[ofa-general] [PATCH] ipoib: racing uses of ipoib_neigh in IPoIB-CM

[Yossi Etigin]

akepner at sgi.com wrote:
> On Thu, Jun 11, 2009 at 08:11:10PM +0300, Yossi Etigin wrote:
>> ....
>> How about checking the refcnt in the neigh_cleanup function, and if it's
>> not 1 give up the ipoib cleanup?
>>
> 
> ipoib_neigh_cleanup() can be invoked even when there are > 1 references 
> (e.g., if the interface is brought down that can happen).
> 
> So using the refcnt in the neighbour isn't going to work. 
> 
> I'm going to send a very simple patch in just a minute.
> 

If the interface is brought down, isn't ipoib cleaning the neighbours anyway?
(ipoib_flush_paths)