[openib-general] makiing ibverb.h transport neutral -- 2nd draft

[James Lentini]

On Wed, 13 Jul 2005, Caitlin Bestler wrote:

>
>
>> -----Original Message-----
>> From: James Lentini [mailto:jlentini at netapp.com]
>> Sent: Wednesday, July 13, 2005 7:48 AM
>> To: Caitlin Bestler
>> Cc: openib-general; rdma-developers at lists.sourceforge.net
>> Subject: Re: [openib-general] makiing ibverb.h transport
>> neutral -- 2nd draft
>>
>>
>> Why do the transport dependent include files, iwarp_verbs.h
>> and ibonly_verbs.h, include the generic include file,
>> rdma_verbs.h? This will require consumer code to be aware of
>> the different iwarp and ib verbs.
>>
>> Consumers want to use a generic RDMA interface that hides the
>> differences between the various transports. Is this proposal
>> intended to address that requirement?
>>
>
> The transport dependent include file reference rdma_verbs.h
> because they are dependent on it. My usual style is to explicitly
> include such dependencies and rely on the #ifdefs to filter out
> redundant includes.
>
> Most consumer code would only #include rdma_verbs.h. Only those
> that needed to perform transport specific conneciton establishment
> or error recovery would use the transport specific headers.
> Providers, on the other hand, would inlude their transport specific
> header file.

Ok, so each file will be (directly) included by a different type of 
user: rdma_verbs.h by generic RDMA consumers, ibonly_verbs.h (which 
will become ib_verbs.h) by IB consumers, and iwarp_verbs.h by iwarp 
consumers.

In that case, why does the rdma_device structure in rdma_verbs.h 
contain entries for functions with IB specific types (query_gid, 
modify_device, process_mad, etc.)? Shouldn't these be in the ib 
specific header?

Keeping knowledge of the IB types in ib_verbs.h would remove the need 
to define the ib_gid type before including rdma_verbs.h in ib_verbs.h. 
Include files that require certain symbols to be defined before they 
can be included can quickly become difficult to maintain.