[PATCH v2 0/3] KVM: x86: Minor emulator cleanup

[PATCH v2 0/3] KVM: x86: Minor emulator cleanup

[Sean Christopherson]

Three small patches to move emulator specific variables from 'struct
kvm_vcpu_arch' to 'struct x86_emulate_ctxt'.

v2:
  - Rebase to kvm/queue, 2c2787938512 ("KVM: selftests: Stop ...")

Sean Christopherson (3):
  KVM: x86: Add EMULTYPE_PF when emulation is triggered by a page fault
  KVM: x86: Move gpa_val and gpa_available into the emulator context
  KVM: x86: Move #PF retry tracking variables into emulation context

 arch/x86/include/asm/kvm_emulate.h |  8 ++++++
 arch/x86/include/asm/kvm_host.h    | 19 ++++++-------
 arch/x86/kvm/mmu/mmu.c             | 10 ++-----
 arch/x86/kvm/x86.c                 | 45 +++++++++++++++++++-----------
 4 files changed, 48 insertions(+), 34 deletions(-)

-- 
2.24.1

[PATCH v2 1/3] KVM: x86: Add EMULTYPE_PF when emulation is triggered by a page fault

[Sean Christopherson]

Add a new emulation type flag to explicitly mark emulation related to a
page fault.  Move the propation of the GPA into the emulator from the
page fault handler into x86_emulate_instruction, using EMULTYPE_PF as an
indicator that cr2 is valid.  Similarly, don't propagate cr2 into the
exception.address when it's *not* valid.

Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
---
 arch/x86/include/asm/kvm_host.h | 12 +++++++++---
 arch/x86/kvm/mmu/mmu.c          | 10 ++--------
 arch/x86/kvm/x86.c              | 25 +++++++++++++++++++------
 3 files changed, 30 insertions(+), 17 deletions(-)

diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 4dffbc10d3f8..10c1e8f472b6 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -1370,8 +1370,9 @@ extern u64 kvm_mce_cap_supported;
  *		   decode the instruction length.  For use *only* by
  *		   kvm_x86_ops->skip_emulated_instruction() implementations.
  *
- * EMULTYPE_ALLOW_RETRY - Set when the emulator should resume the guest to
- *			  retry native execution under certain conditions.
+ * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
+ *			     retry native execution under certain conditions,
+ *			     Can only be set in conjunction with EMULTYPE_PF.
  *
  * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
  *			     triggered by KVM's magic "force emulation" prefix,
@@ -1384,13 +1385,18 @@ extern u64 kvm_mce_cap_supported;
  *			backdoor emulation, which is opt in via module param.
  *			VMware backoor emulation handles select instructions
  *			and reinjects the #GP for all other cases.
+ *
+ * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
+ *		 case the CR2/GPA value pass on the stack is valid.
  */
 #define EMULTYPE_NO_DECODE	    (1 << 0)
 #define EMULTYPE_TRAP_UD	    (1 << 1)
 #define EMULTYPE_SKIP		    (1 << 2)
-#define EMULTYPE_ALLOW_RETRY	    (1 << 3)
+#define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)
 #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
 #define EMULTYPE_VMWARE_GP	    (1 << 5)
+#define EMULTYPE_PF		    (1 << 6)
+
 int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
 int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
 					void *insn, int insn_len);
diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index 7011a4e54866..258624d46588 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -5416,18 +5416,12 @@ EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page_virt);
 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
 		       void *insn, int insn_len)
 {
-	int r, emulation_type = 0;
+	int r, emulation_type = EMULTYPE_PF;
 	bool direct = vcpu->arch.mmu->direct_map;
 
 	if (WARN_ON(!VALID_PAGE(vcpu->arch.mmu->root_hpa)))
 		return RET_PF_RETRY;
 
-	/* With shadow page tables, fault_address contains a GVA or nGPA.  */
-	if (vcpu->arch.mmu->direct_map) {
-		vcpu->arch.gpa_available = true;
-		vcpu->arch.gpa_val = cr2_or_gpa;
-	}
-
 	r = RET_PF_INVALID;
 	if (unlikely(error_code & PFERR_RSVD_MASK)) {
 		r = handle_mmio_page_fault(vcpu, cr2_or_gpa, direct);
@@ -5472,7 +5466,7 @@ int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
 	 * for L1 isn't going to magically fix whatever issue cause L2 to fail.
 	 */
 	if (!mmio_info_in_cache(vcpu, cr2_or_gpa, direct) && !is_guest_mode(vcpu))
-		emulation_type = EMULTYPE_ALLOW_RETRY;
+		emulation_type |= EMULTYPE_ALLOW_RETRY_PF;
 emulate:
 	/*
 	 * On AMD platforms, under certain conditions insn_len may be zero on #NPF.
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index fbabb2f06273..92af6c5a69e3 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -6483,10 +6483,11 @@ static bool reexecute_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
 	gpa_t gpa = cr2_or_gpa;
 	kvm_pfn_t pfn;
 
-	if (!(emulation_type & EMULTYPE_ALLOW_RETRY))
+	if (!(emulation_type & EMULTYPE_ALLOW_RETRY_PF))
 		return false;
 
-	if (WARN_ON_ONCE(is_guest_mode(vcpu)))
+	if (WARN_ON_ONCE(is_guest_mode(vcpu)) ||
+	    WARN_ON_ONCE(!(emulation_type & EMULTYPE_PF)))
 		return false;
 
 	if (!vcpu->arch.mmu->direct_map) {
@@ -6574,10 +6575,11 @@ static bool retry_instruction(struct x86_emulate_ctxt *ctxt,
 	 */
 	vcpu->arch.last_retry_eip = vcpu->arch.last_retry_addr = 0;
 
-	if (!(emulation_type & EMULTYPE_ALLOW_RETRY))
+	if (!(emulation_type & EMULTYPE_ALLOW_RETRY_PF))
 		return false;
 
