Re: [PATCH v1] arm64/mm: Fix Boot panic on Ampere Altra

[Will Deacon <will@kernel.org>]

On Tue, Feb 25, 2025 at 07:05:35PM +0100, Ard Biesheuvel wrote:
> Apologies for the breakage, and thanks for the fix.
> 
> I have to admit that I was a bit overzealous here: there is no point
> yet in using the sanitised value, given that we don't actually
> override the PA range in the first place. This is something I've
> prototyped for Android use, so that linear map randomization can be
> force enabled on CPUs with a wide PArange, but right now, mainline
> does not have that capability, and so I'd be inclined to just revert
> the hunk that introduces the call to read_sanitised_ftr_reg() into
> arm64_memblock_init(), especially given the fact that commit
> 62cffa496aac was tagged for stable, and was already pulled into 6.13
> and 6.12
> 
> In any case, it would be good if we could get a fix into Linus's tree asap

Makes sense. So the patch below?

Will

--->8

From b76ddd40dd6fe350727a4b2ec50709fd919d8408 Mon Sep 17 00:00:00 2001
From: Ryan Roberts <ryan.roberts@arm.com>
Date: Tue, 25 Feb 2025 11:46:36 +0000
Subject: [PATCH] arm64/mm: Fix Boot panic on Ampere Altra

When the range of present physical memory is sufficiently small enough
and the reserved address space for the linear map is sufficiently large
enough, The linear map base address is randomized in
arm64_memblock_init().

Prior to commit 62cffa496aac ("arm64/mm: Override PARange for !LPA2 and
use it consistently"), we decided if the sizes were suitable with the
help of the raw mmfr0.parange. But the commit changed this to use the
sanitized version instead. But the function runs before the register has
been sanitized so this returns 0, interpreted as a parange of 32 bits.
Some fun wrapping occurs and the logic concludes that there is enough
room to randomize the linear map base address, when really there isn't.
So the top of the linear map ends up outside the reserved address space.

Since the PA range cannot be overridden in the first place, restore the
mmfr0 reading logic to its state prior to 62cffa496aac, where the raw
register value is used.

Reported-by: Luiz Capitulino <luizcap@redhat.com>
Suggested-by: Ard Biesheuvel <ardb@kernel.org>
Closes: https://lore.kernel.org/all/a3d9acbe-07c2-43b6-9ba9-a7585f770e83@redhat.com/
Fixes: 62cffa496aac ("arm64/mm: Override PARange for !LPA2 and use it consistently")
Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>
Link: https://lore.kernel.org/r/20250225114638.2038006-1-ryan.roberts@arm.com
Signed-off-by: Will Deacon <will@kernel.org>
---
 arch/arm64/mm/init.c | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)

diff --git a/arch/arm64/mm/init.c b/arch/arm64/mm/init.c
index 9c0b8d9558fc..ccdef53872a0 100644
--- a/arch/arm64/mm/init.c
+++ b/arch/arm64/mm/init.c
@@ -279,12 +279,7 @@ void __init arm64_memblock_init(void)
 
 	if (IS_ENABLED(CONFIG_RANDOMIZE_BASE)) {
 		extern u16 memstart_offset_seed;
-
-		/*
-		 * Use the sanitised version of id_aa64mmfr0_el1 so that linear
-		 * map randomization can be enabled by shrinking the IPA space.
-		 */
-		u64 mmfr0 = read_sanitised_ftr_reg(SYS_ID_AA64MMFR0_EL1);
+		u64 mmfr0 = read_cpuid(ID_AA64MMFR0_EL1);
 		int parange = cpuid_feature_extract_unsigned_field(
 					mmfr0, ID_AA64MMFR0_EL1_PARANGE_SHIFT);
 		s64 range = linear_region_size -
-- 
2.48.1.658.g4767266eb4-goog