[PATCH 0/2] arch/x86: Enable MPK feature on AMD

[PATCH 0/2] arch/x86: Enable MPK feature on AMD

[Babu Moger]

AMD's next generation of EPYC processors support the MPK (Memory
Protection Keys) feature.

AMD documentation for MPK feature is available at "AMD64 Architecture
Programmer’s Manual Volume 2: System Programming, Pub. 24593 Rev. 3.34,
Section 5.6.6 Memory Protection Keys (MPK) Bit".

The documentation can be obtained at the link below:
https://bugzilla.kernel.org/show_bug.cgi?id=206537

This series enables the feature on AMD and updates config parameters
to reflect the MPK support on generic x86 platforms.

---

Babu Moger (2):
      arch/x86: Rename config X86_INTEL_MEMORY_PROTECTION_KEYS to generic x86
      KVM: SVM: Add support for MPK feature on AMD


 Documentation/core-api/protection-keys.rst     |    3 ++-
 arch/x86/Kconfig                               |    6 +++---
 arch/x86/include/asm/disabled-features.h       |    4 ++--
 arch/x86/include/asm/mmu.h                     |    2 +-
 arch/x86/include/asm/mmu_context.h             |    4 ++--
 arch/x86/include/asm/pgtable.h                 |    4 ++--
 arch/x86/include/asm/pgtable_types.h           |    2 +-
 arch/x86/include/asm/special_insns.h           |    2 +-
 arch/x86/include/uapi/asm/mman.h               |    2 +-
 arch/x86/kernel/cpu/common.c                   |    2 +-
 arch/x86/kvm/svm/svm.c                         |   20 ++++++++++++++++++++
 arch/x86/kvm/svm/svm.h                         |    2 ++
 arch/x86/mm/Makefile                           |    2 +-
 arch/x86/mm/pkeys.c                            |    2 +-
 scripts/headers_install.sh                     |    2 +-
 tools/arch/x86/include/asm/disabled-features.h |    4 ++--
 16 files changed, 43 insertions(+), 20 deletions(-)

--

[PATCH 1/2] arch/x86: Rename config X86_INTEL_MEMORY_PROTECTION_KEYS to generic x86

[Babu Moger]

AMD's next generation of EPYC processors support the MPK (Memory
Protection Keys) feature.

So, rename X86_INTEL_MEMORY_PROTECTION_KEYS to X86_MEMORY_PROTECTION_KEYS.

No functional changes.

Signed-off-by: Babu Moger <babu.moger@amd.com>
---
 Documentation/core-api/protection-keys.rst     |    3 ++-
 arch/x86/Kconfig                               |    6 +++---
 arch/x86/include/asm/disabled-features.h       |    4 ++--
 arch/x86/include/asm/mmu.h                     |    2 +-
 arch/x86/include/asm/mmu_context.h             |    4 ++--
 arch/x86/include/asm/pgtable.h                 |    4 ++--
 arch/x86/include/asm/pgtable_types.h           |    2 +-
 arch/x86/include/asm/special_insns.h           |    2 +-
 arch/x86/include/uapi/asm/mman.h               |    2 +-
 arch/x86/kernel/cpu/common.c                   |    2 +-
 arch/x86/mm/Makefile                           |    2 +-
 arch/x86/mm/pkeys.c                            |    2 +-
 scripts/headers_install.sh                     |    2 +-
 tools/arch/x86/include/asm/disabled-features.h |    4 ++--
 14 files changed, 21 insertions(+), 20 deletions(-)

diff --git a/Documentation/core-api/protection-keys.rst b/Documentation/core-api/protection-keys.rst
index 49d9833af871..d25e89e53c59 100644
--- a/Documentation/core-api/protection-keys.rst
+++ b/Documentation/core-api/protection-keys.rst
@@ -6,7 +6,8 @@ Memory Protection Keys
 
 Memory Protection Keys for Userspace (PKU aka PKEYs) is a feature
 which is found on Intel's Skylake "Scalable Processor" Server CPUs.
-It will be avalable in future non-server parts.
+It will be available in future non-server parts. Also, AMD64
+Architecture Programmer’s Manual defines PKU feature in AMD processors.
 
 For anyone wishing to test or use this feature, it is available in
 Amazon's EC2 C5 instances and is known to work there using an Ubuntu
diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 1197b5596d5a..8630b9fa06f5 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -1886,11 +1886,11 @@ config X86_UMIP
 	  specific cases in protected and virtual-8086 modes. Emulated
 	  results are dummy.
 
