Messages in this thread | | | From | Matt Fleming <> | Subject | [PATCH 0/2] efi: Remove conditional in_nmi() runtime locking | Date | Wed, 1 Oct 2014 13:06:38 +0100 |
| |
From: Matt Fleming <matt.fleming@intel.com>
These two patches should address the concerns raised by Ingo and Peter in relation to the EFI pull request containing v3.18 material,
https://lkml.kernel.org/r/20140928202702.GB18635@console-pimps.org
We can drop the in_nmi() checks altogether and just provide normal locking semantics if we introduce a non-blocking SetVariable() operation for the particular case of writing pstore data to the EFI backend from the kdump callback, which aborts in the contended case.
This is currently holding up merging of the v3.18 EFI patches, so please be timely with comments, if any.
@Linaro guys, I don't think you'll actually care all that much about these changes since the in_nmi() goo was for x86's benefit, but I'm Cc'ing you anyway to make sure everything looks OK.
Matt Fleming (2): efi: Provide a non-blocking SetVariable() operation efi: Delete the in_nmi() conditional runtime locking
arch/x86/include/asm/efi.h | 2 -- drivers/firmware/efi/runtime-wrappers.c | 36 ++++++++++++++++--------- drivers/firmware/efi/vars.c | 47 +++++++++++++++++++++++++++++++++ include/linux/efi.h | 6 +++++ 4 files changed, 76 insertions(+), 15 deletions(-)
-- 1.9.3
| |