Re: [mpls-linux-general] How to do it?

[Abhijit G. <gab...@ee...>]

James R. Leu wrote :

>
>Since the MPLS forwarding plane is in the kernel and 'mplsadm' runs as
>a userland process, their are a few mechanisms availble for transfering
>information between the two:
>- /proc/sys/*
>- system calls ('ioctl' 'fnctl' etc..)
>- netlink
>
>Currently I use a mix of system calls and netlink (although I'm leaning more
>to using less and less netlink).

May I know why this is so? I believe netlink is becoming a pretty standard
way of user-kernel communications. Easily extensible, and right now
perhaps the only mechanism (I know) for seemlessly exchanging data between
user-space and kernel-space. Are there any major factors against using
netlink?


>
>If you look in the patched kernel source you can look at the two files that
>are the entry points for netlink and ioctl calls:  net/mpls/mpls_ioctl.c
>net/mpls/mpls_netlink.c.  Both of these files end up calling functions
>in differnt files (mpls_tunnel.c mpls_in_info.c mpls_out_info.c etc).
>
>Jim

-abhijit