[Ocfs2-users] 1.2.4 symbols

[Randy Ramsdell]

Joel Becker wrote:
> On Fri, Feb 09, 2007 at 02:55:37PM -0500, Randy Ramsdell wrote:
>   
>> I found the issue. There are two locations that have the same named
>> modules.
>>
>> /lib/modules/2.6.16.27-0.6-smp/extra/ocfs2/   <--- this is where the
>> default "make install" puts them
>> /lib/modules/2.6.16.27-0.6-smp/kernel/fs/ocfs2/ <--- these are the
>> modules that are called with the init scripts etc ...
>>     
>
> 	Yeah, this is known.
>
>   
>> 2. Remove the offending ocfs2 modules.
>> 3. Manually edit the modules.dep and remove the location of old modules.
>> 4. Or have the "make install" portion store the modules in second
>> location which overwrites the old.
>>     
>
> 	The 'extra' location is a direct result of "make
> modules_install".  That is, it's part of kbuild and can't really be
> changed.
> 	We mostly do 2 or symlink the regular modules to the other
> location.  That's when we're developing.  Otherwise, we use packages
> that put them in the "correct" place.
>
> Joel
>
>   


Ok.  But do different vendors or at least the top three define different
locations?

Looking at the source from the ocfs2 package.

I find:

endor/rhel4/ocfs2.spec-generic:EXTRA_MODLIB="${RPM_BUILD_ROOT}/lib/modules/%{kver}/extra"
vendor/rhel4/ocfs2.spec-generic:EXTRA_MODLIB="${RPM_BUILD_ROOT}/lib/modules/%{kver}smp/extra"
vendor/rhel4/ocfs2.spec-generic:EXTRA_MODLIB="${RPM_BUILD_ROOT}/lib/modules/%{kver}hugemem/extra"
vendor/rhel4/ocfs2.spec-generic:EXTRA_MODLIB="${RPM_BUILD_ROOT}/lib/modules/%{kver}largesmp/extra"

So this looks like it is defined in the ocfs2 source although the vendor
suppile modules are from a kernel compile as you say.

Anyway, I brought the cluster online today so things are good. Thanks
for the help.

rcr