Skip to contentSkip to conversation

E0PD support

PATCH Superseded Mark Brown 2 participants 8 messages Nov 4, 2019 – Nov 6, 2019 arm64

v5 of 9 Superseded Latest is v9 →

lore
Share & tools
lore

Mark Brown

author
#0 lore raw

v5 cover E0PD support

This series adds support for E0PD. We enable E0PD unconditionally where present on systems where all the CPUs in the system support E0PD and change to not enabling KPTI by default on systems where we have enabled E0PD. It also converts the runtime checks for use of non-global mappings into a variable.

v5: Rebase on v5.4-rc3 (there will be conflicts in cpucaps.h) and add
code in patch 4 to start making non-global mappings from boot if we know then that we will need KPTI, preserving an existing performance improvement.
v4: Use a variable to store our decision about using non-global mappings
rather than rechecking constantly at runtime. I've added this as a separate patch mainly for bisection.
v3: Make E0PD a system wide feature.

Mark Brown (4):
arm64: Add initial support for E0PD
arm64: Factor out checks for KASLR in KPTI code into separate function
arm64: Don't use KPTI where we have E0PD
arm64: Use a variable to store non-global mappings decision

 arch/arm64/Kconfig                     | 15 ++++++
 arch/arm64/include/asm/cpucaps.h       |  3 +-
 arch/arm64/include/asm/mmu.h           | 48 ++----------------
 arch/arm64/include/asm/pgtable-hwdef.h |  2 +
 arch/arm64/include/asm/pgtable-prot.h  |  4 +-
 arch/arm64/include/asm/sysreg.h        |  1 +
 arch/arm64/kernel/cpufeature.c         | 70 ++++++++++++++++++++++++--
 arch/arm64/mm/init.c                   |  7 +++
 8 files changed, 99 insertions(+), 51 deletions(-)
Signature · 7 lines
2.20.1


_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
author
#1 lore raw

PATCH v5 1/4 arm64: Add initial support for E0PD

Kernel Page Table Isolation (KPTI) is used to mitigate some speculation based security issues by ensuring that the kernel is not mapped when userspace is running but this approach is expensive and is incompatible with SPE. E0PD, introduced in the ARMv8.5 extensions, provides an alternative to this which ensures that accesses from userspace to the kernel's half of the memory map to always fault with constant time, preventing timing attacks without requiring constant unmapping and remapping or preventing legitimate accesses.

Currently this feature will only be enabled if all CPUs in the system support E0PD, if some CPUs do not support the feature at boot time then the feature will not be enabled and in the unlikely event that a late CPU is the first CPU to lack the feature then we will reject that CPU.

This initial patch does not yet integrate with KPTI, this will be dealt with in followup patches. Ideally we could ensure that by default we don't use KPTI on CPUs where E0PD is present.

---
 arch/arm64/Kconfig                     | 15 ++++++++++++++
 arch/arm64/include/asm/cpucaps.h       |  3 ++-
 arch/arm64/include/asm/pgtable-hwdef.h |  2 ++
 arch/arm64/include/asm/sysreg.h        |  1 +
 arch/arm64/kernel/cpufeature.c         | 27 ++++++++++++++++++++++++++
 5 files changed, 47 insertions(+), 1 deletion(-)
arch/arm64/Kconfig +15 -0
index 950a56b71ff0..9f881acb7acf 100644
@@ -1421,6 +1421,21 @@ config ARM64_PTR_AUTH
 
 endmenu
 
+menu "ARMv8.5 architectural features"
+
+config ARM64_E0PD
+	bool "Enable support for E0PD"
+	default y
+	help
+	   E0PD (part of the ARMv8.5 extensions) allows us to ensure
+	   that EL0 accesses made via TTBR1 always fault in constant time,
+	   providing benefits to KPTI with lower overhead and without
+	   disrupting legitimate access to kernel memory such as SPE.
+
+	   This option enables E0PD for TTBR1 where available.
+
+endmenu
+
 config ARM64_SVE
 	bool "ARM Scalable Vector Extension support"
 	default y
arch/arm64/include/asm/cpucaps.h +2 -1
index f19fe4b9acc4..f25388981075 100644
@@ -52,7 +52,8 @@
 #define ARM64_HAS_IRQ_PRIO_MASKING		42
 #define ARM64_HAS_DCPODP			43
 #define ARM64_WORKAROUND_1463225		44
+#define ARM64_HAS_E0PD				45
 
