Lustre discuss - May 2006 - EM64T support in pre-built rpms 1.4.5.1 and up

taylorba@iglou.com wrote:
> I actually tried building the kernel, but I had a lot of patches that
> failed (using the rhel-2.4.21 series).  I tried resolving the issues by
> hand, but I kept getting compile errors.  After the fourth or fifth
> failure, I figured I''d done too much hacking to be safe, so I
thought I''d
> post to the list.
> 
> I''ll give it another try, though.  If you have any suggestions,
let me know.

Use our kernel-source rpm, it already has the patches required.
cliffw
> 
> Also, sorry for the double post.  I got a send failure, so I thought it
> didn''t go through.
> 
> Thanks,
> Brock Taylor
> 
> -----Original Message-----
> From: cliff white <cliffw@clusterfs.com>
> To: taylorba@iglou.com
> Date: Wed, 22 Mar 2006 16:40:42 -0800
> Subject: Re: [Lustre-discuss] EM64T support in pre-built rpms 1.4.5.1 and
up
> 
> 
>>taylorba@iglou.com wrote:
>>
>>>I installed version 1.4.5 on an EM64T system with 4GB of memory, and
>>
>>I 
>>
>>>hit the issue where the SWIOTLB option is not set.  I need to run
>>
>>with 
>>
>>>at least 4GB on my client nodes, so turning off memory remap is not 
>>>really an option.
>>>
>>>I tried 1.4.5.1, 1.4.5.2 and 1.4.6 rpms, but each one of these
causes
>>
>>a 
>>
>>>kernel panic at boot saying there''s ''... no EM64T
support in this
>>>kernel ...''
>>>
>>>Is there something obvious that I''ve missed, or are the
1.4.5.1 and
>>
>>up 
>>
>>>rpms built without EM64T (IA32E option) support?
>>>
>>>Any help (especially if it keeps me from rebuilding the kernel) is 
>>>greatly appreciated.
>>
>>Yes, apparently our kernels are being built with some incorrect .config
>>options, we are working on a fix. Are you comfortable re-compiling the 
>>kernel?
>>cliffw
>>
>>
>>>-Brock Taylor
>>>
>>>_______________________________________________
>>>Lustre-discuss mailing list
>>>Lustre-discuss@clusterfs.com
>>>https://mail.clusterfs.com/mailman/listinfo/lustre-discuss
>>
>>
>

taylorba@iglou.com wrote:
> I installed version 1.4.5 on an EM64T system with 4GB of memory, and I 
> hit the issue where the SWIOTLB option is not set.  I need to run with 
> at least 4GB on my client nodes, so turning off memory remap is not 
> really an option.
> 
> I tried 1.4.5.1, 1.4.5.2 and 1.4.6 rpms, but each one of these causes a 
> kernel panic at boot saying there''s ''... no EM64T support
in this
> kernel ...''
> 
> Is there something obvious that I''ve missed, or are the 1.4.5.1
and up
> rpms built without EM64T (IA32E option) support?
> 
> Any help (especially if it keeps me from rebuilding the kernel) is 
> greatly appreciated.
Yes, apparently our kernels are being built with some incorrect .config 
options, we are working on a fix. Are you comfortable re-compiling the 
kernel?
cliffw
> 
> -Brock Taylor
> 
> _______________________________________________
> Lustre-discuss mailing list
> Lustre-discuss@clusterfs.com
> https://mail.clusterfs.com/mailman/listinfo/lustre-discuss

I installed version 1.4.5 on an EM64T system with 4GB of memory, and I 
hit the issue where the SWIOTLB option is not set.  I need to run with 
at least 4GB on my client nodes, so turning off memory remap is not 
really an option.

I tried 1.4.5.1, 1.4.5.2 and 1.4.6 rpms, but each one of these causes a 
kernel panic at boot saying there''s ''... no EM64T support in
this
kernel ...''

Is there something obvious that I''ve missed, or are the 1.4.5.1 and up 
rpms built without EM64T (IA32E option) support?

Any help (especially if it keeps me from rebuilding the kernel) is 
greatly appreciated.

-Brock Taylor