-	if (WARN_ON_ONCE(is_guest_mode(vcpu)))
+	if (WARN_ON_ONCE(is_guest_mode(vcpu)) ||
+	    WARN_ON_ONCE(!(emulation_type & EMULTYPE_PF)))
 		return false;
 
 	if (x86_page_table_writing_insn(ctxt))
@@ -6830,8 +6832,19 @@ int x86_emulate_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
 	}
 
 restart:
-	/* Save the faulting GPA (cr2) in the address field */
-	ctxt->exception.address = cr2_or_gpa;
+	if (emulation_type & EMULTYPE_PF) {
+		/* Save the faulting GPA (cr2) in the address field */
+		ctxt->exception.address = cr2_or_gpa;
+
+		/* With shadow page tables, cr2 contains a GVA or nGPA. */
+		if (vcpu->arch.mmu->direct_map) {
+			vcpu->arch.gpa_available = true;
+			vcpu->arch.gpa_val = cr2_or_gpa;
+		}
+	} else {
+		/* Sanitize the address out of an abundance of paranoia. */
+		ctxt->exception.address = 0;
+	}
 
 	r = x86_emulate_insn(ctxt);
 
-- 
2.24.1

Re: [PATCH v2 1/3] KVM: x86: Add EMULTYPE_PF when emulation is triggered by a page fault

[Xiaoyao Li]

On 2/19/2020 7:03 AM, Sean Christopherson wrote:
> Add a new emulation type flag to explicitly mark emulation related to a
> page fault.  Move the propation of the GPA into the emulator from the
> page fault handler into x86_emulate_instruction, using EMULTYPE_PF as an
> indicator that cr2 is valid.  Similarly, don't propagate cr2 into the
> exception.address when it's *not* valid.
> 
> Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
> ---
>   arch/x86/include/asm/kvm_host.h | 12 +++++++++---
>   arch/x86/kvm/mmu/mmu.c          | 10 ++--------
>   arch/x86/kvm/x86.c              | 25 +++++++++++++++++++------
>   3 files changed, 30 insertions(+), 17 deletions(-)
> 
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index 4dffbc10d3f8..10c1e8f472b6 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -1370,8 +1370,9 @@ extern u64 kvm_mce_cap_supported;
>    *		   decode the instruction length.  For use *only* by
>    *		   kvm_x86_ops->skip_emulated_instruction() implementations.
>    *
> - * EMULTYPE_ALLOW_RETRY - Set when the emulator should resume the guest to
> - *			  retry native execution under certain conditions.
> + * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
> + *			     retry native execution under certain conditions,
> + *			     Can only be set in conjunction with EMULTYPE_PF.
>    *
>    * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
>    *			     triggered by KVM's magic "force emulation" prefix,
> @@ -1384,13 +1385,18 @@ extern u64 kvm_mce_cap_supported;
>    *			backdoor emulation, which is opt in via module param.
>    *			VMware backoor emulation handles select instructions
>    *			and reinjects the #GP for all other cases.
> + *
> + * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
> + *		 case the CR2/GPA value pass on the stack is valid.
>    */
>   #define EMULTYPE_NO_DECODE	    (1 << 0)
>   #define EMULTYPE_TRAP_UD	    (1 << 1)
>   #define EMULTYPE_SKIP		    (1 << 2)
> -#define EMULTYPE_ALLOW_RETRY	    (1 << 3)
> +#define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)

How about naming it as EMULTYPE_PF_ALLOW_RETRY and exchanging the bit 
position with EMULTYPE_PF ?

>   #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
>   #define EMULTYPE_VMWARE_GP	    (1 << 5)
> +#define EMULTYPE_PF		    (1 << 6)
> +
>   int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
>   int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
>   					void *insn, int insn_len);
> diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
> index 7011a4e54866..258624d46588 100644
> --- a/arch/x86/kvm/mmu/mmu.c
> +++ b/arch/x86/kvm/mmu/mmu.c
> @@ -5416,18 +5416,12 @@ EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page_virt);
>   int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
>   		       void *insn, int insn_len)
>   {
> -	int r, emulation_type = 0;
> +	int r, emulation_type = EMULTYPE_PF;
>   	bool direct = vcpu->arch.mmu->direct_map;
>   
>   	if (WARN_ON(!VALID_PAGE(vcpu->arch.mmu->root_hpa)))
>   		return RET_PF_RETRY;
>   
> -	/* With shadow page tables, fault_address contains a GVA or nGPA.  */
> -	if (vcpu->arch.mmu->direct_map) {
> -		vcpu->arch.gpa_available = true;
> -		vcpu->arch.gpa_val = cr2_or_gpa;
> -	}
> -
>   	r = RET_PF_INVALID;
>   	if (unlikely(error_code & PFERR_RSVD_MASK)) {
>   		r = handle_mmio_page_fault(vcpu, cr2_or_gpa, direct);
> @@ -5472,7 +5466,7 @@ int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
>   	 * for L1 isn't going to magically fix whatever issue cause L2 to fail.
>   	 */
>   	if (!mmio_info_in_cache(vcpu, cr2_or_gpa, direct) && !is_guest_mode(vcpu))
> -		emulation_type = EMULTYPE_ALLOW_RETRY;
> +		emulation_type |= EMULTYPE_ALLOW_RETRY_PF;
>   emulate:
>   	/*
>   	 * On AMD platforms, under certain conditions insn_len may be zero on #NPF.
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index fbabb2f06273..92af6c5a69e3 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -6483,10 +6483,11 @@ static bool reexecute_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
>   	gpa_t gpa = cr2_or_gpa;
>   	kvm_pfn_t pfn;
>   
> -	if (!(emulation_type & EMULTYPE_ALLOW_RETRY))
> +	if (!(emulation_type & EMULTYPE_ALLOW_RETRY_PF))
>   		return false;
>   
> -	if (WARN_ON_ONCE(is_guest_mode(vcpu)))
> +	if (WARN_ON_ONCE(is_guest_mode(vcpu)) ||
> +	    WARN_ON_ONCE(!(emulation_type & EMULTYPE_PF)))
>   		return false;
>   
>   	if (!vcpu->arch.mmu->direct_map) {
> @@ -6574,10 +6575,11 @@ static bool retry_instruction(struct x86_emulate_ctxt *ctxt,
>   	 */
>   	vcpu->arch.last_retry_eip = vcpu->arch.last_retry_addr = 0;
>   
> -	if (!(emulation_type & EMULTYPE_ALLOW_RETRY))
> +	if (!(emulation_type & EMULTYPE_ALLOW_RETRY_PF))
>   		return false;
>   
> -	if (WARN_ON_ONCE(is_guest_mode(vcpu)))
> +	if (WARN_ON_ONCE(is_guest_mode(vcpu)) ||
> +	    WARN_ON_ONCE(!(emulation_type & EMULTYPE_PF)))
>   		return false;
>   
>   	if (x86_page_table_writing_insn(ctxt))
> @@ -6830,8 +6832,19 @@ int x86_emulate_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
>   	}
>   
>   restart:
> -	/* Save the faulting GPA (cr2) in the address field */
> -	ctxt->exception.address = cr2_or_gpa;
> +	if (emulation_type & EMULTYPE_PF) {
> +		/* Save the faulting GPA (cr2) in the address field */
> +		ctxt->exception.address = cr2_or_gpa;
> +
> +		/* With shadow page tables, cr2 contains a GVA or nGPA. */
> +		if (vcpu->arch.mmu->direct_map) {
> +			vcpu->arch.gpa_available = true;
> +			vcpu->arch.gpa_val = cr2_or_gpa;
> +		}
> +	} else {
> +		/* Sanitize the address out of an abundance of paranoia. */
> +		ctxt->exception.address = 0;
> +	}
>   
>   	r = x86_emulate_insn(ctxt);
>   
> 