-config X86_INTEL_MEMORY_PROTECTION_KEYS
-	prompt "Intel Memory Protection Keys"
+config X86_MEMORY_PROTECTION_KEYS
+	prompt "Memory Protection Keys"
 	def_bool y
 	# Note: only available in 64-bit mode
-	depends on CPU_SUP_INTEL && X86_64
+	depends on X86_64 && (CPU_SUP_INTEL || CPU_SUP_AMD)
 	select ARCH_USES_HIGH_VMA_FLAGS
 	select ARCH_HAS_PKEYS
 	---help---
diff --git a/arch/x86/include/asm/disabled-features.h b/arch/x86/include/asm/disabled-features.h
index 4ea8584682f9..52dbdfed8043 100644
--- a/arch/x86/include/asm/disabled-features.h
+++ b/arch/x86/include/asm/disabled-features.h
@@ -36,13 +36,13 @@
 # define DISABLE_PCID		(1<<(X86_FEATURE_PCID & 31))
 #endif /* CONFIG_X86_64 */
 
-#ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS
+#ifdef CONFIG_X86_MEMORY_PROTECTION_KEYS
 # define DISABLE_PKU		0
 # define DISABLE_OSPKE		0
 #else
 # define DISABLE_PKU		(1<<(X86_FEATURE_PKU & 31))
 # define DISABLE_OSPKE		(1<<(X86_FEATURE_OSPKE & 31))
-#endif /* CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS */
+#endif /* CONFIG_X86_MEMORY_PROTECTION_KEYS */
 
 #ifdef CONFIG_X86_5LEVEL
 # define DISABLE_LA57	0
diff --git a/arch/x86/include/asm/mmu.h b/arch/x86/include/asm/mmu.h
index bdeae9291e5c..351d22152709 100644
--- a/arch/x86/include/asm/mmu.h
+++ b/arch/x86/include/asm/mmu.h
@@ -42,7 +42,7 @@ typedef struct {
 	const struct vdso_image *vdso_image;	/* vdso image in use */
 
 	atomic_t perf_rdpmc_allowed;	/* nonzero if rdpmc is allowed */
-#ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS
+#ifdef CONFIG_X86_MEMORY_PROTECTION_KEYS
 	/*
 	 * One bit per protection key says whether userspace can
 	 * use it or not.  protected by mmap_sem.
diff --git a/arch/x86/include/asm/mmu_context.h b/arch/x86/include/asm/mmu_context.h
index 4e55370e48e8..33f4a7ccac5e 100644
--- a/arch/x86/include/asm/mmu_context.h
+++ b/arch/x86/include/asm/mmu_context.h
@@ -118,7 +118,7 @@ static inline int init_new_context(struct task_struct *tsk,
 	mm->context.ctx_id = atomic64_inc_return(&last_mm_ctx_id);
 	atomic64_set(&mm->context.tlb_gen, 0);
 
-#ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS
+#ifdef CONFIG_X86_MEMORY_PROTECTION_KEYS
 	if (cpu_feature_enabled(X86_FEATURE_OSPKE)) {
 		/* pkey 0 is the default and allocated implicitly */
 		mm->context.pkey_allocation_map = 0x1;
@@ -163,7 +163,7 @@ do {						\
 static inline void arch_dup_pkeys(struct mm_struct *oldmm,
 				  struct mm_struct *mm)
 {
-#ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS
+#ifdef CONFIG_X86_MEMORY_PROTECTION_KEYS
 	if (!cpu_feature_enabled(X86_FEATURE_OSPKE))
 		return;
 
diff --git a/arch/x86/include/asm/pgtable.h b/arch/x86/include/asm/pgtable.h
index 4d02e64af1b3..4265720d62c2 100644
--- a/arch/x86/include/asm/pgtable.h
+++ b/arch/x86/include/asm/pgtable.h
@@ -1451,7 +1451,7 @@ static inline pmd_t pmd_swp_clear_uffd_wp(pmd_t pmd)
 #define PKRU_WD_BIT 0x2
 #define PKRU_BITS_PER_PKEY 2
 
-#ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS
+#ifdef CONFIG_X86_MEMORY_PROTECTION_KEYS
 extern u32 init_pkru_value;
 #else
 #define init_pkru_value	0
@@ -1475,7 +1475,7 @@ static inline bool __pkru_allows_write(u32 pkru, u16 pkey)
 
