On Wed, Nov 26, 2003 at 03:09:10AM -0500, Jeff McAdams wrote: > Martijn van Oosterhout wrote: > >We've > >actually made some significant patches to that version at our work, > >including support for /dev/pts and a few other things. I'll see if I can > >get > >them for you. We fixed a *lot* of things just to get it to work on our > >systems. > > > Cool...I look forward to them. > > As much as you can break them out into discrete changes as possible > would be extremely helpful. Well, it's out of CVS so all I really need to do is find a way of getting CVS to show me all the patches plus descriptions. Would that be what you're looking for? Incidently, I worked out where the three second delay came from. The issue is that both pppd's are start simultaneously and the first one gets its LCP request off before the other end has run PPPIOCGCHAN so the packet simply gets queued up on the socket. As a result it's lost to the other end and is retransmitted after three seconds, triggering the rest of the negotiation. I'd like to be able to transmit all queued packets as soon as we receive the ioctl() but we can't because it never reaches the module. Two solutions: 1. Add a test to recv_core to pass on any queued packets when another comes in. Not ideal but certainly implementable. 2. Change the PPPoX code to execute ppp_input for any queued packets as soon as the ioctl is run. I'd prefer this but it would require changes to the main kernel source. 3. Change the PPPoX code to pass the ioctl() to the module. Might break stuff. -- Martijn van Oosterhout <kleptog@svana.org> http://svana.org/kleptog/ > "All that is needed for the forces of evil to triumph is for enough good > men to do nothing." - Edmond Burke > "The penalty good people pay for not being interested in politics is to be > governed by people worse than themselves." - Plato
Attachment:
pgp00135.pgp
Description: PGP signature