-#define ARM64_NCAPS				45
+#define ARM64_NCAPS				46
 
 #endif /* __ASM_CPUCAPS_H */
arch/arm64/include/asm/pgtable-hwdef.h +2 -0
index 3df60f97da1f..685842e52c3d 100644
@@ -292,6 +292,8 @@
 #define TCR_HD			(UL(1) << 40)
 #define TCR_NFD0		(UL(1) << 53)
 #define TCR_NFD1		(UL(1) << 54)
+#define TCR_E0PD0		(UL(1) << 55)
+#define TCR_E0PD1		(UL(1) << 56)
 
 /*
  * TTBR.
arch/arm64/include/asm/sysreg.h +1 -0
index 972d196c7714..36227a5a22ba 100644
@@ -655,6 +655,7 @@
 #define ID_AA64MMFR1_VMIDBITS_16	2
 
 /* id_aa64mmfr2 */
+#define ID_AA64MMFR2_E0PD_SHIFT		60
 #define ID_AA64MMFR2_FWB_SHIFT		40
 #define ID_AA64MMFR2_AT_SHIFT		32
 #define ID_AA64MMFR2_LVA_SHIFT		16
arch/arm64/kernel/cpufeature.c +27 -0
index cabebf1a7976..2cf2b129ebb4 100644
@@ -220,6 +220,7 @@ static const struct arm64_ftr_bits ftr_id_aa64mmfr1[] = {
 };
 
 static const struct arm64_ftr_bits ftr_id_aa64mmfr2[] = {
+	ARM64_FTR_BITS(FTR_HIDDEN, FTR_NONSTRICT, FTR_LOWER_SAFE, ID_AA64MMFR2_E0PD_SHIFT, 4, 0),
 	ARM64_FTR_BITS(FTR_HIDDEN, FTR_STRICT, FTR_LOWER_SAFE, ID_AA64MMFR2_FWB_SHIFT, 4, 0),
 	ARM64_FTR_BITS(FTR_VISIBLE, FTR_STRICT, FTR_LOWER_SAFE, ID_AA64MMFR2_AT_SHIFT, 4, 0),
 	ARM64_FTR_BITS(FTR_HIDDEN, FTR_STRICT, FTR_LOWER_SAFE, ID_AA64MMFR2_LVA_SHIFT, 4, 0),
@@ -1245,6 +1246,19 @@ static void cpu_enable_address_auth(struct arm64_cpu_capabilities const *cap)
 }
 #endif /* CONFIG_ARM64_PTR_AUTH */
 
+#ifdef CONFIG_ARM64_E0PD
+static void cpu_enable_e0pd(struct arm64_cpu_capabilities const *cap)
+{
+	/*
+	 * The cpu_enable() callback gets called even on CPUs that
+	 * don't detect the feature so we need to verify if we can
+	 * enable.
+	 */
+	if (this_cpu_has_cap(ARM64_HAS_E0PD))
+		sysreg_clear_set(tcr_el1, 0, TCR_E0PD1);
+}
+#endif /* CONFIG_ARM64_E0PD */
+
 #ifdef CONFIG_ARM64_PSEUDO_NMI
 static bool enable_pseudo_nmi;
 
@@ -1560,6 +1574,19 @@ static const struct arm64_cpu_capabilities arm64_features[] = {
 		.sign = FTR_UNSIGNED,
 		.min_field_value = 1,
 	},
+#endif
+#ifdef CONFIG_ARM64_E0PD
+	{
+		.desc = "E0PD",
+		.capability = ARM64_HAS_E0PD,
+		.type = ARM64_CPUCAP_SYSTEM_FEATURE,
+		.sys_reg = SYS_ID_AA64MMFR2_EL1,
+		.sign = FTR_UNSIGNED,
+		.field_pos = ID_AA64MMFR2_E0PD_SHIFT,
+		.matches = has_cpuid_feature,
+		.min_field_value = 1,
+		.cpu_enable = cpu_enable_e0pd,
+	},
 #endif
 	{},
 };
Signature · 7 lines
2.20.1


_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
author
#2 lore raw

PATCH v5 2/4 arm64: Factor out checks for KASLR in KPTI code into separate function

In reply to Mark Brown

In preparation for integrating E0PD support with KASLR factor out the checks for interaction between KASLR and KPTI done in boot context into a new function kaslr_requires_kpti(), in the process clarifying the distinction between what we do in boot context and what we do at runtime.