 static inline u16 pte_flags_pkey(unsigned long pte_flags)
 {
-#ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS
+#ifdef CONFIG_X86_MEMORY_PROTECTION_KEYS
 	/* ifdef to avoid doing 59-bit shift on 32-bit values */
 	return (pte_flags & _PAGE_PKEY_MASK) >> _PAGE_BIT_PKEY_BIT0;
 #else
diff --git a/arch/x86/include/asm/pgtable_types.h b/arch/x86/include/asm/pgtable_types.h
index b6606fe6cfdf..c61a1ff71d53 100644
--- a/arch/x86/include/asm/pgtable_types.h
+++ b/arch/x86/include/asm/pgtable_types.h
@@ -56,7 +56,7 @@
 #define _PAGE_PAT_LARGE (_AT(pteval_t, 1) << _PAGE_BIT_PAT_LARGE)
 #define _PAGE_SPECIAL	(_AT(pteval_t, 1) << _PAGE_BIT_SPECIAL)
 #define _PAGE_CPA_TEST	(_AT(pteval_t, 1) << _PAGE_BIT_CPA_TEST)
-#ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS
+#ifdef CONFIG_X86_MEMORY_PROTECTION_KEYS
 #define _PAGE_PKEY_BIT0	(_AT(pteval_t, 1) << _PAGE_BIT_PKEY_BIT0)
 #define _PAGE_PKEY_BIT1	(_AT(pteval_t, 1) << _PAGE_BIT_PKEY_BIT1)
 #define _PAGE_PKEY_BIT2	(_AT(pteval_t, 1) << _PAGE_BIT_PKEY_BIT2)
diff --git a/arch/x86/include/asm/special_insns.h b/arch/x86/include/asm/special_insns.h
index 6d37b8fcfc77..70eaae7e8f04 100644
--- a/arch/x86/include/asm/special_insns.h
+++ b/arch/x86/include/asm/special_insns.h
@@ -73,7 +73,7 @@ static inline unsigned long native_read_cr4(void)
 
 void native_write_cr4(unsigned long val);
 
-#ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS
+#ifdef CONFIG_X86_MEMORY_PROTECTION_KEYS
 static inline u32 rdpkru(void)
 {
 	u32 ecx = 0;
diff --git a/arch/x86/include/uapi/asm/mman.h b/arch/x86/include/uapi/asm/mman.h
index d4a8d0424bfb..d4da414a9de2 100644
--- a/arch/x86/include/uapi/asm/mman.h
+++ b/arch/x86/include/uapi/asm/mman.h
@@ -4,7 +4,7 @@
 
 #define MAP_32BIT	0x40		/* only give out 32bit addresses */
 
-#ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS
+#ifdef CONFIG_X86_MEMORY_PROTECTION_KEYS
 /*
  * Take the 4 protection key bits out of the vma->vm_flags
  * value and turn them in to the bits that we can put in
diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index bed0cb83fe24..e5fb9955214c 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -448,7 +448,7 @@ static __always_inline void setup_pku(struct cpuinfo_x86 *c)
 	set_cpu_cap(c, X86_FEATURE_OSPKE);
 }
 
-#ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS
+#ifdef CONFIG_X86_MEMORY_PROTECTION_KEYS
 static __init int setup_disable_pku(char *arg)
 {
 	/*
diff --git a/arch/x86/mm/Makefile b/arch/x86/mm/Makefile
index 98f7c6fa2eaa..17ebf12ba8ff 100644
--- a/arch/x86/mm/Makefile
+++ b/arch/x86/mm/Makefile
@@ -45,7 +45,7 @@ obj-$(CONFIG_AMD_NUMA)		+= amdtopology.o
 obj-$(CONFIG_ACPI_NUMA)		+= srat.o
 obj-$(CONFIG_NUMA_EMU)		+= numa_emulation.o
 
-obj-$(CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)	+= pkeys.o
+obj-$(CONFIG_X86_MEMORY_PROTECTION_KEYS)	+= pkeys.o
 obj-$(CONFIG_RANDOMIZE_MEMORY)			+= kaslr.o
 obj-$(CONFIG_PAGE_TABLE_ISOLATION)		+= pti.o
 
