Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Wed, 18 Nov 2020 09:25:47 +0000
From:   Quentin Perret <qperret@google.com>
To:     Rob Herring <robh+dt@kernel.org>
Cc:     Catalin Marinas <catalin.marinas@arm.com>,
        Will Deacon <will@kernel.org>, Marc Zyngier <maz@kernel.org>,
        James Morse <james.morse@arm.com>,
        Julien Thierry <julien.thierry.kdev@gmail.com>,
        Suzuki K Poulose <suzuki.poulose@arm.com>,
        Frank Rowand <frowand.list@gmail.com>,
        "moderated list:ARM64 PORT (AARCH64 ARCHITECTURE)" 
        <linux-arm-kernel@lists.infradead.org>,
        open list <linux-kernel@vger.kernel.org>,
        "open list:KERNEL VIRTUAL MACHINE FOR ARM64 (KVM/arm64)" 
        <kvmarm@lists.cs.columbia.edu>,
        "open list:OPEN FIRMWARE AND FLATTENED DEVICE TREE" 
        <devicetree@vger.kernel.org>,
        Android Kernel Team <kernel-team@android.com>,
        android-kvm@google.com
Subject: Re: [RFC PATCH 15/27] of/fdt: Introduce
 early_init_dt_add_memory_hyp()
Message-ID: <20201118092547.GA2031536@google.com>
References: <20201117181607.1761516-1-qperret@google.com>
 <20201117181607.1761516-16-qperret@google.com>
 <CAL_Jsq+xAy9+HjH6vqfmaAEKBe9MMm+wWvUtiz5dFnHmMneqNw@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAL_Jsq+xAy9+HjH6vqfmaAEKBe9MMm+wWvUtiz5dFnHmMneqNw@mail.gmail.com>
Precedence: bulk

Hi Rob,

On Tuesday 17 Nov 2020 at 13:44:53 (-0600), Rob Herring wrote:
> Can this be done right after we add all the memblocks using the
> memblock API?

Possibly, but the thing I'm a bit worried about is the way 'no-map'
regions are removed from memblocks early on.

The EL2 object needs to know about these parts of memory too (and in
fact we may be able to enforce the 'no-map' attribute at the host stage
2 level as well). It's also possible we'll need to have portions of the
guests payload preloaded (and verified) by the bootloader into reserved
memory regions, possibly no-map, to make sure the host does not mess
with them in a normal use-case. So, I couldn't find a much better place
than this one but suggestions are very much welcome.

I'll have a go at the memblock stuff to see if I find a way to make it
work from that angle.

> I thought EFI would also need to be handled, but looks
> like it just calls early_init_dt_add_memory_arch(). That's odd
> especially for ACPI systems...
> 
> I don't really like putting what looks like an arm64 only hook here,
> but then I don't want an arm64 version of
> early_init_dt_add_memory_arch() either. We're almost to the point of
> getting rid of the arch specific ones. But I don't have a better
> suggestion currently.

Ack, the ugly truth is that this is likely to remain arm64-specific. I
figured this was simple enough that we might want to consider it,
though.

Thanks,
Quentin