[Xorp-hackers] The XrlPFSender race

[Bruce Simpson]

Hey guys,

I realize it's the winter break at the moment, and many of us may not be 
actively working over this period. Nevertheless, this is just a ping to 
say...

On 17/12/2009 16:51, Bruce Simpson wrote:
> ...
> Attached is a patch which compiles, and passes a 'scons check'. If you 
> guys could test further, that would be great.

...this is a patch intended to address the XrlPFSender race condition in 
XRL which Ben Greear found, intended to be clear in intent, as well as 
being the first XORP change which starts making wider use of the Boost 
C++ Libraries.

     Li Zhao ran into it too, and the seed for seeing this race 
condition manifest seems to be a high system load factor, either due to 
being run on an embedded target platform with less CPU horsepower, or in 
a situation creating high system load, e.g. multiple processes in Ben's 
virtualization branch.

     I've given the patch some basic testing in a router instance on my 
desktop, and it passes regression tests. It would be great to get test 
feedback from yourselves with this patch -- I would much prefer that it 
sees wider testing before it gets committed.

Thanks again, and enjoy your winter break,
BMS