diff --git a/arch/x86/mm/pkeys.c b/arch/x86/mm/pkeys.c
index 8873ed1438a9..a77497e8d58c 100644
--- a/arch/x86/mm/pkeys.c
+++ b/arch/x86/mm/pkeys.c
@@ -1,6 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0-only
 /*
- * Intel Memory Protection Keys management
+ * Memory Protection Keys management
  * Copyright (c) 2015, Intel Corporation.
  */
 #include <linux/debugfs.h>		/* debugfs_create_u32()		*/
diff --git a/scripts/headers_install.sh b/scripts/headers_install.sh
index a07668a5c36b..6e60e5362d3e 100755
--- a/scripts/headers_install.sh
+++ b/scripts/headers_install.sh
@@ -86,7 +86,7 @@ arch/sh/include/uapi/asm/sigcontext.h:CONFIG_CPU_SH5
 arch/sh/include/uapi/asm/stat.h:CONFIG_CPU_SH5
 arch/x86/include/uapi/asm/auxvec.h:CONFIG_IA32_EMULATION
 arch/x86/include/uapi/asm/auxvec.h:CONFIG_X86_64
-arch/x86/include/uapi/asm/mman.h:CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS
+arch/x86/include/uapi/asm/mman.h:CONFIG_X86_MEMORY_PROTECTION_KEYS
 include/uapi/asm-generic/fcntl.h:CONFIG_64BIT
 include/uapi/linux/atmdev.h:CONFIG_COMPAT
 include/uapi/linux/elfcore.h:CONFIG_BINFMT_ELF_FDPIC
diff --git a/tools/arch/x86/include/asm/disabled-features.h b/tools/arch/x86/include/asm/disabled-features.h
index 4ea8584682f9..52dbdfed8043 100644
--- a/tools/arch/x86/include/asm/disabled-features.h
+++ b/tools/arch/x86/include/asm/disabled-features.h
@@ -36,13 +36,13 @@
 # define DISABLE_PCID		(1<<(X86_FEATURE_PCID & 31))
 #endif /* CONFIG_X86_64 */
 
-#ifdef CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS
+#ifdef CONFIG_X86_MEMORY_PROTECTION_KEYS
 # define DISABLE_PKU		0
 # define DISABLE_OSPKE		0
 #else
 # define DISABLE_PKU		(1<<(X86_FEATURE_PKU & 31))
 # define DISABLE_OSPKE		(1<<(X86_FEATURE_OSPKE & 31))
-#endif /* CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS */
+#endif /* CONFIG_X86_MEMORY_PROTECTION_KEYS */
 
 #ifdef CONFIG_X86_5LEVEL
 # define DISABLE_LA57	0

[PATCH 2/2] KVM: SVM: Add support for MPK feature on AMD

[Babu Moger]

The Memory Protection Key (MPK) feature provides a way for applications
to impose page-based data access protections (read/write, read-only or
no access), without requiring modification of page tables and subsequent
TLB invalidations when the application changes protection domains.

This feature is already available in Intel platforms. Now enable the
feature on AMD platforms.

The host pkru state needs to be saved/restored during the guest/host
switches in SVM.  Other changes are already taken care by the pkru
common code.

AMD documentation for MPK feature is available at "AMD64 Architecture
Programmer’s Manual Volume 2: System Programming, Pub. 24593 Rev. 3.34,
Section 5.6.6 Memory Protection Keys (MPK) Bit". Documentation can be
obtained at the link below.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=206537
Signed-off-by: Babu Moger <babu.moger@amd.com>
---
 arch/x86/kvm/svm/svm.c |   20 ++++++++++++++++++++
 arch/x86/kvm/svm/svm.h |    2 ++
 2 files changed, 22 insertions(+)

diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
index 2f379bacbb26..de327f02470f 100644
--- a/arch/x86/kvm/svm/svm.c
+++ b/arch/x86/kvm/svm/svm.c
@@ -818,6 +818,10 @@ static __init void svm_set_cpu_caps(void)
 	if (boot_cpu_has(X86_FEATURE_LS_CFG_SSBD) ||
 	    boot_cpu_has(X86_FEATURE_AMD_SSBD))
 		kvm_cpu_cap_set(X86_FEATURE_VIRT_SSBD);
+
+	/* PKU is not yet implemented for shadow paging. */
+	if (npt_enabled && boot_cpu_has(X86_FEATURE_OSPKE))
+		kvm_cpu_cap_check_and_set(X86_FEATURE_PKU);
 }
 
 static __init int svm_hardware_setup(void)
@@ -1300,6 +1304,8 @@ static void svm_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 		indirect_branch_prediction_barrier();
 	}
 	avic_vcpu_load(vcpu, cpu);
+
+	svm->host_pkru = read_pkru();
 }
 
 static void svm_vcpu_put(struct kvm_vcpu *vcpu)
@@ -3318,6 +3324,12 @@ static void svm_vcpu_run(struct kvm_vcpu *vcpu)
 	clgi();
 	kvm_load_guest_xsave_state(vcpu);
 
+	/* Load the guest pkru state */
+	if (static_cpu_has(X86_FEATURE_PKU) &&
+	    kvm_read_cr4_bits(vcpu, X86_CR4_PKE) &&
+	    vcpu->arch.pkru != svm->host_pkru)
+		__write_pkru(vcpu->arch.pkru);
+
 	if (lapic_in_kernel(vcpu) &&
 		vcpu->arch.apic->lapic_timer.timer_advance_ns)
 		kvm_wait_lapic_expire(vcpu);
@@ -3371,6 +3383,14 @@ static void svm_vcpu_run(struct kvm_vcpu *vcpu)
 	if (unlikely(svm->vmcb->control.exit_code == SVM_EXIT_NMI))
 		kvm_before_interrupt(&svm->vcpu);
 
