Skip to contentSkip to conversation

arm64: cpufeature: Trim feature reporting and include PAN emulation

PATCH In review Kees Cook 3 participants 5 messages Feb 20, 2018 – Feb 21, 2018 arm64

lore
Share & tools
lore

Kees Cook

author
#0 lore raw

PATCH 1/1 arm64: cpufeature: Trim feature reporting and include PAN emulation

The PAN emulation notification was only happening for non-boot CPUs if CPU capabilities had already been configured. This seems to be the wrong place, as it's system-wide and isn't attached to capabilities, so its reporting didn't normally happen. Instead, report it once from the boot CPU. Additionally removes the redundant "feature" word from the "CPU features:" line.

Before (redundant "feature", and missing PAN emulation report):

SMP: Total of 4 processors activated.
CPU features: detected feature: 32-bit EL0 Support
CPU features: detected feature: Kernel page table isolation (KPTI)
CPU: All CPU(s) started at EL2

After:

SMP: Total of 4 processors activated.
CPU features: detected: 32-bit EL0 Support
CPU features: detected: Kernel page table isolation (KPTI)
CPU features: emulated: Privileged Access Never (PAN) using TTBR0_EL1 switching
CPU: All CPU(s) started at EL2

---
 arch/arm64/kernel/cpufeature.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
arch/arm64/kernel/cpufeature.c +4 -4
index 29b1f873e337..6c799ca58b53 100644
@@ -1333,9 +1333,6 @@ static void verify_local_cpu_capabilities(void)
 
 	if (system_supports_sve())
 		verify_sve_features();
-
-	if (system_uses_ttbr0_pan())
-		pr_info("Emulating Privileged Access Never (PAN) using TTBR0_EL1 switching\n");
 }
 
 void check_local_cpu_capabilities(void)
@@ -1360,7 +1357,7 @@ void check_local_cpu_capabilities(void)
 
 static void __init setup_feature_capabilities(void)
 {
-	update_cpu_capabilities(arm64_features, "detected feature:");
+	update_cpu_capabilities(arm64_features, "detected:");
 	enable_cpu_capabilities(arm64_features);
 }
 
@@ -1394,6 +1391,9 @@ void __init setup_cpu_features(void)
 	if (system_supports_32bit_el0())
 		setup_elf_hwcaps(compat_elf_hwcaps);
 
+	if (system_uses_ttbr0_pan())
+		pr_info("emulated: Privileged Access Never (PAN) using TTBR0_EL1 switching\n");
+
 	sve_setup();
 
 	/* Advertise that we have computed the system capabilities */
Signature · 6 lines
2.7.4


-- 
Kees Cook
Pixel Security
#1 lore raw
Quoted 42 lines · Kees Cook

The PAN emulation notification was only happening for non-boot CPUs if CPU capabilities had already been configured. This seems to be the wrong place, as it's system-wide and isn't attached to capabilities, so its reporting didn't normally happen. Instead, report it once from the boot CPU. Additionally removes the redundant "feature" word from the "CPU features:" line.

Before (redundant "feature", and missing PAN emulation report):

SMP: Total of 4 processors activated.
CPU features: detected feature: 32-bit EL0 Support
CPU features: detected feature: Kernel page table isolation (KPTI)
CPU: All CPU(s) started at EL2

After:

SMP: Total of 4 processors activated.
CPU features: detected: 32-bit EL0 Support
CPU features: detected: Kernel page table isolation (KPTI)
CPU features: emulated: Privileged Access Never (PAN) using TTBR0_EL1 switching
CPU: All CPU(s) started at EL2

---
 arch/arm64/kernel/cpufeature.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
arch/arm64/kernel/cpufeature.c +0 -3
index 29b1f873e337..6c799ca58b53 100644
@@ -1333,9 +1333,6 @@ static void verify_local_cpu_capabilities(void)
 
 	if (system_supports_sve())
 		verify_sve_features();
-
-	if (system_uses_ttbr0_pan())
-		pr_info("Emulating Privileged Access Never (PAN) using TTBR0_EL1 switching\n");
 }
 
 void check_local_cpu_capabilities(void)

arch/arm64/kernel/cpufeature.c void check_local_cpu_capabilities(void) · line 1360

Quoted 5 lines
arch/arm64/kernel/cpufeature.c +1 -1
@@ -1360,7 +1357,7 @@ void check_local_cpu_capabilities(void)
 
 static void __init setup_feature_capabilities(void)
 {
-	update_cpu_capabilities(arm64_features, "detected feature:");
+	update_cpu_capabilities(arm64_features, "detected:");

Although I get what you're saying about redundant use of the word "features", this feels like cosmetic churn that is unrelated to the problem this patch is addressing.

