[SSI-devel] Re: [SSI-users] no single point of failure

[Brian J. W. <Bri...@hp...>]

Walker, Bruce J wrote:
> I have been hoping someone would try out the "bonding".  Sorry it wasn't
> someone before you.  I am very interested in it working so I expect it
> will get resolved in the next week or 2 (if not before).

My response is more development-related, so I'm copying the devel 
mailing list.

The current system call for initializing the ICS interface during boot 
takes an unconfigured network interface (ethx), an IP address and a 
netmask. The /linuxrc script in the initrd gathers up this information, 
makes the appropriate system call, and the kernel configures the network 
interface and initializes it as the ICS interface.

In order to initialize other types of ICS interfaces, I think we should 
decouple the configuration of the network interface and the 
initialization of ICS. The system call to initialize ICS should take an 
already configured interface, whether it's ethx, bondx, etc. This 
shouldn't take too much effort, since it involves stripping 
functionality out of a system call, rather than adding functionality to 
it. Of course, mkinitrd will also have to be modified to do the network 
configuration itself.

My $.02,

Brian