+	/* Save the guest pkru state and restore the host pkru state back */
+	if (static_cpu_has(X86_FEATURE_PKU) &&
+	    kvm_read_cr4_bits(vcpu, X86_CR4_PKE)) {
+		vcpu->arch.pkru = rdpkru();
+		if (vcpu->arch.pkru != svm->host_pkru)
+			__write_pkru(svm->host_pkru);
+	}
+
 	kvm_load_host_xsave_state(vcpu);
 	stgi();
 
diff --git a/arch/x86/kvm/svm/svm.h b/arch/x86/kvm/svm/svm.h
index df3474f4fb02..5d20a28c1b0e 100644
--- a/arch/x86/kvm/svm/svm.h
+++ b/arch/x86/kvm/svm/svm.h
@@ -158,6 +158,8 @@ struct vcpu_svm {
 	u64 *avic_physical_id_cache;
 	bool avic_is_running;
 
+	u32 host_pkru;
+
 	/*
 	 * Per-vcpu list of struct amd_svm_iommu_ir:
 	 * This is used mainly to store interrupt remapping information used

Re: [PATCH 1/2] arch/x86: Rename config X86_INTEL_MEMORY_PROTECTION_KEYS to generic x86

[Dave Hansen]

> diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
> index 1197b5596d5a..8630b9fa06f5 100644
> --- a/arch/x86/Kconfig
> +++ b/arch/x86/Kconfig
> @@ -1886,11 +1886,11 @@ config X86_UMIP
>  	  specific cases in protected and virtual-8086 modes. Emulated
>  	  results are dummy.
>  
> -config X86_INTEL_MEMORY_PROTECTION_KEYS
> -	prompt "Intel Memory Protection Keys"
> +config X86_MEMORY_PROTECTION_KEYS
> +	prompt "Memory Protection Keys"
>  	def_bool y
>  	# Note: only available in 64-bit mode
> -	depends on CPU_SUP_INTEL && X86_64
> +	depends on X86_64 && (CPU_SUP_INTEL || CPU_SUP_AMD)
>  	select ARCH_USES_HIGH_VMA_FLAGS
>  	select ARCH_HAS_PKEYS
>  	---help---

It's a bit of a bummer that we're going to prompt everybody doing
oldconfig's for this new option.  But, I don't know any way for Kconfig
to suppress it if the name is changed.  Also, I guess the def_bool=y
means that menuconfig and olddefconfig will tend to do the right thing.

Do we *really* need to change the Kconfig name?  The text prompt, sure.
 End users see that and having Intel in there is massively confusing.

If I have to put up with seeing 'amd64' all over my Debian package
names, you can put up with a Kconfig name. :P

I'm really just wondering what the point of the churn is.

Re: [PATCH 2/2] KVM: SVM: Add support for MPK feature on AMD

[Sean Christopherson]

On Wed, May 06, 2020 at 05:02:21PM -0500, Babu Moger wrote:
>  static __init int svm_hardware_setup(void)
> @@ -1300,6 +1304,8 @@ static void svm_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
>  		indirect_branch_prediction_barrier();
>  	}
>  	avic_vcpu_load(vcpu, cpu);
> +
> +	svm->host_pkru = read_pkru();

Move vcpu_vmx's host_prku to kvm_vcpu_arch instead of duplicating it to
SVM.  And I'm 99% certain "vcpu->arch.host_pkru = read_pkru()" can be moved
to kvm_arch_vcpu_load().  The only direct calls to vmx_vcpu_load() are to
get the right VMCS loaded.  Actually, those calls shouldn't be using
vmx_vcpu_load(), especially since that'll trigger IBPB.  I'll send a patch
for that.

>  }
>  
>  static void svm_vcpu_put(struct kvm_vcpu *vcpu)
> @@ -3318,6 +3324,12 @@ static void svm_vcpu_run(struct kvm_vcpu *vcpu)
>  	clgi();
>  	kvm_load_guest_xsave_state(vcpu);
>  
> +	/* Load the guest pkru state */
> +	if (static_cpu_has(X86_FEATURE_PKU) &&
> +	    kvm_read_cr4_bits(vcpu, X86_CR4_PKE) &&
> +	    vcpu->arch.pkru != svm->host_pkru)
> +		__write_pkru(vcpu->arch.pkru);

This and the restoration should be moved to common x86 helpers, at a glance
they look identical.