It could be worth reviewing the CPU errata messages and other miscellaneous printks together to make them less verbose and more consistent all in one go, but that would be a separate patch...

arch/arm64/kernel/cpufeature.c void __init setup_cpu_features(void) · line 1396

Quoted 10 lines
arch/arm64/kernel/cpufeature.c +3 -0
 	enable_cpu_capabilities(arm64_features);
 }
 
@@ -1394,6 +1391,9 @@ void __init setup_cpu_features(void)
 	if (system_supports_32bit_el0())
 		setup_elf_hwcaps(compat_elf_hwcaps);
 
+	if (system_uses_ttbr0_pan())
+		pr_info("emulated: Privileged Access Never (PAN) using TTBR0_EL1 switching\n");
+

Moving this seems sensible. The other option would be to paste it into update_cpu_capabilities(), but the message would still potentially get printed multiple times, so that doesn't feel like the right approach.

[...]

Cheers
---Dave

reviewer Acked-by Acked-by
#2 lore raw
Quoted 44 lines · Dave Martin
Quoted 43 lines · Kees Cook

The PAN emulation notification was only happening for non-boot CPUs if CPU capabilities had already been configured. This seems to be the wrong place, as it's system-wide and isn't attached to capabilities, so its reporting didn't normally happen. Instead, report it once from the boot CPU. Additionally removes the redundant "feature" word from the "CPU features:" line.

Before (redundant "feature", and missing PAN emulation report):

SMP: Total of 4 processors activated.
CPU features: detected feature: 32-bit EL0 Support
CPU features: detected feature: Kernel page table isolation (KPTI)
CPU: All CPU(s) started at EL2

After:

SMP: Total of 4 processors activated.
CPU features: detected: 32-bit EL0 Support
CPU features: detected: Kernel page table isolation (KPTI)
CPU features: emulated: Privileged Access Never (PAN) using TTBR0_EL1 switching
CPU: All CPU(s) started at EL2

---
 arch/arm64/kernel/cpufeature.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
arch/arm64/kernel/cpufeature.c +0 -3
index 29b1f873e337..6c799ca58b53 100644
@@ -1333,9 +1333,6 @@ static void verify_local_cpu_capabilities(void)
 
 	if (system_supports_sve())
 		verify_sve_features();
-
-	if (system_uses_ttbr0_pan())
-		pr_info("Emulating Privileged Access Never (PAN) using TTBR0_EL1 switching\n");
 }
 
 void check_local_cpu_capabilities(void)
@@ -1360,7 +1357,7 @@ void check_local_cpu_capabilities(void)
 
Quoted 8 lines

arch/arm64/kernel/cpufeature.c void check_local_cpu_capabilities(void) · line 1360

Quoted 4 lines
arch/arm64/kernel/cpufeature.c +1 -1
 static void __init setup_feature_capabilities(void)
 {
-	update_cpu_capabilities(arm64_features, "detected feature:");
+	update_cpu_capabilities(arm64_features, "detected:");

Although I get what you're saying about redundant use of the word "features", this feels like cosmetic churn that is unrelated to the problem this patch is addressing.

Given it seems sensible, shall we just split that into a separate patch?

FWIW, for a patch with just this change:

Quoted 11 lines

It could be worth reviewing the CPU errata messages and other miscellaneous printks together to make them less verbose and more consistent all in one go, but that would be a separate patch...

Quoted 7 lines
arch/arm64/kernel/cpufeature.c +0 -0
 	enable_cpu_capabilities(arm64_features);
 }
 
@@ -1394,6 +1391,9 @@ void __init setup_cpu_features(void)
 	if (system_supports_32bit_el0())
 		setup_elf_hwcaps(compat_elf_hwcaps);
 
Quoted 7 lines

arch/arm64/kernel/cpufeature.c void __init setup_cpu_features(void) · line 1396

Quoted 3 lines
arch/arm64/kernel/cpufeature.c +3 -0
+	if (system_uses_ttbr0_pan())
+		pr_info("emulated: Privileged Access Never (PAN) using TTBR0_EL1 switching\n");
+

Moving this seems sensible. The other option would be to paste it into update_cpu_capabilities(), but the message would still potentially get printed multiple times, so that doesn't feel like the right approach.

I think that more ideally, we'd give this an entry in the arm64_features array, but because it's effectively a negative feature, it's a little tricky.

This also looks fine to me, so FWIW:

Mark.

#3 lore raw
Quoted 54 lines · Mark Rutland
Quoted 44 lines · Dave Martin
Quoted 43 lines · Kees Cook

The PAN emulation notification was only happening for non-boot CPUs if CPU capabilities had already been configured. This seems to be the wrong place, as it's system-wide and isn't attached to capabilities, so its reporting didn't normally happen. Instead, report it once from the boot CPU. Additionally removes the redundant "feature" word from the "CPU features:" line.

