[Maverick] [PATCH 0/6] xsaveopt instruction support

Tim Gardner tim.gardner at canonical.com
Tue Aug 24 02:31:16 UTC 2010


On 08/23/2010 07:58 PM, leann.ogasawara at canonical.com wrote:
> It's been requested that we provide xsaveopt instruction support in
> Maverick which can be leveraged by Intel Sandy Bridge.  This is a new
> feature to improve performance of the XSAVE operation by reducing the
> amount of data written during an XSAVE operation.  All patches are
> currently upstream in 2.6.35-rc1.  Proposing these for Maverick.
>
> Thanks,
> Leann
>
> Suresh Siddha (6):
>    x86, cpu: Make init_scattered_cpuid_features() consider cpuid
>      subleaves
>    x86, cpu: Add xsaveopt cpufeature
>    x86, cpu: Enumerate xsaveopt
>    x86, xsave: Track the offset, size of state in the xsave layout
>    x86, xsave: Sync xsave memory layout with its header for user
>      handling
>    x86, xsave: Use xsaveopt in context-switch path when supported
>
>   arch/x86/include/asm/cpufeature.h          |    1 +
>   arch/x86/include/asm/i387.h                |   14 ++++
>   arch/x86/include/asm/xsave.h               |   19 ++++-
>   arch/x86/kernel/cpu/addon_cpuid_features.c |   24 +++---
>   arch/x86/kernel/cpu/common.c               |    8 ++
>   arch/x86/kernel/i387.c                     |   11 +++
>   arch/x86/kernel/xsave.c                    |  116 ++++++++++++++++++++++++++++
>   7 files changed, 179 insertions(+), 14 deletions(-)
>
>

So, how do we know this actually works? I can't see any way to tell if 
the xsave feature is actually in use. Given the low level nature of this 
patch its pretty hard to evaluate. I see the first 3 patches were 
backported. Is there any question about them being correct, or were they 
simple collisions?

rtg
-- 
Tim Gardner tim.gardner at canonical.com




More information about the kernel-team mailing list