In short, pretty much all of this belongs in common x86.

> +
>  	if (lapic_in_kernel(vcpu) &&
>  		vcpu->arch.apic->lapic_timer.timer_advance_ns)
>  		kvm_wait_lapic_expire(vcpu);
> @@ -3371,6 +3383,14 @@ static void svm_vcpu_run(struct kvm_vcpu *vcpu)
>  	if (unlikely(svm->vmcb->control.exit_code == SVM_EXIT_NMI))
>  		kvm_before_interrupt(&svm->vcpu);
>  
> +	/* Save the guest pkru state and restore the host pkru state back */
> +	if (static_cpu_has(X86_FEATURE_PKU) &&
> +	    kvm_read_cr4_bits(vcpu, X86_CR4_PKE)) {
> +		vcpu->arch.pkru = rdpkru();
> +		if (vcpu->arch.pkru != svm->host_pkru)
> +			__write_pkru(svm->host_pkru);
> +	}
> +
>  	kvm_load_host_xsave_state(vcpu);
>  	stgi();
>  
> diff --git a/arch/x86/kvm/svm/svm.h b/arch/x86/kvm/svm/svm.h
> index df3474f4fb02..5d20a28c1b0e 100644
> --- a/arch/x86/kvm/svm/svm.h
> +++ b/arch/x86/kvm/svm/svm.h
> @@ -158,6 +158,8 @@ struct vcpu_svm {
>  	u64 *avic_physical_id_cache;
>  	bool avic_is_running;
>  
> +	u32 host_pkru;
> +
>  	/*
>  	 * Per-vcpu list of struct amd_svm_iommu_ir:
>  	 * This is used mainly to store interrupt remapping information used
> 

Re: [PATCH 1/2] arch/x86: Rename config X86_INTEL_MEMORY_PROTECTION_KEYS to generic x86

[Dave Hansen]

On 5/6/20 3:21 PM, Dave Hansen wrote:
> I'm really just wondering what the point of the churn is.

The config option is also in the manpages, fwiw:

	http://man7.org/linux/man-pages/man7/pkeys.7.html

By the way, I am regretting ever sticking "INTEL_" in there.  Seems like
a good best practice would be to leave those things out in the future if
there's any credible opportunity for another vendor to add support.

Re: [PATCH 1/2] arch/x86: Rename config X86_INTEL_MEMORY_PROTECTION_KEYS to generic x86

[Logan Gunthorpe]

On 2020-05-06 4:21 p.m., Dave Hansen wrote:
>> diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
>> index 1197b5596d5a..8630b9fa06f5 100644
>> --- a/arch/x86/Kconfig
>> +++ b/arch/x86/Kconfig
>> @@ -1886,11 +1886,11 @@ config X86_UMIP
>>  	  specific cases in protected and virtual-8086 modes. Emulated
>>  	  results are dummy.
>>  
>> -config X86_INTEL_MEMORY_PROTECTION_KEYS
>> -	prompt "Intel Memory Protection Keys"
>> +config X86_MEMORY_PROTECTION_KEYS
>> +	prompt "Memory Protection Keys"
>>  	def_bool y
>>  	# Note: only available in 64-bit mode
>> -	depends on CPU_SUP_INTEL && X86_64
>> +	depends on X86_64 && (CPU_SUP_INTEL || CPU_SUP_AMD)
>>  	select ARCH_USES_HIGH_VMA_FLAGS
>>  	select ARCH_HAS_PKEYS
>>  	---help---
> 
> It's a bit of a bummer that we're going to prompt everybody doing
> oldconfig's for this new option.  But, I don't know any way for Kconfig
> to suppress it if the name is changed.  Also, I guess the def_bool=y
> means that menuconfig and olddefconfig will tend to do the right thing.
> 
> Do we *really* need to change the Kconfig name?  The text prompt, sure.
>  End users see that and having Intel in there is massively confusing.
> 
> If I have to put up with seeing 'amd64' all over my Debian package
> names, you can put up with a Kconfig name. :P

Lol, isn't that just Intel's penance for Itanium?

