On Mon, Jun 10, 2019 at 4:50 AM Samuli Seppänen <sam...@openvpn.net> wrote: > > Il 07/06/19 20:19, Selva Nair ha scritto: > > Hi JJK, > > > > On Fri, Jun 7, 2019 at 11:09 AM Jan Just Keijser <janj...@nikhef.nl> wrote: > >> > >> hi all, > >> > >> in the eduVPN project we've run into a strange issue: > >> HP Envy laptops running Windows 10 have a "handy" feature to > >> automatically switch from wifi to a 'wired' adapter if one is detected. > >> The use case behind this is that if you put your laptop in a docking > >> station (with LAN) it will auto-switch to it. > >> How does this relate to OpenVPN: when an OpenVPN connection is > >> established, the HP Envy driver gimmick detects the tap-windows device > >> as a LAN adapter and switches off wifi -regardless of whether you are on > >> a LAN or not. Thus, "plain" OpenVPN 2.4 becomes quite hard to use on > >> these laptops. > >> Now the strange part is, that both Viscosity and NordVPN - each with > >> their own tap-win adapters do NOT have this problem. > >> So my question now is: what could be causing this and how can we > >> resolve it. > >> The patches the Viscosity folks made to the tap-win driver for their > >> setup are in github (pull request #47, IIRC) and I don't see anything > >> truly strange in that patch. > >> The only thing that seems to be different is whether a device is > >> sporting the UI flag in the "characterstics" section. > >> > >> Does anybody have a clue how to best approach this? > > > > One thing I've noticed is that in the driver inf file viscosity has set > > > > characteristics = 0x1 ; NCF_VIRTUAL > > > > It goes into the registry entries for the device node. > > > > They don't seem to have a link to download the patched source (have to > > make a request by email -- no idea whether a paid license is also > > required), so I haven't looked for any other changes in the sources. > > Possibly you could ask for the source code and do a diff. > > > > > > Selva > > Hi, > > This is tangential, but I wonder if these "characteristics" (also) have > an effect on which HLK tests are required? Rozmansi's "Introduce TAP > adapter as a virtual device" does not touch them: > > https://github.com/OpenVPN/tap-windows6/pull/84/files > > That PR, as-is, did not help narrow down the scope of HLK tests. > > The wintun driver has this in characteristics: > > wintun.inf:Characteristics = 0x1 ; NCF_VIRTUAL > > Default tap-windows6 has > > version.m4:define([PRODUCT_TAP_WIN_CHARACTERISTICS], [0x81]) > > Thoughts?
Interesting --- that should get into the inf file, but on my machine inf has [tap0901.ndi] CopyFiles = tap0901.driver, tap0901.files AddReg = tap0901.reg AddReg = tap0901.params.reg Characteristics = *IfType = 0x6 ; IF_TYPE_ETHERNET_CSMACD *MediaType = 0x0 ; NdisMedium802_3 *PhysicalMediaType = 14 ; NdisPhysicalMedium802_3 Note that Characteristics is blank as if @PRODUCT_TAP_WIN_CHARACTERISTICS@ did not get expanded. . Do I have a wrong .inf? By the way, 0x81 stands for NCF_VIRTUAL|NCF_HAS_UI No idea what difference it may make to HLK or anything else. Selva _______________________________________________ Openvpn-users mailing list Openvpn-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/openvpn-users
