled only on server 2016? > > Windows Server 2016 is roughly equivalent to the Windows 10 Anniversary > Update (1607), though it has more changes backported since its > servicing-vs-upgrade cadence is very different from desktop Windows' > cadence. That said, yes, it's the same kernel. > > The question here is whether Server has more features lit up than client, > and my digging suggests 'not in the TCP/IP stack'--but Server 2016 did have > significant changes and does allow a lot more configuration (as described > earlier with 'custom' profiles): > > https://www.speedguide.net/articles/windows-10-anniversary-updates-to-tcp-6266 > > Server is more perf-sensitive than desktop, so it wouldn't surprise me if > the Server 2016 TCP/IP stack started to rely on these OIDs for congestion > control. The Server 2016 HLK tests do require that they be implemented. > (This is one reason it is not (yet!) certified.) > > If you're feeling daring, you could try rebuilding the driver with this > extremely hacky change and see if the perf changes: > > https://github.com/jkunkee/tap-windows6/commit/2706c9a98bd25a0fee6559ac132d75622475c26e > > HTH, > Jon > > -- Best regards, *Rostyslav Maryliak* System Administrator *iDeals™ Solutions Group*| + 38(073)437-72-51 <%2B%2038%28093%29575-35-16> | Skype: rostyslav.maryliak.ideals| *rostyslav.maryl...@idealscorp.com <dmitry.zaporozhche...@idealscorp.com>* | www.idealsvdr.com <http://www.idealscorp.com/> -- CONFIDENTIALITY NOTE: The information transmitted, including attachments, is intended only for the person(s) or entity to which it is addressed and may contain confidential and/or privileged material. Any review, retransmission, dissemination or other use of, or taking of any action in reliance upon this information by persons or entities other than the intended recipient is prohibited. If you received this in error, please contact the sender and destroy any copies of this information.
_______________________________________________ Openvpn-devel mailing list Openvpn-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/openvpn-devel