Logan

Re: [PATCH 2/2] KVM: SVM: Add support for MPK feature on AMD

[Dave Hansen]

On 5/6/20 3:02 PM, Babu Moger wrote:
> --- a/arch/x86/kvm/svm/svm.c
> +++ b/arch/x86/kvm/svm/svm.c
> @@ -818,6 +818,10 @@ static __init void svm_set_cpu_caps(void)
>  	if (boot_cpu_has(X86_FEATURE_LS_CFG_SSBD) ||
>  	    boot_cpu_has(X86_FEATURE_AMD_SSBD))
>  		kvm_cpu_cap_set(X86_FEATURE_VIRT_SSBD);
> +
> +	/* PKU is not yet implemented for shadow paging. */
> +	if (npt_enabled && boot_cpu_has(X86_FEATURE_OSPKE))
> +		kvm_cpu_cap_check_and_set(X86_FEATURE_PKU);
>  }

Reviewed-by: Dave Hansen <dave.hansen@intel.com>

But, I'll also say that we probably shouldn't have put the other code
into arch/x86/kvm/vmx/vmx.c in the first place.  It would be much nicer
if this refactored the current code into a common spot rather than
copying.  But, I do understand the impulse to do it the way it was done.

Re: [PATCH 1/2] arch/x86: Rename config X86_INTEL_MEMORY_PROTECTION_KEYS to generic x86

[Sebastian Andrzej Siewior]

On 2020-05-06 15:21:29 [-0700], Dave Hansen wrote:
> > diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
> > index 1197b5596d5a..8630b9fa06f5 100644
> > --- a/arch/x86/Kconfig
> > +++ b/arch/x86/Kconfig
> > @@ -1886,11 +1886,11 @@ config X86_UMIP
> >  	  specific cases in protected and virtual-8086 modes. Emulated
> >  	  results are dummy.
> >  
> > -config X86_INTEL_MEMORY_PROTECTION_KEYS
> > -	prompt "Intel Memory Protection Keys"
> > +config X86_MEMORY_PROTECTION_KEYS
> > +	prompt "Memory Protection Keys"
> >  	def_bool y
> >  	# Note: only available in 64-bit mode
> > -	depends on CPU_SUP_INTEL && X86_64
> > +	depends on X86_64 && (CPU_SUP_INTEL || CPU_SUP_AMD)
> >  	select ARCH_USES_HIGH_VMA_FLAGS
> >  	select ARCH_HAS_PKEYS
> >  	---help---
> 
> It's a bit of a bummer that we're going to prompt everybody doing
> oldconfig's for this new option.  But, I don't know any way for Kconfig
> to suppress it if the name is changed.  Also, I guess the def_bool=y
> means that menuconfig and olddefconfig will tend to do the right thing.

You could add a new option (X86_MEMORY_PROTECTION_KEYS) which is
def_bool X86_INTEL_MEMORY_PROTECTION_KEYS and avoiding the prompt line.
Soo it is selected based on the old option and the user isn't bother. A
few cycles later you could remove intel option and add prompt to other.
But still little work for…

> Do we *really* need to change the Kconfig name?  The text prompt, sure.
>  End users see that and having Intel in there is massively confusing.
> 
> If I have to put up with seeing 'amd64' all over my Debian package
> names, you can put up with a Kconfig name. :P

:) Right. On AMD you also use the crc32c-intel (if possible) and I
haven't seen people complain about this one.

> I'm really just wondering what the point of the churn is.

Sebastian

Re: [PATCH 2/2] KVM: SVM: Add support for MPK feature on AMD

[Paolo Bonzini]

On 07/05/20 00:26, Sean Christopherson wrote:
>> +	/* Load the guest pkru state */
>> +	if (static_cpu_has(X86_FEATURE_PKU) &&
>> +	    kvm_read_cr4_bits(vcpu, X86_CR4_PKE) &&
>> +	    vcpu->arch.pkru != svm->host_pkru)
>> +		__write_pkru(vcpu->arch.pkru);
> This and the restoration should be moved to common x86 helpers, at a glance
> they look identical.
> 
> In short, pretty much all of this belongs in common x86.
> 

We could stick this in kvm_load_guest_xsave_state
kvm_load_host_xsave_state.  It's not a perfect match, after all the code
itself proves that PKRU can be loaded without XSAVE; but it's close
enough and it's exactly in the right point of vmx_vcpu_run and svm_vcpu_run.

Paolo

Re: [PATCH 1/2] arch/x86: Rename config X86_INTEL_MEMORY_PROTECTION_KEYS to generic x86

[Dave Hansen]

