[ASL] Update on route_check.o problems

[Vikas K. <ka...@co...>]

	I spent the last few hours trying to debug route_check.o, and in
the process managed to irrepairably corrupt my filesystem. Before that I
could discover that the problem is essentially due to redefinition of some
netfilter symbols in newer kernel. Hence there are unresolved symbols and
null pointer dereferencing. 

	route_check.o works for stock 2.4.18 downloaded off 
http://www.kernel.org/pub/linux/kernel/v2.4/  After some effort it could 
be made to work with later kernels. But then it may not work with 2.4.18. 
In general it appears messy to keep the module uptodate with kernel 
versions.

	I now believe that ASL should do away with route_check.o and
become truly and completely a userspace library. Let me present my case.
route_check.o is there to duplicate route caching information which the
kernel already maintains but does not output to /proc/net/rt_cache.
route_check.o was written since we did not want to change the kernel
source AT ALL. See section 4.1.2 of the paper:
http://aslib.sourceforge.net/papers/HTML-paper-327/index.html


	It turns out there is another way to get the last_use_time out of 
the kernel, and that is through a NETLINK socket. The popular tool 
iproute2 does that. For example try the command: 

ip -s -s route list cache

What we want is the "age field". The information is there in the output,
we could put it to a file and grep it or something. But thats ugly. So
what ASL really should do is open a netlink socket to the kernel and read
this route cache information. This would have to be done in the function 
query_route_idle_time() in libASL/api.c.  This is in essence a replication 
of part of what iproute2 does.

The source of iproute2 is here:
(http://www-courses.cs.uiuc.edu/~cs397hou/iproute2-2.4.7-now-ss010824.tar.gz)
The relevant function is rtnl_rtcache_request(struct rtnl_handle *rth, int 
family) in ip/iproute.c and the file lib/libnetlink.c. So one needs to 
carefully look into these two files and incorporate that functionality in 
query_route_idle_time().


	
Effect on AODV.
--------------

AODV uses route_check.o in a more complicated way through the function  
query_route_idle_time_aodv(addr_t k,  int destination_flag) defined in 
ASL-a. This is to support a seemingly redundant subte optimization, see 
section 5.1.3 of the above paper. We could do away with it at the risk of 
not completely complying with the AODV draft. In anycase, 
query_route_idle_time_aodv() can also be modified to use all the 
information it needs from the kernel cache.


I have outlined my thoughts and possible options. I would appreciate input 
from the list.

The changes I propose are straightforward but tedious. Rolling out a well
tested new version may be 3-4 days of hard work. Unfortunately I am really
caught up in other work, and cant afford to do this for another 2 weeks or
so at least. In the unlikely event that someone on the list feels 
motivated, I would be glad to help. Inputs on the proposed changes are 
appreciated anyways.


Thanks
-vikas