Re: [PATCH v2 1/3] KVM: x86: Add EMULTYPE_PF when emulation is triggered by a page fault

[Sean Christopherson]

On Wed, Feb 19, 2020 at 05:01:41PM +0800, Xiaoyao Li wrote:
> On 2/19/2020 7:03 AM, Sean Christopherson wrote:
> >Add a new emulation type flag to explicitly mark emulation related to a
> >page fault.  Move the propation of the GPA into the emulator from the
> >page fault handler into x86_emulate_instruction, using EMULTYPE_PF as an
> >indicator that cr2 is valid.  Similarly, don't propagate cr2 into the
> >exception.address when it's *not* valid.
> >
> >Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
> >---
> >  arch/x86/include/asm/kvm_host.h | 12 +++++++++---
> >  arch/x86/kvm/mmu/mmu.c          | 10 ++--------
> >  arch/x86/kvm/x86.c              | 25 +++++++++++++++++++------
> >  3 files changed, 30 insertions(+), 17 deletions(-)
> >
> >diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> >index 4dffbc10d3f8..10c1e8f472b6 100644
> >--- a/arch/x86/include/asm/kvm_host.h
> >+++ b/arch/x86/include/asm/kvm_host.h
> >@@ -1370,8 +1370,9 @@ extern u64 kvm_mce_cap_supported;
> >   *		   decode the instruction length.  For use *only* by
> >   *		   kvm_x86_ops->skip_emulated_instruction() implementations.
> >   *
> >- * EMULTYPE_ALLOW_RETRY - Set when the emulator should resume the guest to
> >- *			  retry native execution under certain conditions.
> >+ * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
> >+ *			     retry native execution under certain conditions,
> >+ *			     Can only be set in conjunction with EMULTYPE_PF.
> >   *
> >   * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
> >   *			     triggered by KVM's magic "force emulation" prefix,
> >@@ -1384,13 +1385,18 @@ extern u64 kvm_mce_cap_supported;
> >   *			backdoor emulation, which is opt in via module param.
> >   *			VMware backoor emulation handles select instructions
> >   *			and reinjects the #GP for all other cases.
> >+ *
> >+ * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
> >+ *		 case the CR2/GPA value pass on the stack is valid.
> >   */
> >  #define EMULTYPE_NO_DECODE	    (1 << 0)
> >  #define EMULTYPE_TRAP_UD	    (1 << 1)
> >  #define EMULTYPE_SKIP		    (1 << 2)
> >-#define EMULTYPE_ALLOW_RETRY	    (1 << 3)
> >+#define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)
> 
> How about naming it as EMULTYPE_PF_ALLOW_RETRY and exchanging the bit
> position with EMULTYPE_PF ?

Hmm, EMULTYPE_PF_ALLOW_RETRY does sound better.  I'm on the fence regarding
shuffling the bits.  If I were to shuffle the bits, I'd do a more thorough
reorder so that the #UD and #PF types are consecutive, e.g.

	#define EMULTYPE_NO_DECODE	    (1 << 0)
	#define EMULTYPE_TRAP_UD	    (1 << 1)
	#define EMULTYPE_TRAP_UD_FORCED	    (1 << 2)
	#define EMULTYPE_SKIP		    (1 << 3)
	#define EMULTYPE_VMWARE_GP	    (1 << 4)
	#define EMULTYPE_PF		    (1 << 5)
	#define EMULTYPE_PF_ALLOW_RETRY	    (1 << 6)

Part of me really wants to do that, the other part of me thinks it's
unnecessary thrash.

