Re: [PATCH v2 1/7] arm64: Unconditionally call unflatten_device_tree()

From: Rob Herring <robh@kernel.org>
To: Stephen Boyd <sboyd@kernel.org>
Cc: linux-kernel@vger.kernel.org, patches@lists.linux.dev,
	linux-um@lists.infradead.org,
	linux-arm-kernel@lists.infradead.org, kunit-dev@googlegroups.com,
	linux-kselftest@vger.kernel.org, devicetree@vger.kernel.org,
	Frank Rowand <frowand.list@gmail.com>,
	Catalin Marinas <catalin.marinas@arm.com>,
	Will Deacon <will@kernel.org>,
	Mark Rutland <mark.rutland@arm.com>
Subject: Re: [PATCH v2 1/7] arm64: Unconditionally call unflatten_device_tree()
Date: Fri, 2 Feb 2024 10:30:06 -0600	[thread overview]
Message-ID: <20240202163006.GA209874-robh@kernel.org> (raw)
In-Reply-To: <efe6a7886c3491cc9c225a903efa2b1e.sboyd@kernel.org>

On Wed, Jan 31, 2024 at 02:59:53PM -0800, Stephen Boyd wrote:
> Quoting Rob Herring (2024-01-31 12:54:05)
> > On Mon, Jan 29, 2024 at 04:45:00PM -0800, Stephen Boyd wrote:
> > > Call this function unconditionally so that we can populate an empty DTB
> > > on platforms that don't boot with a firmware provided or builtin DTB.
> > > Override 'initial_boot_params' to NULL when ACPI is in use but the
> > > bootloader has loaded a DTB so that we don't allow both ACPI and DT to
> > > be used during boot. If there isn't a valid initial_boot_params dtb then
> > > unflatten_device_tree() returns early so this is fine.
> > > 
> > > Cc: Rob Herring <robh+dt@kernel.org>
> > > Cc: Frank Rowand <frowand.list@gmail.com>
> > > Cc: Catalin Marinas <catalin.marinas@arm.com>
> > > Cc: Will Deacon <will@kernel.org>
> > > Cc: Mark Rutland <mark.rutland@arm.com>
> > > Cc: <linux-arm-kernel@lists.infradead.org>
> > > Signed-off-by: Stephen Boyd <sboyd@kernel.org>
> > > ---
> > >  arch/arm64/kernel/setup.c | 7 +++++--
> > >  1 file changed, 5 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/arch/arm64/kernel/setup.c b/arch/arm64/kernel/setup.c
> > > index 417a8a86b2db..ffb1942724ae 100644
> > > --- a/arch/arm64/kernel/setup.c
> > > +++ b/arch/arm64/kernel/setup.c
> > > @@ -351,8 +351,11 @@ void __init __no_sanitize_address setup_arch(char **cmdline_p)
> > >       /* Parse the ACPI tables for possible boot-time configuration */
> > >       acpi_boot_table_init();
> > >  
> > > -     if (acpi_disabled)
> > > -             unflatten_device_tree();
> > > +     /* Don't use the FDT from boot if ACPI is in use */
> > > +     if (!acpi_disabled)
> > > +             initial_boot_params = NULL;
> > 
> > I still think this is a problem for kexec. See 
> > of_kexec_alloc_and_setup_fdt(). You see it uses initial_boot_params. At 
> > first glance it looks like it would just write out everything we need. 
> > But for UEFI boot, I think we need all the chosen properties like 
> > linux,uefi-mmap-start preserved from the current boot for the next 
> > kernel we kexec.
> 
> Ok, got it.
> 
> > 
> > I think you'll have to check acpi_disabled in unflatten_device_tree() 
> > and unflatten the empty tree leaving initial_boot_params alone. That 
> > means our FDT and unflattened tree will be different DTs, but I think 
> > that's fine.
> 
> It's sort of scary given that 'initial_boot_params' is an exported
> global. Maybe that should be hidden away and accessed with a function
> instead so that this mismatch doesn't break something later on?

What you see now is after I did a pass of minimizing the number of 
users.

BTW, I've now got another use for unconditionally calling 
unflatten_device_tree() with this[1].

Rob

[1] https://lore.kernel.org/all/20240201194653.GA1328565-robh@kernel.org/