Before (redundant "feature", and missing PAN emulation report):

SMP: Total of 4 processors activated.
CPU features: detected feature: 32-bit EL0 Support
CPU features: detected feature: Kernel page table isolation (KPTI)
CPU: All CPU(s) started at EL2

After:

SMP: Total of 4 processors activated.
CPU features: detected: 32-bit EL0 Support
CPU features: detected: Kernel page table isolation (KPTI)
CPU features: emulated: Privileged Access Never (PAN) using TTBR0_EL1 switching
CPU: All CPU(s) started at EL2

---
 arch/arm64/kernel/cpufeature.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
arch/arm64/kernel/cpufeature.c +0 -3
index 29b1f873e337..6c799ca58b53 100644
@@ -1333,9 +1333,6 @@ static void verify_local_cpu_capabilities(void)
 
 	if (system_supports_sve())
 		verify_sve_features();
-
-	if (system_uses_ttbr0_pan())
-		pr_info("Emulating Privileged Access Never (PAN) using TTBR0_EL1 switching\n");
 }
 
 void check_local_cpu_capabilities(void)
@@ -1360,7 +1357,7 @@ void check_local_cpu_capabilities(void)
 
Quoted 8 lines

arch/arm64/kernel/cpufeature.c void check_local_cpu_capabilities(void) · line 1360

Quoted 4 lines
arch/arm64/kernel/cpufeature.c +1 -1
 static void __init setup_feature_capabilities(void)
 {
-	update_cpu_capabilities(arm64_features, "detected feature:");
+	update_cpu_capabilities(arm64_features, "detected:");

Although I get what you're saying about redundant use of the word "features", this feels like cosmetic churn that is unrelated to the problem this patch is addressing.

Quoted 1 line

Given it seems sensible, shall we just split that into a separate patch?

Can do, though I still think it's a bit dubious. The original message as it stands isn't wrong. I won't lose sleep over it though.

[...]

Cheers
---Dave

author
#4 lore raw
In reply to Mark Rutland
Quoted 54 lines · Mark Rutland
Quoted 44 lines · Dave Martin
Quoted 43 lines · Kees Cook

The PAN emulation notification was only happening for non-boot CPUs if CPU capabilities had already been configured. This seems to be the wrong place, as it's system-wide and isn't attached to capabilities, so its reporting didn't normally happen. Instead, report it once from the boot CPU. Additionally removes the redundant "feature" word from the "CPU features:" line.

Before (redundant "feature", and missing PAN emulation report):

SMP: Total of 4 processors activated.
CPU features: detected feature: 32-bit EL0 Support
CPU features: detected feature: Kernel page table isolation (KPTI)
CPU: All CPU(s) started at EL2

After:

SMP: Total of 4 processors activated.
CPU features: detected: 32-bit EL0 Support
CPU features: detected: Kernel page table isolation (KPTI)
CPU features: emulated: Privileged Access Never (PAN) using TTBR0_EL1 switching
CPU: All CPU(s) started at EL2

---
 arch/arm64/kernel/cpufeature.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
arch/arm64/kernel/cpufeature.c +0 -3
index 29b1f873e337..6c799ca58b53 100644
@@ -1333,9 +1333,6 @@ static void verify_local_cpu_capabilities(void)
 
    if (system_supports_sve())
            verify_sve_features();
-
-   if (system_uses_ttbr0_pan())
-           pr_info("Emulating Privileged Access Never (PAN) using TTBR0_EL1 switching\n");
 }
 
 void check_local_cpu_capabilities(void)
@@ -1360,7 +1357,7 @@ void check_local_cpu_capabilities(void)
 
Quoted 8 lines

arch/arm64/kernel/cpufeature.c void check_local_cpu_capabilities(void) · line 1360

Quoted 4 lines
arch/arm64/kernel/cpufeature.c +1 -1
 static void __init setup_feature_capabilities(void)
 {
-   update_cpu_capabilities(arm64_features, "detected feature:");
+   update_cpu_capabilities(arm64_features, "detected:");

Although I get what you're saying about redundant use of the word "features", this feels like cosmetic churn that is unrelated to the problem this patch is addressing.

Quoted 5 lines

Given it seems sensible, shall we just split that into a separate patch?

FWIW, for a patch with just this change:

Sure!

Quoted 19 lines
Quoted 11 lines

It could be worth reviewing the CPU errata messages and other miscellaneous printks together to make them less verbose and more consistent all in one go, but that would be a separate patch...

Quoted 7 lines
arch/arm64/kernel/cpufeature.c +0 -0
    enable_cpu_capabilities(arm64_features);
 }
 
@@ -1394,6 +1391,9 @@ void __init setup_cpu_features(void)
    if (system_supports_32bit_el0())
            setup_elf_hwcaps(compat_elf_hwcaps);
 
Quoted 7 lines

arch/arm64/kernel/cpufeature.c void __init setup_cpu_features(void) · line 1396

Quoted 3 lines
arch/arm64/kernel/cpufeature.c +3 -0
+   if (system_uses_ttbr0_pan())
+           pr_info("emulated: Privileged Access Never (PAN) using TTBR0_EL1 switching\n");
+

Moving this seems sensible. The other option would be to paste it into update_cpu_capabilities(), but the message would still potentially get printed multiple times, so that doesn't feel like the right approach.

Quoted 2 lines

I think that more ideally, we'd give this an entry in the arm64_features array, but because it's effectively a negative feature, it's a little tricky.

Yeah, I looked at that but it seemed like it would needlessly burn a capability bit for no real gain.

Quoted 3 lines

This also looks fine to me, so FWIW:

Thanks, I'll split the patches!

-Kees

Signature · 2 lines
Kees Cook
Pixel Security