[PATCH v2 2/2] arm64: Don't use KPTI where we have E0PD

From: Mark Brown <broonie@kernel•org>
To: Catalin Marinas <catalin.marinas@arm•com>, Will Deacon <will@kernel•org>
Cc: Mark Brown <broonie@kernel•org>,
	linux-arm-kernel@lists•infradead.org,
	Suzuki K Poulose <suzuki.poulose@arm•com>
Subject: [PATCH v2 2/2] arm64: Don't use KPTI where we have E0PD
Date: Wed, 14 Aug 2019 19:31:03 +0100	[thread overview]
Message-ID: <20190814183103.33707-3-broonie@kernel.org> (raw)
In-Reply-To: <20190814183103.33707-1-broonie@kernel.org>

Since E0PD is intended to fulfil the same role as KPTI we don't need to
use KPTI on CPUs where E0PD is available, we can rely on E0PD instead.
Change the check that forces KPTI on when KASLR is enabled to check for
E0PD before doing so, CPUs with E0PD are not expected to be affected by
meltdown so should not need to enable KPTI for other reasons.

Since we repeat the KPTI check for all CPUs we will still enable KPTI if
any of the CPUs in the system lacks E0PD. Since KPTI itself is not changed
by this patch once we enable KPTI we will do so for all CPUs. This is safe
but not optimally performant for such systems.

In order to ensure that we don't install any non-global mappings in
cases where we use E0PD for the system instead we add a check for E0PD
to the early checks in arm64_kernel_use_ng_mappings(), not installing NG
mappings if the current CPU has E0PD. This will incur an overhead on
systems where the boot CPU has E0PD but some others do not, however it
is expected that systems with very large memories which benefit most
from this optimization will be symmetric.

KPTI can still be forced on from the command line if required.

Signed-off-by: Mark Brown <broonie@kernel•org>
---

Added a check in arm64_kernel_use_ng_mappings() to suppress non-global
mappings when E0PD is present and KPTI isn't forced on.

 arch/arm64/include/asm/mmu.h   | 13 ++++++++++++-
 arch/arm64/kernel/cpufeature.c |  2 +-
 2 files changed, 13 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/include/asm/mmu.h b/arch/arm64/include/asm/mmu.h
index fd6161336653..85552f6fceda 100644
--- a/arch/arm64/include/asm/mmu.h
+++ b/arch/arm64/include/asm/mmu.h
@@ -38,6 +38,7 @@ static inline bool arm64_kernel_unmapped_at_el0(void)
 static inline bool arm64_kernel_use_ng_mappings(void)
 {
 	bool tx1_bug;
+	u64 ftr;
 
 	/* What's a kpti? Use global mappings if we don't know. */
 	if (!IS_ENABLED(CONFIG_UNMAP_KERNEL_AT_EL0))
@@ -59,7 +60,7 @@ static inline bool arm64_kernel_use_ng_mappings(void)
 	 * KASLR is enabled so we're going to be enabling kpti on non-broken
 	 * CPUs regardless of their susceptibility to Meltdown. Rather
 	 * than force everybody to go through the G -> nG dance later on,
-	 * just put down non-global mappings from the beginning.
+	 * just put down non-global mappings from the beginning...
 	 */
 	if (!IS_ENABLED(CONFIG_CAVIUM_ERRATUM_27456)) {
 		tx1_bug = false;
@@ -74,6 +75,16 @@ static inline bool arm64_kernel_use_ng_mappings(void)
 		tx1_bug = __cpus_have_const_cap(ARM64_WORKAROUND_CAVIUM_27456);
 	}
 
+	/*
+	 * ...unless we have E0PD in which case we may use that in
+	 * preference to unmapping the kernel.
+	 */
+	if (IS_ENABLED(CONFIG_ARM64_E0PD)) {
+		ftr = read_sysreg_s(SYS_ID_AA64MMFR2_EL1);
+		if ((ftr >> ID_AA64MMFR2_E0PD_SHIFT) & 0xf)
+			return false;
+	}
+
 	return !tx1_bug && kaslr_offset() > 0;
 }
 
diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
index 62b01fc35ef6..6bed144867ad 100644
--- a/arch/arm64/kernel/cpufeature.c
+++ b/arch/arm64/kernel/cpufeature.c
@@ -1003,7 +1003,7 @@ static bool unmap_kernel_at_el0(const struct arm64_cpu_capabilities *entry,
 
 	/* Useful for KASLR robustness */
 	if (IS_ENABLED(CONFIG_RANDOMIZE_BASE) && kaslr_offset() > 0) {
-		if (!__kpti_forced) {
+		if (!__kpti_forced && !this_cpu_has_cap(ARM64_HAS_E0PD)) {
 			str = "KASLR";
 			__kpti_forced = 1;
 		}
-- 
2.20.1


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists•infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