---
 arch/arm64/include/asm/mmu.h   | 53 +++++++++++++++++++++++-----------
 arch/arm64/kernel/cpufeature.c |  2 +-
 2 files changed, 37 insertions(+), 18 deletions(-)
arch/arm64/include/asm/mmu.h +36 -17
index f217e3292919..55e285fff262 100644
@@ -35,10 +35,37 @@ static inline bool arm64_kernel_unmapped_at_el0(void)
 	       cpus_have_const_cap(ARM64_UNMAP_KERNEL_AT_EL0);
 }
 
-static inline bool arm64_kernel_use_ng_mappings(void)
+static inline bool kaslr_requires_kpti(void)
 {
 	bool tx1_bug;
 
+	if (!IS_ENABLED(CONFIG_RANDOMIZE_BASE))
+		return false;
+
+	/*
+	 * Systems affected by Cavium erratum 24756 are incompatible
+	 * with KPTI.
+	 */
+	if (!IS_ENABLED(CONFIG_CAVIUM_ERRATUM_27456)) {
+		tx1_bug = false;
+#ifndef MODULE
+	} else if (!static_branch_likely(&arm64_const_caps_ready)) {
+		extern const struct midr_range cavium_erratum_27456_cpus[];
+
+		tx1_bug = is_midr_in_range_list(read_cpuid_id(),
+						cavium_erratum_27456_cpus);
+#endif
+	} else {
+		tx1_bug = __cpus_have_const_cap(ARM64_WORKAROUND_CAVIUM_27456);
+	}
+	if (tx1_bug)
+		return false;
+
+	return kaslr_offset() > 0;
+}
+
+static inline bool arm64_kernel_use_ng_mappings(void)
+{
 	/* What's a kpti? Use global mappings if we don't know. */
 	if (!IS_ENABLED(CONFIG_UNMAP_KERNEL_AT_EL0))
 		return false;
@@ -52,29 +79,21 @@ static inline bool arm64_kernel_use_ng_mappings(void)
 	if (arm64_kernel_unmapped_at_el0())
 		return true;
 
-	if (!IS_ENABLED(CONFIG_RANDOMIZE_BASE))
+	/*
+	 * Once we are far enough into boot for capabilities to be
+	 * ready we will have confirmed if we are using non-global
+	 * mappings so don't need to consider anything else here.
+	 */
+	if (static_branch_likely(&arm64_const_caps_ready))
 		return false;
 
 	/*
 	 * 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;
-#ifndef MODULE
-	} else if (!static_branch_likely(&arm64_const_caps_ready)) {
-		extern const struct midr_range cavium_erratum_27456_cpus[];
-
-		tx1_bug = is_midr_in_range_list(read_cpuid_id(),
-						cavium_erratum_27456_cpus);
-#endif
-	} else {
-		tx1_bug = __cpus_have_const_cap(ARM64_WORKAROUND_CAVIUM_27456);
-	}
-
-	return !tx1_bug && kaslr_offset() > 0;
+	return kaslr_requires_kpti();
 }
 
 typedef void (*bp_hardening_cb_t)(void);
arch/arm64/kernel/cpufeature.c +1 -1
index 2cf2b129ebb4..0d551af06421 100644
@@ -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 (kaslr_requires_kpti()) {
 		if (!__kpti_forced) {
 			str = "KASLR";
 			__kpti_forced = 1;
Signature · 7 lines
2.20.1


_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
author
#3 lore raw

PATCH v5 3/4 arm64: Don't use KPTI where we have E0PD

In reply to Mark Brown

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 E0PD is a system capability we will still enable KPTI if any of the CPUs in the system lacks E0PD, this will rewrite any global mappings that were established in systems where some but not all CPUs support E0PD. We may transiently have a mix of global and non-global mappings while booting since we use the local CPU when deciding if KPTI will be required prior to completing CPU enumeration but any global mappings will be converted to non-global ones when KPTI is applied.

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

---
 arch/arm64/include/asm/mmu.h | 11 +++++++++++
 1 file changed, 11 insertions(+)
arch/arm64/include/asm/mmu.h +11 -0
index 55e285fff262..d61908bf4c9c 100644
@@ -38,10 +38,21 @@ static inline bool arm64_kernel_unmapped_at_el0(void)
 static inline bool kaslr_requires_kpti(void)
 {
 	bool tx1_bug;
+	u64 ftr;
 
 	if (!IS_ENABLED(CONFIG_RANDOMIZE_BASE))
 		return false;
 
+	/*
+	 * E0PD does a similar job to KPTI so can be used instead
+	 * where available.
+	 */
+	if (IS_ENABLED(CONFIG_ARM64_E0PD)) {
+		ftr = read_sysreg_s(SYS_ID_AA64MMFR2_EL1);
+		if ((ftr >> ID_AA64MMFR2_E0PD_SHIFT) & 0xf)
+			return false;
+	}
+
 	/*
 	 * Systems affected by Cavium erratum 24756 are incompatible
 	 * with KPTI.
Signature · 7 lines
2.20.1


_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
author
#4 lore raw

PATCH v5 4/4 arm64: Use a variable to store non-global mappings decision

In reply to Mark Brown

Refactor the code which checks to see if we need to use non-global mappings to use a variable instead of checking with the CPU capabilities each time, doing the initial check for KPTI early in boot before we start allocating memory so we still avoid transitioning to non-global mappings in common cases.

