On 05/ 5/09 09:33 PM, Nicolas Morey-Chaisemartin wrote:
> Le 05/05/2009 15:25, [email protected] a écrit :
>> The hops table of ftree_sw_t is too small to hold the hop count
>> of max_lid. Changed sw_create() to allocate hops[max_lid+1]
>> not hops[max_lid].
>>
>> Signed-off-by: Line Holen <[email protected]>
> 
> 
> This patch seems right to me (at least agrees with other checks).
> However, I've been using the ftree algorithm without this fix in thousands of 
> tests and never had any seg fault problem and valgrind showed nothing 
> either...
> Would it be possible that the actual value is always < max_lid_ho ? 
> 
> Nicolas
> 

I haven't experienced any seg fault either. But I have seen lack of connectivity
to the node having lid = max_lid. This was because hop[max_lid] contained a 
value 
of 0 rather than 0xff (for some of the switches) which made the routing stop too
early.

Line


_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general

To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

Reply via email to