Re: [PATCH 0/5] EFI memmap and other fixes, v3

[Toshi Kani <toshi.kani@hp.com>]

On Thu, 2014-01-23 at 09:06 +0000, Matt Fleming wrote:
> On Sat, 18 Jan, at 12:48:13PM, Borislav Petkov wrote:
> > From: Borislav Petkov <bp@suse.de>
> > 
> > Hi all,
> > 
> > this is the latest incarnation which should hopefully work just fine. It
> > tpasses esting at least on all our boxes so we have *some* coverage.
> > 
> > It took us a long time to debug the unmapping path and realize how
> > exactly we're doing the PGD sharing between the kernel and the EFI page
> > table. I've commented on this verbosely in 4/5 for future reference
> > because this stuff is not trivial.
> > 
> > We definitely should have this in mind when we do more changes on how we
> > handle the EFI page table and what we do with it, maybe even decouple it
> > completely from the kernel one. I dunno - it is all future stuff to pay
> > attention to.
> > 
> > Anyway, if anyone with an EFI box wants to give it a run, feel free to
> > do so.
> > 
> > As always, all comments and suggestions are appreciated.
> 
> Thanks Borislav. I picked these up and applied them to my 'next' branch.
> 
> You'll notice that I haven't stuck them in my 'urgent' branch, so these
> won't be sent to Linus before v3.15 despite the fact that they're bug
> fixes. I think Toshi's box is special enough that most people shouldn't
> hit this issue, and since these patches involve rewriting the way we do
> SetVirtualAddressMap() I'm happy for people to hammer on the stuff that
> did make it into the merge window without this added complication.
> 
> Having said that, if people think it's worth the effort I don't mind
> setting up a new branch containing this series that can be sent to Linus
> much sooner.

Since this issue is quite serious on our systems (3.14 kernels won't
boot up), I'd really appreciate if the fixes can be made into some
3.14-rc.  It can hit on any EFI systems with large memory.

Thanks,
-Toshi