[openib-general] mthca v. current ib_verbs

Masanori ITOH itoumsn at nttdata.co.jp
Mon Aug 9 21:17:51 PDT 2004


Hi Roland,

From: Roland Dreier <roland at topspin.com>
Subject: Re: [openib-general] mthca v. current ib_verbs
Date: Mon, 09 Aug 2004 18:23:34 -0700

>     Masanori> Those messages sometimes appear when you are building
>     Masanori> your source outside linux kernel source tree. In this
>     Masanori> case, especially the modules containing ib_req_notify_cq
>     Masanori> or so.
> 
> Do you have any idea why?  If the code includes ib_verbs.h properly,
> then the definitions below mean that no references to either
> ib_poll_cq() or ib_req_notify_cq() are generated:
> 
> 	static inline int ib_poll_cq(struct ib_cq *cq, int num_entries,
> 				     struct ib_wc *wc)
> 	{
> 		return cq->device->poll_cq(cq, num_entries, wc);
> 	}
> 
> 	static inline int ib_req_notify_cq(struct ib_cq *cq,
> 					   enum ib_cq_notify cq_notify)
> 	{
> 		return cq->device->req_notify_cq(cq, cq_notify);
> 	}
> 
> Thanks,
>   Roland

Humm...
There are 15 ib_verbs now, and it would be better to ask Hal which 'ib_verb.h'
he is including.

ritafsdev:/opt/svn/openib[11] find . -name 'ib_verbs.h'
./trunk/contrib/voltaire/voltaire-ibhost/src/user-protocols/dapl/dapl-sf-031229/dapl/include/ib/IBM/ke/ib_verbs.h
./trunk/contrib/voltaire/voltaire-ibhost/src/user-protocols/dapl/dapl-sf-031229/dapl/include/ib/IBM/us/ib_verbs.h
./trunk/contrib/intel/ib_verbs.h
./trunk/contrib/infinicon/ALL_HOST/Dat/dapl/dapl/include/ib/IBM/ke/ib_verbs.h
./trunk/contrib/infinicon/ALL_HOST/Dat/dapl/dapl/include/ib/IBM/us/ib_verbs.h
./trunk/contrib/infinicon/latest/ALL_HOST/Dat/dapl/dapl/include/ib/IBM/ke/ib_verbs.h
./trunk/contrib/infinicon/latest/ALL_HOST/Dat/dapl/dapl/include/ib/IBM/us/ib_verbs.h
./trunk/openib/1.0/ib-support/host/ulp/dapl/dapl3/dapl/include/ib/IBM/ke/ib_verbs.h
./trunk/openib/1.0/ib-support/host/ulp/dapl/dapl3/dapl/include/ib/IBM/us/ib_verbs.h
./gen2/branches/openib-candidate/src/linux-kernel/infiniband/include/ib_verbs.h
./gen2/branches/roland-merge/src/linux-kernel/infiniband/include/ib_verbs.h
./tags/infinicon-latest-pre2.6/ALL_HOST/Dat/dapl/dapl/include/ib/IBM/ke/ib_verbs.h
./tags/infinicon-latest-pre2.6/ALL_HOST/Dat/dapl/dapl/include/ib/IBM/us/ib_verbs.h
./tags/infinicon2.1/ALL_HOST/Dat/dapl/dapl/include/ib/IBM/ke/ib_verbs.h
./tags/infinicon2.1/ALL_HOST/Dat/dapl/dapl/include/ib/IBM/us/ib_verbs.h


By the way, why not defining macros like the following instead of static 
inline functions?

#define ib_poll_cq(cq, num_entries, wc) cq->device->poll_cq(cq, num_entries, wc);


Thanks,
Masanori

---
Masanori ITOH  Open Source Software Development Center, NTT DATA CORPORATION
               e-mail: itoumsn at nttdata.co.jp
               phone : +81-3-3523-8122 (ext. 7354)



More information about the general mailing list