[openib-general] opensm errors with ehca

[Brad Benton]

Troy Benjegerdes wrote on 10/30/2005 05:55:04 PM:

> The firmware on the IBM eHCA causes opensm to spit out these kinds of
> errors all the time..
> 
> Is there a way we can either not send P_KeyTable requests to any eHCA
> guids, or figure out what (if anything) is broken in their firmware?
> 
> Is this a spec violation, or just ambiguities in implementation?
...
> Oct 30 17:49:46 053861 [43005960] -> SMP dump:
>                                 base_ver................0x1
>                                 mgmt_class..............0x81
>                                 class_ver...............0x1
>                                 method..................0x1 (SubnGet)
>                                 D bit...................0x0
>                                 status..................0x0
>                                 hop_ptr.................0x0
>                                 hop_count...............0x2
>                                 trans_id................0x158c
>                                 attr_id.................0x16 
(P_KeyTable)
>                                 resv....................0x0
>                                 attr_mod................0x260000

Here is what is happening:  The attribute modifier for the P_KeyTable 
attribute is divided into two, 16-bit halves.  The most significant 16 
bits 
is information that is only valid for switches.  The problem here is that 
this SubnGet is for an HCA.  The firmware currently sees that the upper 
bits are non-zero and since it is not a switch, throws the packet away. 
The proper response would be for it to ignore the upper bits and process 
the MAD.  However, this is in firmware that won't be able to be changed 
quickly.  So, in the meantime as a work around, would it be possible to 
have the opensm clear out the upper 16 bits of the attribute modifier when 

making a P_KeyTable request of an HCA?

Thanks,
--Brad
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/general/attachments/20051101/ff52514c/attachment.html>