[ofa-general] Re: [PATCH] ibsim: Add support for vendor ID and system image GUID

[Sasha Khapyorsky]

Hi Hal,

On 17:45 Wed 27 Aug     , Hal Rosenstock wrote:
>>> diff --git a/ibsim/sim_net.c b/ibsim/sim_net.c
>>> index 6e3c0e9..146bcde 100644
>>> --- a/ibsim/sim_net.c
>>> +++ b/ibsim/sim_net.c
>>> @@ -190,7 +190,9 @@ char (*aliases)[NODEIDLEN + NODEPREFIX + 1];	// 
>>> aliases map format: "%s@%s"
>>>   int netnodes, netswitches, netports, netaliases;
>>>  char netprefix[NODEPREFIX + 1];
>>> +int netvendid;
>>>  int netdevid;
>>> +uint64_t netsysimgguid;
>>>  int netwidth = DEFAULT_LINKWIDTH;
>>>  int netspeed = DEFAULT_LINKSPEED;
>>>  @@ -324,11 +326,12 @@ static Node *new_node(int type, char *nodename, 
>>> char *nodedesc, int nodeports)
>>>  	}
>>>   	mad_set_field(nd->nodeinfo, 0, IB_NODE_NPORTS_F, nd->numports);
>>> +	mad_set_field(nd->nodeinfo, 0, IB_NODE_VENDORID_F, netvendid);
>>>  	mad_set_field(nd->nodeinfo, 0, IB_NODE_DEVID_F, netdevid);
>>>   	mad_encode_field(nd->nodeinfo, IB_NODE_GUID_F, &nd->nodeguid);
>>>  	mad_encode_field(nd->nodeinfo, IB_NODE_PORT_GUID_F, &nd->nodeguid);
>>> -	mad_encode_field(nd->nodeinfo, IB_NODE_SYSTEM_GUID_F, &nd->nodeguid);
>>> +	mad_encode_field(nd->nodeinfo, IB_NODE_SYSTEM_GUID_F, &netsysimgguid);
>>>     
>>
>> And when netsysimgguid was not parsed for this node, it will put previous
>> value there (or "0" if it was never parsed)?
>>   
> Is "state" for a node in the topology file needed to deal with this ? 
> Something like the following: When the vendor ID line is seen, reset 
> netsysimgguid and if 0 when new_node is invoked, then use the node GUID as 
> currently done. Does that make sense ?

Why to not reset netsysimgguid unconditionally at end of new_node()?

The rest could be as you said:

	mad_encode_field(nd->nodeinfo, IB_NODE_SYSTEM_GUID_F,
			 netsysimgguid ? &netsysimgguid : &nd->nodeguid);

Sasha