[openib-general] mapping between IP address and device name

[James Lentini]

On Thu, 23 Jun 2005, Fab Tillier wrote:

>> From: Roland Dreier [mailto:roland at topspin.com]
>> Sent: Thursday, June 23, 2005 10:32 AM
>>
>>     James> Perhaps a bit of motivation of how the GID->IP service can
>>     James> be used is in order.
>>
>>     James> kDAPL uses this feature to provide the passive side of a
>>     James> connection with the IP address of the remote peer. kDAPL
>>     James> consumers can use this information as a weak authentication
>>     James> mechanism.
>>
>> This seems so weak as to be not useful, and rather expensive to boot.
>> To implement this, a system receiving a connection request would have
>> to perform an SA query to map the remote LID back to a GuidInfo
>> record, and then for each GID attached to the remote LID, somehow
>> retrieve the set of IP addresses configured for that GID (assuming
>> that is somehow even possible).
>
> This reverse lookup was something that I worked to accommodate in my 
> proposed changes to expand DAPL ATS to support multiple IP 
> addresses.  The revised DAPL ATS proposal establishes the notion of 
> a primary IP address that would be used for such validation. 
> However, I still think the reverse lookup (GID->IP) is weak as there 
> is no way to tell which IP the source really used.
>
> IMO it would be much better to put the source and destination 
> addresses into the CM private data, but this supposedly creates a 
> wire protocol which the DAT collaborative wants to avoid at all 
> costs.

If we place the IP address into the CM private data it will create 
interoperability problems. Upper layer protocols like NFS-RDMA, would 
only be able to communicate with implementations that also placed the 
address in the CM private data.

How would an NFS-RDMA client written directly to the IB verbs layer 
communicate with an NFS-RDMA server written to kDAPL?

The NFS-RDMA client in this configuration would need to place the IP 
address in the CM private data. In effect, we've added a new (albeit 
small) component to the NFS-RDMA protocol.

Rather than add this mechanism into every ULP, I think it would be 
better to provide it as a core service of the network.

james