On Sun, 17 Oct 2004, Alberto Gonzalez Iniesta wrote: > On Sun, Oct 17, 2004 at 01:36:33PM -0000, James Yonan wrote: > > Alberto Gonzalez Iniesta <a...@inittab.org> said: > > > > > Hi all, > > > > > > After I decided to push OpenVPN 2.0 into Debian for future inclusion in > > > Sarge, I got this [1] bug report from one Debian user. It seems that 2.0 > > > doesn't get along well with udev, as opposed to 1.6. I don't use udev, > > > so I can't really tell, but has anything changed from 1.6 to 2.0 in the > > > way the device is handled? > > > > > > The suspicious message is: > > > > > > Oct 16 16:31:12 gimli wait_for_sysfs[5873]: error: wait_for_sysfs needs an > > update to handle the device '/class/net/tun0' properly, please report to > > <linux-hotplug-de...@lists.sourceforge.net> > > > > > > The bug submitter claims that OpenVPN 1.6 deals correctly with it. Maybe > > > it has something to do with device probing? > > > > The TUN/TAP handling code in 2.0 is nearly identical to the 1.6 code. The > > only change I can think of is the setting of txqueuelen on the TUN/TAP > > driver. > > Ok > > > Incidentally, the error shown in the bug report looks like he is trying to > > make a tunnel between 1.6 and 2.0 using incompatible key-method parameters. > > He needs to set --key-method 2 on the 1.6 side. > > > Has this changed between 1.6 and 2 beta11? > The man page says --key-method defaults to 1 and the ChangeLog does not say > otherwise.
Ah yes, the man page entry for --key-method hasn't been updated for 2.0. I will fix this for beta12. > I don't think this guy changed that to 2 just after he upgraded the > package. I'll ask him anyway, yo never know... > Could a problem with the TUN/TAP driver confuse the --key-method handling > code? I doubt it. The SSL/TLS handling code has almost no interactions with the TUN/TAP code. James
