[ofa-general] OFED 1.4's autoconf.h conflicting with kernel

[Liang Zhen]

Hi, could somebody help us on this? We are blocked somehow by this, of
couse we can remove the "autoconf.h" before we build our software
depending on OFED, but I don't think it's the supposed way...

Regards
Liang

Brian J. Murrell wrote:
> On Fri, 2009-01-23 at 11:01 -0800, Jeff Becker wrote:
>   
>> Hi Brian
>>     
>
> Hi (again) Jeff (and everyone else, especially however maintains the
> packaging of /usr/src/ofa_kernel),
>
>   
>> Brian J. Murrell wrote:
>>     
>>> Some research has led me to a message
>>> (http://www.mail-archive.com/general@lists.openfabrics.org/msg18161.html) from Jeff Becker back on Thu, 10 Jul 2008 15:58:53 -0700 in which he submitted a patch to integrate NFSRDMA into OFED 1.4 which is what appears to have brought these changes into OFED 1.4.
>>>       
>
> The more I look at this, the more I'm convinced there is either an angle
> I am completely missing or this is just plainly not the way to do this.
>
> It just cannot work to have two "linux/autoconf.h" files for a third
> (where the first two parties are OFED and the kernel) party module
> build.  There is no guarantee that the third party module won't need to
> query about various CONFIG_ definitions of both the kernel and the OFED
> stack.
>
> The only way I can think of making this work is to "somehow" "unionize"
> these two files (i.e. so there is a single "superset" of them both).
> Perhaps it's doable with some kind of #include_next magic, perhaps not.
>
>   
>> I usually build my kernel first (usually with NFS). Then I build OFED.
>>     
>
> Right.  This is simple enough.  It's when you want to build another
> kernel module that wants the OFED stack that things get sticky.
>
> I realize this is probably not really your area of responsibility and
> this dual autoconf.h problem pre-existed your patch, but your patch has
> really exacerbated the issue by directly conflicting (CONFIG_SYSCTL is
> the particular example I have on hand at the moment) some of the non-IB
> kernel defines.
>
> I'd love to engage whoever is directly responsible for this area of the
> stack but nobody seems to be responding to my queries or the bug which I
> filed yesterday (which is admittedly a short time ago).  I'd try just
> posting a patch to fix it, but I think this needs some discussion on how
> to really achieve the end goal.
>
> Cheers,
> b.
>
>
> _______________________________________________
> general mailing list
> general at lists.openfabrics.org
> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
>
> To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
>