> >  #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
> >  #define EMULTYPE_VMWARE_GP	    (1 << 5)
> >+#define EMULTYPE_PF		    (1 << 6)
> >+
> >  int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
> >  int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
> >  					void *insn, int insn_len);
> >diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
> >index 7011a4e54866..258624d46588 100644
> >--- a/arch/x86/kvm/mmu/mmu.c
> >+++ b/arch/x86/kvm/mmu/mmu.c
> >@@ -5416,18 +5416,12 @@ EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page_virt);
> >  int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
> >  		       void *insn, int insn_len)
> >  {
> >-	int r, emulation_type = 0;
> >+	int r, emulation_type = EMULTYPE_PF;
> >  	bool direct = vcpu->arch.mmu->direct_map;
> >  	if (WARN_ON(!VALID_PAGE(vcpu->arch.mmu->root_hpa)))
> >  		return RET_PF_RETRY;
> >-	/* With shadow page tables, fault_address contains a GVA or nGPA.  */
> >-	if (vcpu->arch.mmu->direct_map) {
> >-		vcpu->arch.gpa_available = true;
> >-		vcpu->arch.gpa_val = cr2_or_gpa;
> >-	}
> >-
> >  	r = RET_PF_INVALID;
> >  	if (unlikely(error_code & PFERR_RSVD_MASK)) {
> >  		r = handle_mmio_page_fault(vcpu, cr2_or_gpa, direct);
> >@@ -5472,7 +5466,7 @@ int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
> >  	 * for L1 isn't going to magically fix whatever issue cause L2 to fail.
> >  	 */
> >  	if (!mmio_info_in_cache(vcpu, cr2_or_gpa, direct) && !is_guest_mode(vcpu))
> >-		emulation_type = EMULTYPE_ALLOW_RETRY;
> >+		emulation_type |= EMULTYPE_ALLOW_RETRY_PF;
> >  emulate:
> >  	/*
> >  	 * On AMD platforms, under certain conditions insn_len may be zero on #NPF.
> >diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> >index fbabb2f06273..92af6c5a69e3 100644
> >--- a/arch/x86/kvm/x86.c
> >+++ b/arch/x86/kvm/x86.c
> >@@ -6483,10 +6483,11 @@ static bool reexecute_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
> >  	gpa_t gpa = cr2_or_gpa;
> >  	kvm_pfn_t pfn;
> >-	if (!(emulation_type & EMULTYPE_ALLOW_RETRY))
> >+	if (!(emulation_type & EMULTYPE_ALLOW_RETRY_PF))
> >  		return false;
> >-	if (WARN_ON_ONCE(is_guest_mode(vcpu)))
> >+	if (WARN_ON_ONCE(is_guest_mode(vcpu)) ||
> >+	    WARN_ON_ONCE(!(emulation_type & EMULTYPE_PF)))
> >  		return false;
> >  	if (!vcpu->arch.mmu->direct_map) {
> >@@ -6574,10 +6575,11 @@ static bool retry_instruction(struct x86_emulate_ctxt *ctxt,
> >  	 */
> >  	vcpu->arch.last_retry_eip = vcpu->arch.last_retry_addr = 0;
> >-	if (!(emulation_type & EMULTYPE_ALLOW_RETRY))
> >+	if (!(emulation_type & EMULTYPE_ALLOW_RETRY_PF))
> >  		return false;
> >-	if (WARN_ON_ONCE(is_guest_mode(vcpu)))
> >+	if (WARN_ON_ONCE(is_guest_mode(vcpu)) ||
> >+	    WARN_ON_ONCE(!(emulation_type & EMULTYPE_PF)))
> >  		return false;
> >  	if (x86_page_table_writing_insn(ctxt))
> >@@ -6830,8 +6832,19 @@ int x86_emulate_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
> >  	}
> >  restart:
> >-	/* Save the faulting GPA (cr2) in the address field */
> >-	ctxt->exception.address = cr2_or_gpa;
> >+	if (emulation_type & EMULTYPE_PF) {
> >+		/* Save the faulting GPA (cr2) in the address field */
> >+		ctxt->exception.address = cr2_or_gpa;
> >+
> >+		/* With shadow page tables, cr2 contains a GVA or nGPA. */
> >+		if (vcpu->arch.mmu->direct_map) {
> >+			vcpu->arch.gpa_available = true;
> >+			vcpu->arch.gpa_val = cr2_or_gpa;
> >+		}
> >+	} else {
> >+		/* Sanitize the address out of an abundance of paranoia. */
> >+		ctxt->exception.address = 0;
> >+	}
> >  	r = x86_emulate_insn(ctxt);
> >
> 

Re: [PATCH v2 1/3] KVM: x86: Add EMULTYPE_PF when emulation is triggered by a page fault

[Xiaoyao Li]

On 2/21/2020 4:11 AM, Sean Christopherson wrote:
> On Wed, Feb 19, 2020 at 05:01:41PM +0800, Xiaoyao Li wrote:
>> On 2/19/2020 7:03 AM, Sean Christopherson wrote:
>>> Add a new emulation type flag to explicitly mark emulation related to a
>>> page fault.  Move the propation of the GPA into the emulator from the
>>> page fault handler into x86_emulate_instruction, using EMULTYPE_PF as an
>>> indicator that cr2 is valid.  Similarly, don't propagate cr2 into the
>>> exception.address when it's *not* valid.
>>>
>>> Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
>>> ---
>>>   arch/x86/include/asm/kvm_host.h | 12 +++++++++---
>>>   arch/x86/kvm/mmu/mmu.c          | 10 ++--------
>>>   arch/x86/kvm/x86.c              | 25 +++++++++++++++++++------
>>>   3 files changed, 30 insertions(+), 17 deletions(-)
>>>
>>> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
>>> index 4dffbc10d3f8..10c1e8f472b6 100644
>>> --- a/arch/x86/include/asm/kvm_host.h
>>> +++ b/arch/x86/include/asm/kvm_host.h
>>> @@ -1370,8 +1370,9 @@ extern u64 kvm_mce_cap_supported;
>>>    *		   decode the instruction length.  For use *only* by
>>>    *		   kvm_x86_ops->skip_emulated_instruction() implementations.
>>>    *
>>> - * EMULTYPE_ALLOW_RETRY - Set when the emulator should resume the guest to
>>> - *			  retry native execution under certain conditions.
>>> + * EMULTYPE_ALLOW_RETRY_PF - Set when the emulator should resume the guest to
>>> + *			     retry native execution under certain conditions,
>>> + *			     Can only be set in conjunction with EMULTYPE_PF.
>>>    *
>>>    * EMULTYPE_TRAP_UD_FORCED - Set when emulating an intercepted #UD that was
>>>    *			     triggered by KVM's magic "force emulation" prefix,
>>> @@ -1384,13 +1385,18 @@ extern u64 kvm_mce_cap_supported;
>>>    *			backdoor emulation, which is opt in via module param.
>>>    *			VMware backoor emulation handles select instructions
>>>    *			and reinjects the #GP for all other cases.
>>> + *
>>> + * EMULTYPE_PF - Set when emulating MMIO by way of an intercepted #PF, in which
>>> + *		 case the CR2/GPA value pass on the stack is valid.
>>>    */
>>>   #define EMULTYPE_NO_DECODE	    (1 << 0)
>>>   #define EMULTYPE_TRAP_UD	    (1 << 1)
>>>   #define EMULTYPE_SKIP		    (1 << 2)
>>> -#define EMULTYPE_ALLOW_RETRY	    (1 << 3)
>>> +#define EMULTYPE_ALLOW_RETRY_PF	    (1 << 3)
>>
>> How about naming it as EMULTYPE_PF_ALLOW_RETRY and exchanging the bit
>> position with EMULTYPE_PF ?
> 
> Hmm, EMULTYPE_PF_ALLOW_RETRY does sound better.  I'm on the fence regarding
> shuffling the bits.  If I were to shuffle the bits, I'd do a more thorough
> reorder so that the #UD and #PF types are consecutive, e.g.
> 
> 	#define EMULTYPE_NO_DECODE	    (1 << 0)
> 	#define EMULTYPE_TRAP_UD	    (1 << 1)
> 	#define EMULTYPE_TRAP_UD_FORCED	    (1 << 2)
> 	#define EMULTYPE_SKIP		    (1 << 3)
> 	#define EMULTYPE_VMWARE_GP	    (1 << 4)
> 	#define EMULTYPE_PF		    (1 << 5)
> 	#define EMULTYPE_PF_ALLOW_RETRY	    (1 << 6)
> 
> Part of me really wants to do that, the other part of me thinks it's
> unnecessary thrash.
>

I'm fine with thorough reorder, it helps read the codes.
It's up to Paolo, anyway.