Since this variable always matches our decision about non-global mappings this means we can also combine arm64_kernel_use_ng_mappings() and arm64_unmap_kernel_at_el0() into a single function, the variable simply stores the result and the decision code is elsewhere. We could just have the users check the variable directly but having a function makes it clear that these uses are read-only.

The result is that we simplify the code a bit and reduces the amount of code executed at runtime.

---
 arch/arm64/include/asm/mmu.h          | 78 ++-------------------------
 arch/arm64/include/asm/pgtable-prot.h |  4 +-
 arch/arm64/kernel/cpufeature.c        | 41 ++++++++++++--
 arch/arm64/mm/init.c                  |  7 +++
 4 files changed, 51 insertions(+), 79 deletions(-)
arch/arm64/include/asm/mmu.h +4 -74
index d61908bf4c9c..e4d862420bb4 100644
@@ -29,82 +29,11 @@ typedef struct {
  */
 #define ASID(mm)	((mm)->context.id.counter & 0xffff)
 
-static inline bool arm64_kernel_unmapped_at_el0(void)
-{
-	return IS_ENABLED(CONFIG_UNMAP_KERNEL_AT_EL0) &&
-	       cpus_have_const_cap(ARM64_UNMAP_KERNEL_AT_EL0);
-}
+extern bool arm64_use_ng_mappings;
 
-static inline bool kaslr_requires_kpti(void)
-{
-	bool tx1_bug;
-	u64 ftr;
-
-	if (!IS_ENABLED(CONFIG_RANDOMIZE_BASE))
-		return false;
-
-	/*
-	 * E0PD does a similar job to KPTI so can be used instead
-	 * where available.
-	 */
-	if (IS_ENABLED(CONFIG_ARM64_E0PD)) {
-		ftr = read_sysreg_s(SYS_ID_AA64MMFR2_EL1);
-		if ((ftr >> ID_AA64MMFR2_E0PD_SHIFT) & 0xf)
-			return false;
-	}
-
-	/*
-	 * Systems affected by Cavium erratum 24756 are incompatible
-	 * with KPTI.
-	 */
-	if (!IS_ENABLED(CONFIG_CAVIUM_ERRATUM_27456)) {
-		tx1_bug = false;
-#ifndef MODULE
-	} else if (!static_branch_likely(&arm64_const_caps_ready)) {
-		extern const struct midr_range cavium_erratum_27456_cpus[];
-
-		tx1_bug = is_midr_in_range_list(read_cpuid_id(),
-						cavium_erratum_27456_cpus);
-#endif
-	} else {
-		tx1_bug = __cpus_have_const_cap(ARM64_WORKAROUND_CAVIUM_27456);
-	}
-	if (tx1_bug)
-		return false;
-
-	return kaslr_offset() > 0;
-}
-
-static inline bool arm64_kernel_use_ng_mappings(void)
+static inline bool arm64_kernel_unmapped_at_el0(void)
 {
-	/* What's a kpti? Use global mappings if we don't know. */
-	if (!IS_ENABLED(CONFIG_UNMAP_KERNEL_AT_EL0))
-		return false;
-
-	/*
-	 * Note: this function is called before the CPU capabilities have
-	 * been configured, so our early mappings will be global. If we
-	 * later determine that kpti is required, then
-	 * kpti_install_ng_mappings() will make them non-global.
-	 */
-	if (arm64_kernel_unmapped_at_el0())
-		return true;
-
-	/*
-	 * Once we are far enough into boot for capabilities to be
-	 * ready we will have confirmed if we are using non-global
-	 * mappings so don't need to consider anything else here.
-	 */
-	if (static_branch_likely(&arm64_const_caps_ready))
-		return false;
-
-	/*
-	 * 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
-	 */
-	return kaslr_requires_kpti();
+	return arm64_use_ng_mappings;
 }
 
 typedef void (*bp_hardening_cb_t)(void);
@@ -158,6 +87,7 @@ extern void create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys,
 			       pgprot_t prot, bool page_mappings_only);
 extern void *fixmap_remap_fdt(phys_addr_t dt_phys, int *size, pgprot_t prot);
 extern void mark_linear_text_alias_ro(void);
+extern bool kaslr_requires_kpti(void);
 