On 5/7/20 12:29 AM, Sebastian Andrzej Siewior wrote:
>>> -config X86_INTEL_MEMORY_PROTECTION_KEYS
>>> -	prompt "Intel Memory Protection Keys"
>>> +config X86_MEMORY_PROTECTION_KEYS
>>> +	prompt "Memory Protection Keys"
>>>  	def_bool y
>>>  	# Note: only available in 64-bit mode
>>> -	depends on CPU_SUP_INTEL && X86_64
>>> +	depends on X86_64 && (CPU_SUP_INTEL || CPU_SUP_AMD)
>>>  	select ARCH_USES_HIGH_VMA_FLAGS
>>>  	select ARCH_HAS_PKEYS
>>>  	---help---
>> It's a bit of a bummer that we're going to prompt everybody doing
>> oldconfig's for this new option.  But, I don't know any way for Kconfig
>> to suppress it if the name is changed.  Also, I guess the def_bool=y
>> means that menuconfig and olddefconfig will tend to do the right thing.
> You could add a new option (X86_MEMORY_PROTECTION_KEYS) which is
> def_bool X86_INTEL_MEMORY_PROTECTION_KEYS and avoiding the prompt line.
> Soo it is selected based on the old option and the user isn't bother. A
> few cycles later you could remove intel option and add prompt to other.
> But still little work for…

That does sound viable, if we decide it's all worth it.

So, for now my preference would be to change the prompt, but leave the
CONFIG_ naming in place.  If we decide that transitioning the config is
the right thing (I don't feel super strongly either way), let's use
Sebastian's trick to avoid the Kconfig prompts.

Re: [PATCH 1/2] arch/x86: Rename config X86_INTEL_MEMORY_PROTECTION_KEYS to generic x86

[Paolo Bonzini]

On 07/05/20 16:44, Dave Hansen wrote:
>> You could add a new option (X86_MEMORY_PROTECTION_KEYS) which is
>> def_bool X86_INTEL_MEMORY_PROTECTION_KEYS and avoiding the prompt line.
>> Soo it is selected based on the old option and the user isn't bother. A
>> few cycles later you could remove intel option and add prompt to other.
>> But still little work for…
> That does sound viable, if we decide it's all worth it.
> 
> So, for now my preference would be to change the prompt, but leave the
> CONFIG_ naming in place.

I agree.

What's in a name?  An Intel rose by any other name would smell as sweet.
 Oh wait... :)

Paolo

> If we decide that transitioning the config is
> the right thing (I don't feel super strongly either way), let's use
> Sebastian's trick to avoid the Kconfig prompts.
> 

Re: [PATCH 1/2] arch/x86: Rename config X86_INTEL_MEMORY_PROTECTION_KEYS to generic x86

[Babu Moger]

On 5/7/20 10:16 AM, Paolo Bonzini wrote:
> On 07/05/20 16:44, Dave Hansen wrote:
>>> You could add a new option (X86_MEMORY_PROTECTION_KEYS) which is
>>> def_bool X86_INTEL_MEMORY_PROTECTION_KEYS and avoiding the prompt line.
>>> Soo it is selected based on the old option and the user isn't bother. A
>>> few cycles later you could remove intel option and add prompt to other.
>>> But still little work for…
>> That does sound viable, if we decide it's all worth it.
>>
>> So, for now my preference would be to change the prompt, but leave the
>> CONFIG_ naming in place.
> 
> I agree.
> 
> What's in a name?  An Intel rose by any other name would smell as sweet.

How about X86_MPK? Or I will use already proposed name
X86_MEMORY_PROTECTION_KEYS.

>  Oh wait... :)
> 
> Paolo
> 
>> If we decide that transitioning the config is
>> the right thing (I don't feel super strongly either way), let's use
>> Sebastian's trick to avoid the Kconfig prompts.
>>
> 

Re: [PATCH 1/2] arch/x86: Rename config X86_INTEL_MEMORY_PROTECTION_KEYS to generic x86

[Paolo Bonzini]

On 07/05/20 18:06, Babu Moger wrote:
>>> So, for now my preference would be to change the prompt, but leave the
>>> CONFIG_ naming in place.
>> I agree.
>>
>> What's in a name?  An Intel rose by any other name would smell as sweet.
> 
> How about X86_MPK? Or I will use already proposed name
> X86_MEMORY_PROTECTION_KEYS.

Dave is proposing to keep the CONFIG_ as is and only change the prompt.

Paolo

Re: [PATCH 1/2] arch/x86: Rename config X86_INTEL_MEMORY_PROTECTION_KEYS to generic x86

[Babu Moger]

On 5/7/20 11:07 AM, Paolo Bonzini wrote:
> On 07/05/20 18:06, Babu Moger wrote:
>>>> So, for now my preference would be to change the prompt, but leave the
>>>> CONFIG_ naming in place.
>>> I agree.
>>>
>>> What's in a name?  An Intel rose by any other name would smell as sweet.
>>
>> How about X86_MPK? Or I will use already proposed name
>> X86_MEMORY_PROTECTION_KEYS.
> 
> Dave is proposing to keep the CONFIG_ as is and only change the prompt.

Ok. Got it. thanks