>>>   #define EMULTYPE_TRAP_UD_FORCED	    (1 << 4)
>>>   #define EMULTYPE_VMWARE_GP	    (1 << 5)
>>> +#define EMULTYPE_PF		    (1 << 6)
>>> +
>>>   int kvm_emulate_instruction(struct kvm_vcpu *vcpu, int emulation_type);
>>>   int kvm_emulate_instruction_from_buffer(struct kvm_vcpu *vcpu,
>>>   					void *insn, int insn_len);
>>> diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
>>> index 7011a4e54866..258624d46588 100644
>>> --- a/arch/x86/kvm/mmu/mmu.c
>>> +++ b/arch/x86/kvm/mmu/mmu.c
>>> @@ -5416,18 +5416,12 @@ EXPORT_SYMBOL_GPL(kvm_mmu_unprotect_page_virt);
>>>   int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
>>>   		       void *insn, int insn_len)
>>>   {
>>> -	int r, emulation_type = 0;
>>> +	int r, emulation_type = EMULTYPE_PF;
>>>   	bool direct = vcpu->arch.mmu->direct_map;
>>>   	if (WARN_ON(!VALID_PAGE(vcpu->arch.mmu->root_hpa)))
>>>   		return RET_PF_RETRY;
>>> -	/* With shadow page tables, fault_address contains a GVA or nGPA.  */
>>> -	if (vcpu->arch.mmu->direct_map) {
>>> -		vcpu->arch.gpa_available = true;
>>> -		vcpu->arch.gpa_val = cr2_or_gpa;
>>> -	}
>>> -
>>>   	r = RET_PF_INVALID;
>>>   	if (unlikely(error_code & PFERR_RSVD_MASK)) {
>>>   		r = handle_mmio_page_fault(vcpu, cr2_or_gpa, direct);
>>> @@ -5472,7 +5466,7 @@ int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code,
>>>   	 * for L1 isn't going to magically fix whatever issue cause L2 to fail.
>>>   	 */
>>>   	if (!mmio_info_in_cache(vcpu, cr2_or_gpa, direct) && !is_guest_mode(vcpu))
>>> -		emulation_type = EMULTYPE_ALLOW_RETRY;
>>> +		emulation_type |= EMULTYPE_ALLOW_RETRY_PF;
>>>   emulate:
>>>   	/*
>>>   	 * On AMD platforms, under certain conditions insn_len may be zero on #NPF.
>>> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
>>> index fbabb2f06273..92af6c5a69e3 100644
>>> --- a/arch/x86/kvm/x86.c
>>> +++ b/arch/x86/kvm/x86.c
>>> @@ -6483,10 +6483,11 @@ static bool reexecute_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
>>>   	gpa_t gpa = cr2_or_gpa;
>>>   	kvm_pfn_t pfn;
>>> -	if (!(emulation_type & EMULTYPE_ALLOW_RETRY))
>>> +	if (!(emulation_type & EMULTYPE_ALLOW_RETRY_PF))
>>>   		return false;
>>> -	if (WARN_ON_ONCE(is_guest_mode(vcpu)))
>>> +	if (WARN_ON_ONCE(is_guest_mode(vcpu)) ||
>>> +	    WARN_ON_ONCE(!(emulation_type & EMULTYPE_PF)))
>>>   		return false;
>>>   	if (!vcpu->arch.mmu->direct_map) {
>>> @@ -6574,10 +6575,11 @@ static bool retry_instruction(struct x86_emulate_ctxt *ctxt,
>>>   	 */
>>>   	vcpu->arch.last_retry_eip = vcpu->arch.last_retry_addr = 0;
>>> -	if (!(emulation_type & EMULTYPE_ALLOW_RETRY))
>>> +	if (!(emulation_type & EMULTYPE_ALLOW_RETRY_PF))
>>>   		return false;
>>> -	if (WARN_ON_ONCE(is_guest_mode(vcpu)))
>>> +	if (WARN_ON_ONCE(is_guest_mode(vcpu)) ||
>>> +	    WARN_ON_ONCE(!(emulation_type & EMULTYPE_PF)))
>>>   		return false;
>>>   	if (x86_page_table_writing_insn(ctxt))
>>> @@ -6830,8 +6832,19 @@ int x86_emulate_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
>>>   	}
>>>   restart:
>>> -	/* Save the faulting GPA (cr2) in the address field */
>>> -	ctxt->exception.address = cr2_or_gpa;
>>> +	if (emulation_type & EMULTYPE_PF) {
>>> +		/* Save the faulting GPA (cr2) in the address field */
>>> +		ctxt->exception.address = cr2_or_gpa;
>>> +
>>> +		/* With shadow page tables, cr2 contains a GVA or nGPA. */
>>> +		if (vcpu->arch.mmu->direct_map) {
>>> +			vcpu->arch.gpa_available = true;
>>> +			vcpu->arch.gpa_val = cr2_or_gpa;
>>> +		}
>>> +	} else {
>>> +		/* Sanitize the address out of an abundance of paranoia. */
>>> +		ctxt->exception.address = 0;
>>> +	}
>>>   	r = x86_emulate_insn(ctxt);
>>>
>>

Re: [PATCH v2 1/3] KVM: x86: Add EMULTYPE_PF when emulation is triggered by a page fault

[Paolo Bonzini]

On 20/02/20 21:11, Sean Christopherson wrote:
>> How about naming it as EMULTYPE_PF_ALLOW_RETRY and exchanging the bit
>> position with EMULTYPE_PF ?
> Hmm, EMULTYPE_PF_ALLOW_RETRY does sound better.  I'm on the fence regarding
> shuffling the bits.  If I were to shuffle the bits, I'd do a more thorough
> reorder so that the #UD and #PF types are consecutive, e.g.

Let's just change the name, I can do it.

Paolo

[PATCH v2 2/3] KVM: x86: Move gpa_val and gpa_available into the emulator context

[Sean Christopherson]

Move the GPA tracking into the emulator context now that the context is
guaranteed to be initialized via __init_emulate_ctxt() prior to
dereferencing gpa_{available,val}, i.e. now that seeing a stale
gpa_available will also trigger a WARN due to an invalid context.

Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
---
 arch/x86/include/asm/kvm_emulate.h |  4 ++++
 arch/x86/include/asm/kvm_host.h    |  4 ----
 arch/x86/kvm/x86.c                 | 13 ++++++-------
 3 files changed, 10 insertions(+), 11 deletions(-)

diff --git a/arch/x86/include/asm/kvm_emulate.h b/arch/x86/include/asm/kvm_emulate.h
index 03946eb3e2b9..a4ef19a6e612 100644
--- a/arch/x86/include/asm/kvm_emulate.h
+++ b/arch/x86/include/asm/kvm_emulate.h
@@ -311,6 +311,10 @@ struct x86_emulate_ctxt {
 	bool have_exception;
 	struct x86_exception exception;
 
+	/* GPA available */
+	bool gpa_available;
+	gpa_t gpa_val;
+
 	/*
 	 * decode cache
 	 */
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 10c1e8f472b6..9c79c41eb5f6 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -798,10 +798,6 @@ struct kvm_vcpu_arch {
 	int pending_ioapic_eoi;
 	int pending_external_vector;
 
-	/* GPA available */
-	bool gpa_available;
-	gpa_t gpa_val;
-
 	/* be preempted when it's in kernel-mode(cpl=0) */
 	bool preempted_in_kernel;
 
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 92af6c5a69e3..f88b72932c35 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -5736,10 +5736,9 @@ static int emulator_read_write_onepage(unsigned long addr, void *val,
 	 * operation using rep will only have the initial GPA from the NPF
 	 * occurred.
 	 */
-	if (vcpu->arch.gpa_available &&
-	    emulator_can_use_gpa(ctxt) &&
-	    (addr & ~PAGE_MASK) == (vcpu->arch.gpa_val & ~PAGE_MASK)) {
-		gpa = vcpu->arch.gpa_val;
+	if (ctxt->gpa_available && emulator_can_use_gpa(ctxt) &&
+	    (addr & ~PAGE_MASK) == (ctxt->gpa_val & ~PAGE_MASK)) {
+		gpa = ctxt->gpa_val;
 		ret = vcpu_is_mmio_gpa(vcpu, addr, gpa, write);
 	} else {
 		ret = vcpu_mmio_gva_to_gpa(vcpu, addr, &gpa, exception, write);
@@ -6408,6 +6407,7 @@ static void init_emulate_ctxt(struct kvm_vcpu *vcpu)
 
 	kvm_x86_ops->get_cs_db_l_bits(vcpu, &cs_db, &cs_l);
 
+	ctxt->gpa_available = false;
 	ctxt->eflags = kvm_get_rflags(vcpu);
 	ctxt->tf = (ctxt->eflags & X86_EFLAGS_TF) != 0;
 
@@ -6838,8 +6838,8 @@ int x86_emulate_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa,
 
 		/* With shadow page tables, cr2 contains a GVA or nGPA. */
 		if (vcpu->arch.mmu->direct_map) {
-			vcpu->arch.gpa_available = true;
-			vcpu->arch.gpa_val = cr2_or_gpa;
+			ctxt->gpa_available = true;
+			ctxt->gpa_val = cr2_or_gpa;
 		}
 	} else {
 		/* Sanitize the address out of an abundance of paranoia. */
@@ -8443,7 +8443,6 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu)
 	if (vcpu->arch.apic_attention)
 		kvm_lapic_sync_from_vapic(vcpu);
 
-	vcpu->arch.gpa_available = false;
 	r = kvm_x86_ops->handle_exit(vcpu, exit_fastpath);
 	return r;
 
-- 
2.24.1

[PATCH v2 3/3] KVM: x86: Move #PF retry tracking variables into emulation context

[Sean Christopherson]

Move last_retry_eip and last_retry_addr into the emulation context as
they are specific to retrying an instruction after emulation failure.

Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
---
 arch/x86/include/asm/kvm_emulate.h |  4 ++++
 arch/x86/include/asm/kvm_host.h    |  3 ---
 arch/x86/kvm/x86.c                 | 11 ++++++-----
 3 files changed, 10 insertions(+), 8 deletions(-)

diff --git a/arch/x86/include/asm/kvm_emulate.h b/arch/x86/include/asm/kvm_emulate.h
index a4ef19a6e612..a26c8de414e8 100644
--- a/arch/x86/include/asm/kvm_emulate.h
+++ b/arch/x86/include/asm/kvm_emulate.h
@@ -315,6 +315,10 @@ struct x86_emulate_ctxt {
 	bool gpa_available;
 	gpa_t gpa_val;
 
+	/* Track EIP and CR2/GPA when retrying faulting instruction on #PF. */
+	unsigned long last_retry_eip;
+	unsigned long last_retry_addr;
+
 	/*
 	 * decode cache
 	 */
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 9c79c41eb5f6..6312ea32bb41 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -752,9 +752,6 @@ struct kvm_vcpu_arch {
 
 	cpumask_var_t wbinvd_dirty_mask;
 
-	unsigned long last_retry_eip;
-	unsigned long last_retry_addr;
-
 	struct {
 		bool halted;
 		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index f88b72932c35..d19eb776f297 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -6407,6 +6407,7 @@ static void init_emulate_ctxt(struct kvm_vcpu *vcpu)
 
 	kvm_x86_ops->get_cs_db_l_bits(vcpu, &cs_db, &cs_l);
 
+	/* last_retry_{eip,addr} are persistent and must not be init'd here. */
 	ctxt->gpa_available = false;
 	ctxt->eflags = kvm_get_rflags(vcpu);
 	ctxt->tf = (ctxt->eflags & X86_EFLAGS_TF) != 0;
@@ -6557,8 +6558,8 @@ static bool retry_instruction(struct x86_emulate_ctxt *ctxt,
 	struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
 	unsigned long last_retry_eip, last_retry_addr, gpa = cr2_or_gpa;
 
-	last_retry_eip = vcpu->arch.last_retry_eip;
-	last_retry_addr = vcpu->arch.last_retry_addr;
+	last_retry_eip = ctxt->last_retry_eip;
+	last_retry_addr = ctxt->last_retry_addr;
 
