Re: [PATCH V4 5/6] arm64: mm: introduce 52-bit userspace support

[Steve Capper <Steve.Capper@arm.com>]

On Thu, Dec 06, 2018 at 02:35:20PM +0000, Suzuki K Poulose wrote:
> 
> 
> On 06/12/2018 12:26, Steve Capper wrote:
> > On Wed, Dec 05, 2018 at 06:22:27PM +0000, Suzuki K Poulose wrote:
> > > Hi Steve,
> > > 
> > [...]
> > > I think we may need a check for the secondary CPUs to make sure that they have
> > > the 52bit support once the boot CPU has decided to use the feature and fail the
> > > CPU bring up (just like we do for the granule support).
> > > 
> > > Suzuki
> > 
> > Hi Suzuki,
> > I have just written a patch to detect a mismatch between 52-bit VA that
> > is being tested now.
> > 
> > As 52-bit kernel VA support is coming in future, the patch checks for a
> > mismatch during the secondary boot path and, if one is found, prevents
> > the secondary from booting (and displays an error message to the user).
> 
> Right now, it is the boot CPU which decides the Userspace 52bit VA, isn't it ?
> Irrespective of the kernel VA support, the userspace must be able to run on
> all the CPUs on the system, right ? So don't we need it now, with this series ?

Hi Suzuki,

Yes the boot CPU determines vabits_user. My idea was to have the
secondary CPUs check to see if vabits_user was 52, and if so, then check
to see if it's capable of supporting 52-bit. If not, then it stops
booting (and sets a flag to indicate why).

This check will be valid for 52-bit userspace support and also valid for
52-bit kernel support (as the check is performed before the secondary
mmu is enabled). I didn't want to write a higher level detection
routine for the userspace support and then have to re-write it later
when introducing 52-bit kernel support.

I'm happy to do what works though, I thought this way was simplest :-).

Cheers,
-- 
Steve