SRU request for LP#239666

Wang, Yong Y yong.y.wang at intel.com
Fri Jun 20 02:49:04 UTC 2008


> -----Original Message-----
> From: Tim Gardner [mailto:tim.gardner at canonical.com] 
> Sent: 2008年6月20日 10:37
> To: Wang, Yong Y
> Cc: kernel-team at lists.ubuntu.com
> Subject: Re: SRU request for LP#239666
> 
> Wang, Yong Y wrote:
> >> -----Original Message-----
> >> From: Tim Gardner [mailto:tim.gardner at canonical.com]
> >> Sent: 2008年6月20日 1:28
> >> To: Wang, Yong Y
> >> Cc: kernel-team at lists.ubuntu.com
> >> Subject: Re: SRU request for LP#239666
> >>
> >> Wang, Yong Y wrote:
> >>> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/239666
> >>>
> >>> SRU justification:
> >>>
> >>> Impact: Current 64 bit MCE reporting code in Hardy 2.6.24 kernel
> >>> hardcodes bank limitation as 6 and this is not an issue for
> >> processors
> >>> with <=6 MCE banks. But for latest processors, e.g. 
> Dunnington, with
> >>> more than 6 MCE banks, additional MCE banks available in processor
> >>> hardware are not enabled to record hardware error
> >> information which is
> >>> extremely helpful for diagnostics.
> >>>
> >>> For those who do not know what MCA & MCE are, here are some nice
> >>> documents:
> >>> 1. http://one.firstfloor.org/~andi/mce.pdf
> >>> 2. kernel-source-dir/Documentation/x86_64/machinecheck
> >>>
> >>> Original patch can be found at http://lkml.org/lkml/2008/5/8/412.
> >>>
> >>> Fix is available below:
> >>>
> >>> The following changes since commit
> >>> cafcac43a64ef93b5929ec8d88103761862474d2:
> >>>   Amit Kucheria (1):
> >>>         UBUNTU: LPIA: More conservative culling of LPIA config
> >>>
> >>> are available in the git repository at:
> >>>
> >>>   git://kernel.ubuntu.com/ywang/ubuntu-hardy.git master
> >>>
> >>> Yong Wang (1):
> >>>       UBUNTU: x86: Remove 6 bank limitation in 64 bit MCE
> >> reporting code
> >>>  arch/x86/kernel/cpu/mcheck/mce_64.c |   20 ++++++++++++--------
> >>>  1 files changed, 12 insertions(+), 8 deletions(-)
> >>>
> >>> -Yong
> >>>
> >> I notice that this disappeared from your git repository. I was also
> >> reviewing the upstream discussion, and I don't think this issue is
> >> completely resolved.
> >>
> > 
> > Oh, it is now in 
> git://kernel.ubuntu.com/ywang/ubuntu-hardy-mca.git master.
> > I met some problems when working on my git tree as 
> indicated by Amit's email.
> > What part of the issue do you think is not completely resolved yet?
> > 
> > -Yong 
> > 
> 
> Peter Anvin's comment in that thread was that he would like to see a
> symmetrical implementation, e.g., support in mce_32.c as well 
> as mce_64.c.
> 
> Even though Ingo queued the 64 bit patch, it has yet to land in Linus'
> tree, so I would like to wait and see what the final patch looks like.
> 

mce_32.c was written by Alan Cox and it does not have the hardcoded limitation problem.
However, mce_64.c was re-ritten by Andi Kleen several years ago and that introduced the
problem we are trying to solve here. I don't think there will be a symmetrical implementation
because the 32  bit code needs to handle a lot of legacy stuff.

-Yong




More information about the kernel-team mailing list