 	/*
 	 * If the emulation is caused by #PF and it is non-page_table
@@ -6573,7 +6574,7 @@ static bool retry_instruction(struct x86_emulate_ctxt *ctxt,
 	 * and the address again, we can break out of the potential infinite
 	 * loop.
 	 */
-	vcpu->arch.last_retry_eip = vcpu->arch.last_retry_addr = 0;
+	ctxt->last_retry_eip = ctxt->last_retry_addr = 0;
 
 	if (!(emulation_type & EMULTYPE_ALLOW_RETRY_PF))
 		return false;
@@ -6588,8 +6589,8 @@ static bool retry_instruction(struct x86_emulate_ctxt *ctxt,
 	if (ctxt->eip == last_retry_eip && last_retry_addr == cr2_or_gpa)
 		return false;
 
-	vcpu->arch.last_retry_eip = ctxt->eip;
-	vcpu->arch.last_retry_addr = cr2_or_gpa;
+	ctxt->last_retry_eip = ctxt->eip;
+	ctxt->last_retry_addr = cr2_or_gpa;
 
 	if (!vcpu->arch.mmu->direct_map)
 		gpa = kvm_mmu_gva_to_gpa_write(vcpu, cr2_or_gpa, NULL);
-- 
2.24.1

Re: [PATCH v2 3/3] KVM: x86: Move #PF retry tracking variables into emulation context

[Paolo Bonzini]

On 19/02/20 00:03, Sean Christopherson wrote:
> Move last_retry_eip and last_retry_addr into the emulation context as
> they are specific to retrying an instruction after emulation failure.
> 
> Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>

I'm not sure about this, since it's not used by emulate.c.  The other
two patches are good.

Paolo

> ---
>  arch/x86/include/asm/kvm_emulate.h |  4 ++++
>  arch/x86/include/asm/kvm_host.h    |  3 ---
>  arch/x86/kvm/x86.c                 | 11 ++++++-----
>  3 files changed, 10 insertions(+), 8 deletions(-)
> 
> diff --git a/arch/x86/include/asm/kvm_emulate.h b/arch/x86/include/asm/kvm_emulate.h
> index a4ef19a6e612..a26c8de414e8 100644
> --- a/arch/x86/include/asm/kvm_emulate.h
> +++ b/arch/x86/include/asm/kvm_emulate.h
> @@ -315,6 +315,10 @@ struct x86_emulate_ctxt {
>  	bool gpa_available;
>  	gpa_t gpa_val;
>  
> +	/* Track EIP and CR2/GPA when retrying faulting instruction on #PF. */
> +	unsigned long last_retry_eip;
> +	unsigned long last_retry_addr;
> +
>  	/*
>  	 * decode cache
>  	 */
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index 9c79c41eb5f6..6312ea32bb41 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -752,9 +752,6 @@ struct kvm_vcpu_arch {
>  
>  	cpumask_var_t wbinvd_dirty_mask;
>  
> -	unsigned long last_retry_eip;
> -	unsigned long last_retry_addr;
> -
>  	struct {
>  		bool halted;
>  		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index f88b72932c35..d19eb776f297 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -6407,6 +6407,7 @@ static void init_emulate_ctxt(struct kvm_vcpu *vcpu)
>  
>  	kvm_x86_ops->get_cs_db_l_bits(vcpu, &cs_db, &cs_l);
>  
> +	/* last_retry_{eip,addr} are persistent and must not be init'd here. */
>  	ctxt->gpa_available = false;
>  	ctxt->eflags = kvm_get_rflags(vcpu);
>  	ctxt->tf = (ctxt->eflags & X86_EFLAGS_TF) != 0;
> @@ -6557,8 +6558,8 @@ static bool retry_instruction(struct x86_emulate_ctxt *ctxt,
>  	struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
>  	unsigned long last_retry_eip, last_retry_addr, gpa = cr2_or_gpa;
>  
> -	last_retry_eip = vcpu->arch.last_retry_eip;
> -	last_retry_addr = vcpu->arch.last_retry_addr;
> +	last_retry_eip = ctxt->last_retry_eip;
> +	last_retry_addr = ctxt->last_retry_addr;
>  
>  	/*
>  	 * If the emulation is caused by #PF and it is non-page_table
> @@ -6573,7 +6574,7 @@ static bool retry_instruction(struct x86_emulate_ctxt *ctxt,
>  	 * and the address again, we can break out of the potential infinite
>  	 * loop.
>  	 */
> -	vcpu->arch.last_retry_eip = vcpu->arch.last_retry_addr = 0;
> +	ctxt->last_retry_eip = ctxt->last_retry_addr = 0;
>  
>  	if (!(emulation_type & EMULTYPE_ALLOW_RETRY_PF))
>  		return false;
> @@ -6588,8 +6589,8 @@ static bool retry_instruction(struct x86_emulate_ctxt *ctxt,
>  	if (ctxt->eip == last_retry_eip && last_retry_addr == cr2_or_gpa)
>  		return false;
>  
> -	vcpu->arch.last_retry_eip = ctxt->eip;
> -	vcpu->arch.last_retry_addr = cr2_or_gpa;
> +	ctxt->last_retry_eip = ctxt->eip;
> +	ctxt->last_retry_addr = cr2_or_gpa;
>  
>  	if (!vcpu->arch.mmu->direct_map)
>  		gpa = kvm_mmu_gva_to_gpa_write(vcpu, cr2_or_gpa, NULL);
> 

Re: [PATCH v2 3/3] KVM: x86: Move #PF retry tracking variables into emulation context

[Sean Christopherson]

On Wed, Feb 19, 2020 at 09:13:00AM +0100, Paolo Bonzini wrote:
> On 19/02/20 00:03, Sean Christopherson wrote:
> > Move last_retry_eip and last_retry_addr into the emulation context as
> > they are specific to retrying an instruction after emulation failure.
> > 
> > Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
> 
> I'm not sure about this, since it's not used by emulate.c.  The other
> two patches are good.

The easy solution to that is to move retry_instruction() into emulate.c.
That would also allow making x86_page_table_writing_insn() static.  All
other functions invoked from retry_instruction() are exposed via kvm_host.h.

