[PATCH v4 0/2] arm64: Clamp ID overrides and gate the GMID_EL1 read

[PATCH v4 0/2] arm64: Clamp ID overrides and gate the GMID_EL1 read

[Fuad Tabba]

Hi folks,

Changes since v3 [1]:
  - Gate on __read_sysreg_by_encoding() and drop the helper's argument,
    so the command-line override is not open-coded in the header.
    (Catalin)
  - Pick up Suzuki's override clamp as patch 1, rather than adding a
    raw-presence check to the gate. (Will)

Patch 2 fixes a GMID_EL1 read that traps to EL2 when the kernel has
disabled MTE, panicking the host under pKVM on a CPU that is offlined
and brought back online. It gates the read on
__read_sysreg_by_encoding().

That needs patch 1, Suzuki's override clamp [2]. Without it
__read_sysreg_by_encoding() applies the command-line override raw, so
id_aa64pfr1.mte=2 on a CPU without FEAT_MTE2 makes the gate true and the
kernel reads a GMID_EL1 that is not there. On the boot CPU that read
comes before init_cpu_ftr_reg() strips the unsafe override, so the gate
cannot rely on it.

Tested on QEMU under pKVM. With -machine virt,mte=on, offlining and
onlining CPU1 with arm64.nomte panics an unpatched v7.2 in
__cpuinfo_store_cpu() and does not with the series. Without tag memory,
where QEMU implements no GMID_EL1, id_aa64pfr1.mte=2 covers the
force-up case.

Based on Linux 7.2 (8d3ae59288f1e).

Cheers,
/fuad

[1] https://lore.kernel.org/all/20260824184155.2644646-1-fuad.tabba@linux.dev/
[2] https://lore.kernel.org/all/afc5bd00-28ca-413b-b047-ee53589c285d@arm.com/

Fuad Tabba (1):
  arm64: Don't read GMID_EL1 when MTE is disabled

Suzuki K Poulose (1):
  arm64: Apply overrides to CPU local capabilities

 arch/arm64/include/asm/cpufeature.h |  9 ++++++
 arch/arm64/kernel/cpufeature.c      | 48 +++++++++++++++++++++--------
 arch/arm64/kernel/cpuinfo.c         |  2 +-
 3 files changed, 46 insertions(+), 13 deletions(-)

-- 
2.39.5

[PATCH v4 1/2] arm64: Apply overrides to CPU local capabilities

[Fuad Tabba]

From: Suzuki K Poulose <suzuki.poulose@arm.com>

If an override has been applied, make sure we apply that for the
secondary CPUs too, to limit the features.

Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lore.kernel.org/all/afc5bd00-28ca-413b-b047-ee53589c285d@arm.com/
Cc: stable@vger.kernel.org
[Fuad: whitespace and comment fixes]
Signed-off-by: Fuad Tabba <fuad.tabba@linux.dev>
---
 arch/arm64/kernel/cpufeature.c | 42 +++++++++++++++++++++++++++-------
 1 file changed, 34 insertions(+), 8 deletions(-)

diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
index 9a22df0c5120f..88b15b5ef2f7f 100644
--- a/arch/arm64/kernel/cpufeature.c
+++ b/arch/arm64/kernel/cpufeature.c
@@ -1232,10 +1232,43 @@ void __init init_cpu_features(struct cpuinfo_arm64 *info)
 		init_cpu_ftr_reg(SYS_GMID_EL1, info->reg_gmid);
 }
 
+/*
+ * Sanitise the register fields to clamp the values to the overrides that
+ * have been applied.
+ */
+static u64 override_cpu_ftr_reg(struct arm64_ftr_reg *reg, u64 val)
+{
+	const struct arm64_ftr_bits *ftrp;
+
+	if (!reg || !reg->override->mask)
+		return val;
+
+	for (ftrp = reg->ftr_bits; ftrp->width; ftrp++) {
+		u64 ftr_mask = arm64_ftr_mask(ftrp);
+		s64 ftr_val, ftr_ovr, ftr_safe;
+
+		/* Skip the fields not overridden */
+		if ((ftr_mask & reg->override->mask) != ftr_mask)
+			continue;
+
+		ftr_val = arm64_ftr_value(ftrp, val);
+		ftr_ovr = arm64_ftr_value(ftrp, reg->override->val);
+		ftr_safe = arm64_ftr_safe_value(ftrp, ftr_ovr, ftr_val);
+
+		if (ftr_safe != ftr_val)
+			val = arm64_ftr_set_value(ftrp, val, ftr_safe);
+	}
+
+	return val;
+}
+
 static void update_cpu_ftr_reg(struct arm64_ftr_reg *reg, u64 new)
 {
 	const struct arm64_ftr_bits *ftrp;
 
+	/* Apply the overrides */
+	new = override_cpu_ftr_reg(reg, new);
+
 	for (ftrp = reg->ftr_bits; ftrp->width; ftrp++) {
 		s64 ftr_cur = arm64_ftr_value(ftrp, reg->sys_val);
 		s64 ftr_new = arm64_ftr_value(ftrp, new);
@@ -1539,7 +1572,6 @@ EXPORT_SYMBOL_GPL(read_sanitised_ftr_reg);
  */
 u64 __read_sysreg_by_encoding(u32 sys_id)
 {
-	struct arm64_ftr_reg *regp;
 	u64 val;
 
 	switch (sys_id) {
@@ -1592,13 +1624,7 @@ u64 __read_sysreg_by_encoding(u32 sys_id)
 		return 0;
 	}
 
