Re: [PATCH v4 2/2] arm64: Don't read GMID_EL1 when MTE is disabled

[Will Deacon <will@kernel.org>]

On Tue, Aug 25, 2026 at 05:42:19PM +0100, Fuad Tabba wrote:
> __cpuinfo_store_cpu() gates the GMID_EL1 read on the raw
> ID_AA64PFR1_EL1, so it reads the register even when the kernel has
> disabled MTE (CONFIG_ARM64_MTE=n or arm64.nomte). KVM sets HCR_EL2.TID5
> in that case, and pKVM injects an UNDEF the host cannot handle:
> 
>   Internal error: Oops - Undefined instruction: 0000000002000000 [#1]  SMP
>   pc : __cpuinfo_store_cpu+0xf4/0x264
>   Kernel panic - not syncing: Attempted to kill the idle task!
> 
> Only pKVM reaches it, and only after a CPU is offlined and brought back
> online: its CPU_ON relay sets the host HCR before the CPU enters EL1,
> while plain nVHE sets it at CPUHP_AP_KVM_ONLINE.
> 
> Gate the read on __read_sysreg_by_encoding(), which applies the cmdline
> override, and on CONFIG_ARM64_MTE, which no register reflects.
> 
> Fixes: f35abcbb8a084 ("KVM: arm64: Trap MTE access and discovery when MTE is disabled")
> Cc: stable@vger.kernel.org # needs "arm64: Apply overrides to CPU local capabilities"
> Signed-off-by: Fuad Tabba <fuad.tabba@linux.dev>
> ---
>  arch/arm64/include/asm/cpufeature.h | 9 +++++++++
>  arch/arm64/kernel/cpufeature.c      | 6 ++----
>  arch/arm64/kernel/cpuinfo.c         | 2 +-
>  3 files changed, 12 insertions(+), 5 deletions(-)
> 
> diff --git a/arch/arm64/include/asm/cpufeature.h b/arch/arm64/include/asm/cpufeature.h
> index a57870fa96db5..0b374e938f708 100644
> --- a/arch/arm64/include/asm/cpufeature.h
> +++ b/arch/arm64/include/asm/cpufeature.h
> @@ -1085,6 +1085,15 @@ static inline bool cpu_has_lpa2(void)
>  #endif
>  }
>  
> +/* No ID register reflects CONFIG_ARM64_MTE. */
> +static inline bool gmid_el1_accessible(void)
> +{
> +	if (!IS_ENABLED(CONFIG_ARM64_MTE))
> +		return false;
> +
> +	return id_aa64pfr1_mte(__read_sysreg_by_encoding(SYS_ID_AA64PFR1_EL1));

I'm planning to internalise __read_sysreg_by_encoding() into cpufeature.c
so I'd prefer to avoid adding another user of it, if possible. In this case,
__cpuinfo_store_cpu() has already read the thing, so all it needs is the
override logic (but see below).

> +}
> +
>  #endif /* __ASSEMBLER__ */
>  
>  #endif
> diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
> index 88b15b5ef2f7f..23f174b3c4e26 100644
> --- a/arch/arm64/kernel/cpufeature.c
> +++ b/arch/arm64/kernel/cpufeature.c
> @@ -1228,7 +1228,7 @@ void __init init_cpu_features(struct cpuinfo_arm64 *info)
>  		init_cpu_ftr_reg(SYS_MPAMIDR_EL1, info->reg_mpamidr);
>  	}
>  
> -	if (id_aa64pfr1_mte(info->reg_id_aa64pfr1))
> +	if (gmid_el1_accessible())
>  		init_cpu_ftr_reg(SYS_GMID_EL1, info->reg_gmid);
>  }
>  
> @@ -1523,11 +1523,9 @@ void update_cpu_features(int cpu,
>  	 * they read/write depends on the GMID_EL1.BS field. Check that the
>  	 * value is the same on all CPUs.
>  	 */
> -	if (IS_ENABLED(CONFIG_ARM64_MTE) &&
> -	    id_aa64pfr1_mte(info->reg_id_aa64pfr1)) {
> +	if (gmid_el1_accessible())
>  		taint |= check_update_ftr_reg(SYS_GMID_EL1, cpu,
>  					      info->reg_gmid, boot->reg_gmid);
> -	}
>  
>  	/*
>  	 * If we don't have AArch32 at all then skip the checks entirely
> diff --git a/arch/arm64/kernel/cpuinfo.c b/arch/arm64/kernel/cpuinfo.c
> index d50e2a9b066b3..389fca84f106b 100644
> --- a/arch/arm64/kernel/cpuinfo.c
> +++ b/arch/arm64/kernel/cpuinfo.c
> @@ -502,7 +502,7 @@ static void __cpuinfo_store_cpu(struct cpuinfo_arm64 *info)
>  	info->reg_id_aa64smfr0 = read_cpuid(ID_AA64SMFR0_EL1);
>  	info->reg_id_aa64fpfr0 = read_cpuid(ID_AA64FPFR0_EL1);
>  
> -	if (id_aa64pfr1_mte(info->reg_id_aa64pfr1))
> +	if (gmid_el1_accessible())
>  		info->reg_gmid = read_cpuid(GMID_EL1);

I'm not sure this is safe. For the boot CPU, __cpuinfo_store_cpu() is
called before init_cpu_features(), so the arm64_ftr_regs[] array hasn't
been sorted and we can't call get_arm64_ftr_reg() reliably. In fact, it
doesn't even look like the overrides will have been processed.

So we're in a bit of a chicken-and-egg problem here. Perhaps we need an
early (__init) function that can apply the override manually to the
value read from hardware using arm64_ftr_safe_value(). You'll probably
need something like my old hack [1] to avoid the bsearch though :/

Will

[1] https://lore.kernel.org/all/20230110161651.GB9436@willie-the-truck/