Moving last_retry_* into the emulator context hopefully makes it more clear
that this code is the only user of the variables, e.g. last_retry_eip can't
be set by some other non-emulator flow.

	if (ctxt->eip == last_retry_eip && last_retry_addr == cr2)
		return false;

	vcpu->arch.last_retry_eip = ctxt->eip;
	vcpu->arch.last_retry_addr = cr2;

> > ---
> >  arch/x86/include/asm/kvm_emulate.h |  4 ++++
> >  arch/x86/include/asm/kvm_host.h    |  3 ---
> >  arch/x86/kvm/x86.c                 | 11 ++++++-----
> >  3 files changed, 10 insertions(+), 8 deletions(-)
> > 
> > diff --git a/arch/x86/include/asm/kvm_emulate.h b/arch/x86/include/asm/kvm_emulate.h
> > index a4ef19a6e612..a26c8de414e8 100644
> > --- a/arch/x86/include/asm/kvm_emulate.h
> > +++ b/arch/x86/include/asm/kvm_emulate.h
> > @@ -315,6 +315,10 @@ struct x86_emulate_ctxt {
> >  	bool gpa_available;
> >  	gpa_t gpa_val;
> >  
> > +	/* Track EIP and CR2/GPA when retrying faulting instruction on #PF. */
> > +	unsigned long last_retry_eip;
> > +	unsigned long last_retry_addr;
> > +
> >  	/*
> >  	 * decode cache
> >  	 */
> > diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> > index 9c79c41eb5f6..6312ea32bb41 100644
> > --- a/arch/x86/include/asm/kvm_host.h
> > +++ b/arch/x86/include/asm/kvm_host.h
> > @@ -752,9 +752,6 @@ struct kvm_vcpu_arch {
> >  
> >  	cpumask_var_t wbinvd_dirty_mask;
> >  
> > -	unsigned long last_retry_eip;
> > -	unsigned long last_retry_addr;
> > -
> >  	struct {
> >  		bool halted;
> >  		gfn_t gfns[roundup_pow_of_two(ASYNC_PF_PER_VCPU)];
> > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> > index f88b72932c35..d19eb776f297 100644
> > --- a/arch/x86/kvm/x86.c
> > +++ b/arch/x86/kvm/x86.c
> > @@ -6407,6 +6407,7 @@ static void init_emulate_ctxt(struct kvm_vcpu *vcpu)
> >  
> >  	kvm_x86_ops->get_cs_db_l_bits(vcpu, &cs_db, &cs_l);
> >  
> > +	/* last_retry_{eip,addr} are persistent and must not be init'd here. */
> >  	ctxt->gpa_available = false;
> >  	ctxt->eflags = kvm_get_rflags(vcpu);
> >  	ctxt->tf = (ctxt->eflags & X86_EFLAGS_TF) != 0;
> > @@ -6557,8 +6558,8 @@ static bool retry_instruction(struct x86_emulate_ctxt *ctxt,
> >  	struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt);
> >  	unsigned long last_retry_eip, last_retry_addr, gpa = cr2_or_gpa;
> >  
> > -	last_retry_eip = vcpu->arch.last_retry_eip;
> > -	last_retry_addr = vcpu->arch.last_retry_addr;
> > +	last_retry_eip = ctxt->last_retry_eip;
> > +	last_retry_addr = ctxt->last_retry_addr;
> >  
> >  	/*
> >  	 * If the emulation is caused by #PF and it is non-page_table
> > @@ -6573,7 +6574,7 @@ static bool retry_instruction(struct x86_emulate_ctxt *ctxt,
> >  	 * and the address again, we can break out of the potential infinite
> >  	 * loop.
> >  	 */
> > -	vcpu->arch.last_retry_eip = vcpu->arch.last_retry_addr = 0;
> > +	ctxt->last_retry_eip = ctxt->last_retry_addr = 0;
> >  
> >  	if (!(emulation_type & EMULTYPE_ALLOW_RETRY_PF))
> >  		return false;
> > @@ -6588,8 +6589,8 @@ static bool retry_instruction(struct x86_emulate_ctxt *ctxt,
> >  	if (ctxt->eip == last_retry_eip && last_retry_addr == cr2_or_gpa)
> >  		return false;
> >  
> > -	vcpu->arch.last_retry_eip = ctxt->eip;
> > -	vcpu->arch.last_retry_addr = cr2_or_gpa;
> > +	ctxt->last_retry_eip = ctxt->eip;
> > +	ctxt->last_retry_addr = cr2_or_gpa;
> >  
> >  	if (!vcpu->arch.mmu->direct_map)
> >  		gpa = kvm_mmu_gva_to_gpa_write(vcpu, cr2_or_gpa, NULL);
> > 
> 

Re: [PATCH v2 3/3] KVM: x86: Move #PF retry tracking variables into emulation context

[Paolo Bonzini]

On 19/02/20 16:16, Sean Christopherson wrote:
> The easy solution to that is to move retry_instruction() into emulate.c.
> That would also allow making x86_page_table_writing_insn() static.  All
> other functions invoked from retry_instruction() are exposed via kvm_host.h.

emulate.c is supposed to invoke no (or almost no) function outside the
ctxt->ops struct.  In particular, retry_instruction() invokes
kvm_mmu_gva_to_gpa_write and kvm_mmu_unprotect_page.

Paolo

Re: [PATCH v2 3/3] KVM: x86: Move #PF retry tracking variables into emulation context

[Sean Christopherson]

On Fri, Feb 21, 2020 at 06:14:03PM +0100, Paolo Bonzini wrote:
> On 19/02/20 16:16, Sean Christopherson wrote:
> > The easy solution to that is to move retry_instruction() into emulate.c.
> > That would also allow making x86_page_table_writing_insn() static.  All
> > other functions invoked from retry_instruction() are exposed via kvm_host.h.
> 
> emulate.c is supposed to invoke no (or almost no) function outside the
> ctxt->ops struct.  In particular, retry_instruction() invokes
> kvm_mmu_gva_to_gpa_write and kvm_mmu_unprotect_page.

Ah, right.  We could split the logic, e.g.

	if (x86_retry_pf_instruction(ctxt, cr2_or_gpa, emulation_type)) {
		gpa_t = gpa = cr2_or_gpa;

		if (!vcpu->arch.mmu->direct_map)
			gpa = kvm_mmu_gva_to_gpa_write(vcpu, cr2_or_gpa, NULL);

		kvm_mmu_unprotect_page(vcpu->kvm, gpa_to_gfn(gpa));
		return 1;
	}

but that's probably a net negative in terms of clarity.  And there's also
vcpu->arch.write_fault_to_shadow_pgtable, which is consumed only by
reexecute_instruction(), and I 100% agree that that variable should stay
in vcpu->arch.  Moving one flag used to retry #PF instructions and not the
other would be weird.

That was a long winded way of saying I agree we should drop this patch :-)