Re: [PULL 18/38] hw/arm/virt: Honor highmem setting when computing the memory map

[Akihiko Odaki]

On Sun, Feb 13, 2022 at 7:46 PM Peter Maydell <peter.maydell@linaro.org> wrote:
>
> On Sun, 13 Feb 2022 at 10:22, Marc Zyngier <maz@kernel.org> wrote:
> >
> > [+ Alex for HVF]
> >
> > On Sun, 13 Feb 2022 05:05:33 +0000,
> > Akihiko Odaki <akihiko.odaki@gmail.com> wrote:
> > > Hi,
> > > This breaks in a case where highmem is disabled but can have more than
> > > 4 GiB of RAM. M1 (Apple Silicon) actually can have 36-bit PA with HVF,
> > > which is not enough for highmem MMIO but is enough to contain 32 GiB
> > > of RAM.
> >
> > Funny. The whole point of this series is to make it all work correctly
> > on M1.
> >
> > > Where the magic number of 4 GiB / 32-bit came from?
> >
> > Not exactly a magic number. From QEMU's docs/system/arm/virt.rst:
> >
> > <quote>
> > highmem
> >   Set ``on``/``off`` to enable/disable placing devices and RAM in physical
> >   address space above 32 bits. The default is ``on`` for machine types
> >   later than ``virt-2.12``.
> > </quote>
> >
> > TL;DR: Removing the bogus 'highmem=off' option from your command-line
> > should get you going with large memory spaces, up to the IPA limit.
>
> Yep. I've tested this with hvf, and we now correctly:
>  * refuse to put RAM above 32-bits if you asked for a 32-bit
>    IPA space with highmem=off
>  * use the full 36-bit address space if you don't say highmem=off
>    on an M1
>
> Note that there is a macos bug where if you don't say highmem=off
> on an M1 Pro then you'll get a macos kernel panic. M1 non-Pro is fine.
>
> thanks
> -- PMM

I found that it actually gets the available PA bit of the emulated CPU
when highmem=on. I used "cortex-a72", which can have more than 36
bits. I just simply switched to "host"; hvf didn't support "host" when
I set up my VM but now it does.
Thanks for your prompt replies.

Regards,
Akihiko Odaki