-	regp  = get_arm64_ftr_reg(sys_id);
-	if (regp) {
-		val &= ~regp->override->mask;
-		val |= (regp->override->val & regp->override->mask);
-	}
-
-	return val;
+	return override_cpu_ftr_reg(get_arm64_ftr_reg(sys_id), val);
 }
 
 #include <linux/irqchip/arm-gic-v3.h>
-- 
2.39.5

Re: [PATCH v4 1/2] arm64: Apply overrides to CPU local capabilities

[Catalin Marinas]

On Tue, Aug 25, 2026 at 05:42:18PM +0100, Fuad Tabba wrote:
> From: Suzuki K Poulose <suzuki.poulose@arm.com>
> 
> If an override has been applied, make sure we apply that for the
> secondary CPUs too, to limit the features.
> 
> Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
> Link: https://lore.kernel.org/all/afc5bd00-28ca-413b-b047-ee53589c285d@arm.com/
> Cc: stable@vger.kernel.org
> [Fuad: whitespace and comment fixes]
> Signed-off-by: Fuad Tabba <fuad.tabba@linux.dev>

For completeness, we should add:

Fixes: b3341ae0efa2  ("arm64: cpufeature: Use IDreg override in __read_sysreg_by_encoding()")

> diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
> index 9a22df0c5120f..88b15b5ef2f7f 100644
> --- a/arch/arm64/kernel/cpufeature.c
> +++ b/arch/arm64/kernel/cpufeature.c
> @@ -1232,10 +1232,43 @@ void __init init_cpu_features(struct cpuinfo_arm64 *info)
>  		init_cpu_ftr_reg(SYS_GMID_EL1, info->reg_gmid);
>  }
>  
> +/*
> + * Sanitise the register fields to clamp the values to the overrides that
> + * have been applied.
> + */
> +static u64 override_cpu_ftr_reg(struct arm64_ftr_reg *reg, u64 val)
> +{
> +	const struct arm64_ftr_bits *ftrp;
> +
> +	if (!reg || !reg->override->mask)
> +		return val;
> +
> +	for (ftrp = reg->ftr_bits; ftrp->width; ftrp++) {
> +		u64 ftr_mask = arm64_ftr_mask(ftrp);
> +		s64 ftr_val, ftr_ovr, ftr_safe;
> +
> +		/* Skip the fields not overridden */
> +		if ((ftr_mask & reg->override->mask) != ftr_mask)
> +			continue;
> +
> +		ftr_val = arm64_ftr_value(ftrp, val);
> +		ftr_ovr = arm64_ftr_value(ftrp, reg->override->val);
> +		ftr_safe = arm64_ftr_safe_value(ftrp, ftr_ovr, ftr_val);
> +
> +		if (ftr_safe != ftr_val)
> +			val = arm64_ftr_set_value(ftrp, val, ftr_safe);
> +	}
> +
> +	return val;
> +}
> +
>  static void update_cpu_ftr_reg(struct arm64_ftr_reg *reg, u64 new)
>  {
>  	const struct arm64_ftr_bits *ftrp;
>  
> +	/* Apply the overrides */
> +	new = override_cpu_ftr_reg(reg, new);

Not sure we need this. The init value has already been clamped, made
safe, so it won't change the result.

Otherwise:

Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>

Re: [PATCH v4 1/2] arm64: Apply overrides to CPU local capabilities

[Fuad Tabba]

Hi Catalin,

On Wed, 26 Aug 2026 at 14:47, Catalin Marinas <catalin.marinas@arm.com> wrote:
...
> For completeness, we should add:
>
> Fixes: b3341ae0efa2  ("arm64: cpufeature: Use IDreg override in __read_sysreg_by_encoding()")

Ack

...
> > +     /* Apply the overrides */
> > +     new = override_cpu_ftr_reg(reg, new);
>
> Not sure we need this. The init value has already been clamped, made
> safe, so it won't change the result.

I'll fix both on the respin. I'll wait to see if Will or anyone else
has further comments before doing that.

> Otherwise:
>
> Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>

Thanks for the reviews!
/fuad

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

[Fuad Tabba]

__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));
+}
+
 #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);
 
 	if (id_aa64pfr0_32bit_el0(info->reg_id_aa64pfr0))
-- 
2.39.5

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

[Catalin Marinas]

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>

Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>

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

[Will Deacon]

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/

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

[Will Deacon]

On Thu, Aug 27, 2026 at 01:56:35PM +0100, Will Deacon wrote:
> On Tue, Aug 25, 2026 at 05:42:19PM +0100, Fuad Tabba wrote:
> > 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 :/

Thinking about this for a few more minutes, perhaps the best option for
a quick fix would be to:

 - Internalise gmid_el1_accessible() in cpufeature.c and take cpuinfo *
   as a parameter

 - It could then refer directly to the ftr reg, along the lines of:

	val = arm64_ftr_safe_value(&ftr_id_aa64pfr1,
				   <value from id_aa64pfr1_override>,
				   <value from cpuinfo>);

 - Then check val before accessing the gmid.

Then we can fix all this properly in the future by moving the override
stuff earlier.

Will