[openib-general] Re: [PATCH] osm: osm_vendor_umad osm_vendor_get_all_port_attr bug

[Hal Rosenstock]

On Tue, 2005-08-30 at 14:52, Eitan Zahavi wrote:
> >>Without the above change I get:
> >>Listing GUIDs:
> >>Port 0:0xd9dffffff3d55 lid:0x0300 state:4
> >>Port 1:0xd9dffffff3d55 lid:0x0400 state:4
> >>Port 2:0xd9dffffff3d56 lid:0x0000 state:0
> >>
> >>After the simple change I get:
> >>Listing GUIDs:
> >>Port 0:0xd9dffffff3d55 lid:0x0300 state:4
> >>Port 1:0xd9dffffff3d55 lid:0x0300 state:4
> >>Port 2:0xd9dffffff3d56 lid:0x0400 state:4
> >>
> >>So as you can see - without the fix the lid of port 2 is presented as 
> >>the lid of port 1...
> > 
> > 
> > I understand the difference in the code and think the difference
> > perhaps relates to either a lack of clarity or confusion with the API as
> > follows: I don't see where it is defined what the index into the port
> > array means. I think we have 2 different interpretations and this
> > relates to how opensm/main.c  handles the results of calling this
> > routine. 

> I do not follow you. Do you suggest it is OK the port at index 1 will have
> the guid of port 1 but the lid and state of port 2?

No.

> I did not complain about what port is reported at what index:
> Just about the mismatch of guids and lids. Please see above.

I'm with you now. Thanks. Applied.

-- Hal