***SPAM*** Re: [ofa-general] ***SPAM*** some problem about the forward tables in up*/down* algorithm

Sasha Khapyorsky sashak at voltaire.com
Wed Apr 15 06:04:54 PDT 2009


On 10:39 Wed 15 Apr     , Jordan wrote:
> Recently I have read the source code of the up*/down* routing algorithm. It
> seems that this algorithm only updates the
> hops[lid_no][port], does  not update the lft (linear forward tables). So ,
> how does the switch forward the packet ?

The routing engine in OpenSM has two methods - build_lid_matrices() and
ucast_build_fwd_tables(). build_lid_matrices() generates min hops tables
(lid matrices), ucast_build_fwd_tables() creates LFTs. You looked only
at build_lid_matrices() implementation.

> Does the switch look up the
> hops[lid_no][port] to forward the packet?

No.

> Another problem is that there are two arrays , lft and new_lft. I don't know
> the difference between these two tables, can anyone tell me ?

new_lft is how how LFT is generated by OpenSM, lft keeps a real LFT
state how it was received in LFT block set responses.

Sasha



More information about the general mailing list