[openfabrics-ewg] Re: ipath updates in svn

Vladimir Sokolovsky vlad at mellanox.co.il
Tue May 23 07:47:11 PDT 2006


Bryan,
modprobe ipath_core still fails on SLES 10 Beta 10:

ipath_core: exports duplicate symbol __iowrite32_copy (owned by kernel)

Kernel: *2.6.16-20-smp*

Regards,
Vladimir

Bryan O'Sullivan wrote:
> On Thu, 2006-05-18 at 14:57 +0300, Tziporet Koren wrote:
>
>   
>> We have build a new version based on RC5 in ofed and these are the results:
>> ipath driver compilation passed on RH EL4 up2 & up3 and FC4.
>>     
>
> Thanks for testing.
>
>   
>> On SLES 10 beta 10 it pass compilation but fails when we try to load it 
>> with:
>> ipath_core: exports duplicate symbol __iowrite32_copy (owned by kernel)
>> ib_ipath: module not supported by Novell, setting U taint flag.
>> ib_ipath: Unknown symbol ipath_layer_get_deviceid
>> ib_ipath: Unknown symbol ipath_layer_get_flags
>> ib_ipath: Unknown symbol ipath_layer_snapshot_counters
>> ib_ipath: Unknown symbol ipath_layer_get_ibmtu
>> ib_ipath: Unknown symbol ipath_layer_get_cr_errpkey
>> ....
>>     
>
> That's strange.  The ipath_layer.c file includes
> ipath_layer_snapshot_counters.  Is ipath_core loaded when you're trying
> to modprobe ib_ipath?
>
>   
>> In file included from 
>> /var/tmp/OFED/tmp/openib/openib/src/linux-kernel/infiniband/hw/ipath/ipath_verbs.h:42,
>> from 
>> /var/tmp/OFED/tmp/openib/openib/src/linux-kernel/infiniband/hw/ipath/ipath_cq.c:36:
>> /var/tmp/OFED/tmp/openib/openib/src/linux-kernel/infiniband/hw/ipath/ipath_backport.h: 
>> In function ‘__iowrite32_copy’:
>> /var/tmp/OFED/tmp/openib/openib/src/linux-kernel/infiniband/hw/ipath/ipath_backport.h:101: 
>> error: implicit declaration of function ‘__raw_writel’
>>     
>
> There's no mention of __raw_writel in the version of ipath_backport.h
> that's in the ipath_backport.patch file in SVN.  Instead, the
> __iowrite32_copy routine is out of line in the file iomem_copy.c.  I
> think you must have an older version of the ipath_backport.h file
> somewhere that is getting included, but which should not be.
>
> 	<b
>
> _______________________________________________
> openfabrics-ewg mailing list
> openfabrics-ewg at openib.org
> http://openib.org/mailman/listinfo/openfabrics-ewg
>
>   




More information about the ewg mailing list