 #define INIT_MM_CONTEXT(name)	\
 	.pgd = init_pg_dir,
arch/arm64/include/asm/pgtable-prot.h +2 -2
index 9a21b84536f2..eb1c6f83343d 100644
@@ -26,8 +26,8 @@
 #define _PROT_DEFAULT		(PTE_TYPE_PAGE | PTE_AF | PTE_SHARED)
 #define _PROT_SECT_DEFAULT	(PMD_TYPE_SECT | PMD_SECT_AF | PMD_SECT_S)
 
-#define PTE_MAYBE_NG		(arm64_kernel_use_ng_mappings() ? PTE_NG : 0)
-#define PMD_MAYBE_NG		(arm64_kernel_use_ng_mappings() ? PMD_SECT_NG : 0)
+#define PTE_MAYBE_NG		(arm64_kernel_unmapped_at_el0() ? PTE_NG : 0)
+#define PMD_MAYBE_NG		(arm64_kernel_unmapped_at_el0() ? PMD_SECT_NG : 0)
 
 #define PROT_DEFAULT		(_PROT_DEFAULT | PTE_MAYBE_NG)
 #define PROT_SECT_DEFAULT	(_PROT_SECT_DEFAULT | PMD_MAYBE_NG)
arch/arm64/kernel/cpufeature.c +38 -3

71 lines not on this page · raw message

arch/arm64/mm/init.c +7 -0
index 45c00a54909c..a09a02b5979a 100644
@@ -456,6 +456,13 @@ void __init bootmem_init(void)
 	zone_sizes_init(min, max);
 
 	memblock_dump_all();
+
+	/*
+	 * If know now we are going to need KPTI then use non-global
+	 * mappings from the start, avoiding the cost of rewriting
+	 * everything later.
+	 */
+	arm64_use_ng_mappings = kaslr_requires_kpti();
 }
 
 #ifndef CONFIG_SPARSEMEM_VMEMMAP
Signature · 7 lines
2.20.1


_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
maintainer
#5 lore raw

Re: [PATCH v5 4/4] arm64: Use a variable to store non-global mappings decision

Quoted 6 lines · Mark Brown
arch/arm64/mm/init.c +0 -0
index 45c00a54909c..a09a02b5979a 100644
@@ -456,6 +456,13 @@ void __init bootmem_init(void)
 	zone_sizes_init(min, max);

arch/arm64/mm/init.c void __init bootmem_init(void) · line 466

Quoted 10 lines
arch/arm64/mm/init.c +7 -0
 
 	memblock_dump_all();
+
+	/*
+	 * If know now we are going to need KPTI then use non-global
+	 * mappings from the start, avoiding the cost of rewriting
+	 * everything later.
+	 */
+	arm64_use_ng_mappings = kaslr_requires_kpti();
 }

I think we need this even earlier in paging_init() before we start creating the kernel mappings.

Signature · 6 lines
Catalin

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
author
#6 lore raw

Re: [PATCH v5 4/4] arm64: Use a variable to store non-global mappings decision

Quoted 12 lines · Catalin Marinas

arch/arm64/mm/init.c void __init bootmem_init(void) · line 466

Quoted 7 lines · Mark Brown
arch/arm64/mm/init.c +6 -0
+	/*
+	 * If know now we are going to need KPTI then use non-global
+	 * mappings from the start, avoiding the cost of rewriting
+	 * everything later.
+	 */
+	arm64_use_ng_mappings = kaslr_requires_kpti();
 }

I think we need this even earlier in paging_init() before we start creating the kernel mappings.

OK, I'll move it. I didn't see any global mappings getting created with the current code but better safe than sorry.

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

maintainer
#7 lore raw

Re: [PATCH v5 4/4] arm64: Use a variable to store non-global mappings decision

Quoted 14 lines · Mark Brown
Quoted 12 lines · Catalin Marinas

arch/arm64/mm/init.c void __init bootmem_init(void) · line 466

Quoted 7 lines · Mark Brown
arch/arm64/mm/init.c +6 -0
+	/*
+	 * If know now we are going to need KPTI then use non-global
+	 * mappings from the start, avoiding the cost of rewriting
+	 * everything later.
+	 */
+	arm64_use_ng_mappings = kaslr_requires_kpti();
 }

I think we need this even earlier in paging_init() before we start creating the kernel mappings.

Quoted 2 lines

OK, I'll move it. I didn't see any global mappings getting created with the current code but better safe than sorry.

I gave these patches a try under kvm/qemu on TX2 and most of the linear map turns out global in ptdump.

I'll try your other series.

Signature · 6 lines
Catalin

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel