Bug 171707

Description of problem:
I would like for Julian Anastasov's kernel patches to be integrated into the 
stock kernel (or, if possible, as a kernel module installable via yum?).  The 
kernel patches can be found here: http://www.ssi.bg/~ja/#routes-2.6

As a related feature: also modify the network scripts, so when, say a gateway 
is defined in ifcfg-eth0 and a gateway is defined in ifcfg-eth1, the rules are 
setup according to the 
instructions "http://www.linuxquestions.org/questions/answers.php?
action=viewarticle&artid=490" (some more info also 
at "http://lartc.org/howto/lartc.rpdb.multiple-links.html")

Basically, the network script change is to set up rules to allow the router to 
use both gateways for internet access, but it causes problems if one gateway 
goes down.  That is where Julian's kernel patches come in.  The patches allow 
better dead gateway detection, so if one gateway goes down then all traffic is 
sent to the other gateway and (somehow, I don't know how) when the downed 
gateway comes back, it is automatically detected and traffic is again sent to 
that gateway.


Version-Release number of selected component (if applicable):


How reproducible: always


Steps to Reproduce:
1. Configure a router with multiple interfaces connected to internet gateways
2. add GATEWAY=blah.blah.blah.blah to the appropriate ifcfg-ethX for each 
interface.
3. run traceroutes to various sites on the net
....
4. break the connection to one gateway, to see failover
  
Actual results:
At step 3, all traceroutes show the traffic going over the same interface

At step 4, if the gateway is lost that all traffic from step 3 is going over, 
then all internet traffic is lost.  After setting up the correct ip route 
rules manually, then when one gateway is lost all traffic that would be sent 
to that gateway is lost and does not failover to the next gateway.

Expected results:
At step 3, each traceroute to a different site should go over a different 
gateway in sequential order.

At step 4, when one gateway is unreachable, that gateway is skipped and 
traffic goes to the next gateway until the unreachable gateway is detected as 
available again.


Additional info:
As I said, Julian Anastasov already did all the hard work.  I just wish his 
patches were in the stock kernels (or available as an addon via yum) so I can 
have the router automatically update with yum without breaking.  Modifying the 
scripts should be a trivial task, it's just running a few extra "ip route add" 
and "ip rule add" commands using variables already in the ifcfg-ethX files.