CentOS - Apr 2011 - FYI - pci_mmcfg_init kernel error after upgrading to 5.6

Just thought I'd pass along an error I received this weekend after 
upgrading two servers to 5.6.  Both are HP Proliant DL380 servers 
running 64-bit, but one is a G6 model and one is a G7.  After the 
upgrade and the reboot to the 2.6.18-238.9.1.el5 kernel, both servers 
displayed the following error at boot:
     "pci_mmcfg_init marking 256MB space uncacheable."

Some Googling found me the following link:
     https://partner-bugzilla.redhat.com/show_bug.cgi?id=581933

There are apparently performance implications due to this issue. 
Fortunately, the solution is simple:  just boot with the kernel parameter
     acpi_mcfg_max_pci_bus_num=on

If everyone is already familiar with this, please ignore. :-)

-- 
Best Regards,

Dave Windsor

Robert Bosch LLC
Team Leader, MES Database Infrastructure Group (AdP/TEF7.1)
4421 Highway 81 North
Anderson, SC 29621 USA
www.bosch.us

Tel: 1 (864) 260-8459
Fax: 1 (864) 260-8422
Dave.Windsor at us.bosch.com

On 04/25/2011 08:58 AM, Windsor Dave L. (AdP/TEF7.1)
wrote:
> Just thought I'd pass along an error I received this weekend after 
> upgrading two servers to 5.6.  Both are HP Proliant DL380 servers 
> running 64-bit, but one is a G6 model and one is a G7.  After the 
> upgrade and the reboot to the 2.6.18-238.9.1.el5 kernel, both servers 
> displayed the following error at boot:
>      "pci_mmcfg_init marking 256MB space uncacheable."
> 
> Some Googling found me the following link:
>      https://partner-bugzilla.redhat.com/show_bug.cgi?id=581933
> 
> There are apparently performance implications due to this issue. 
> Fortunately, the solution is simple:  just boot with the kernel parameter
>      acpi_mcfg_max_pci_bus_num=on
> 
> If everyone is already familiar with this, please ignore. :-)
> 
Supposedly this issue is fixed in the 5.6 kernel ... not introduced in it :D

https://bugzilla.redhat.com/show_bug.cgi?id=581933

However, the advise is good regardless of version of kernel.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 253 bytes
Desc: OpenPGP digital signature
URL:
<http://lists.centos.org/pipermail/centos/attachments/20110425/42531a30/attachment-0005.sig>

On 4/25/2011 10:26 AM, Johnny Hughes wrote:
> On 04/25/2011 08:58 AM, Windsor Dave L. (AdP/TEF7.1) wrote:
>> Just thought I'd pass along an error I received this weekend after
>> upgrading two servers to 5.6.  Both are HP Proliant DL380 servers
>> running 64-bit, but one is a G6 model and one is a G7.  After the
>> upgrade and the reboot to the 2.6.18-238.9.1.el5 kernel, both servers
>> displayed the following error at boot:
>>       "pci_mmcfg_init marking 256MB space uncacheable."
>>
>> Some Googling found me the following link:
>>       https://partner-bugzilla.redhat.com/show_bug.cgi?id=581933
>>
>> There are apparently performance implications due to this issue.
>> Fortunately, the solution is simple:  just boot with the kernel
parameter
>>       acpi_mcfg_max_pci_bus_num=on
>>
>> If everyone is already familiar with this, please ignore. :-)
>>
>
> Supposedly this issue is fixed in the 5.6 kernel ... not introduced in it
:D
>
> https://bugzilla.redhat.com/show_bug.cgi?id=581933
>
> However, the advise is good regardless of version of kernel.
>

Yes, I saw the reference to RHEL 5.5 in the bug description, but I never 
saw the message before the upgrade to CentOS 5.6.  Hardware differences, 
I guess.  The bug description mentions a Dell Precision.

Best Regards,

Dave Windsor

Robert Bosch LLC
Team Leader, MES Database Infrastructure Group (AdP/TEF7.1)
4421 Highway 81 North
Anderson, SC 29621 USA
www.bosch.us

Hello Johnny,

On Mon, 2011-04-25 at 09:26 -0500, Johnny Hughes wrote:
> On 04/25/2011 08:58 AM, Windsor Dave L. (AdP/TEF7.1) wrote:
> > There are apparently performance implications due to this issue. 
> > Fortunately, the solution is simple:  just boot with the kernel
parameter
> >      acpi_mcfg_max_pci_bus_num=on
> Supposedly this issue is fixed in the 5.6 kernel ... not introduced in it
:D
> 
> https://bugzilla.redhat.com/show_bug.cgi?id=581933
> 
> However, the advise is good regardless of version of kernel.
Not exactly. That kernel parameter *is* the fix.

Regards,
Leonard.

-- 
mount -t life -o ro /dev/dna /genetic/research