Re: [PATCH v4 3/3] KVM/arm/arm64: enable enhanced armv8 fp/simd lazy switch

[Mario Smarduch <m.smarduch@samsung.com>]

On 12/3/2015 8:13 AM, Marc Zyngier wrote:
> On 14/11/15 22:12, Mario Smarduch wrote:
>> This patch tracks armv7 and armv8 fp/simd hardware state with a vcpu lazy flag.
>> On vcpu_load for 32 bit guests enable FP access, and later enable fp/simd
>> trapping for 32 and 64 bit guests if lazy flag is not set. On first fp/simd 
>> access trap to handler to save host and restore guest context, disable 
>> trapping and set vcpu lazy flag. On vcpu_put if flag is set save guest and 
>> restore host context and also save guest fpexc register.
>>
>> Signed-off-by: Mario Smarduch <m.smarduch@samsung.com>
>> ---
>>  arch/arm/include/asm/kvm_host.h   |  3 ++
>>  arch/arm/kvm/arm.c                | 18 +++++++--
>>  arch/arm64/include/asm/kvm_asm.h  |  2 +
>>  arch/arm64/include/asm/kvm_host.h | 17 +++++++-
>>  arch/arm64/kernel/asm-offsets.c   |  1 +
>>  arch/arm64/kvm/hyp.S              | 83 +++++++++++++++++++++++++--------------
>>  6 files changed, 89 insertions(+), 35 deletions(-)
>>
>> diff --git a/arch/arm/include/asm/kvm_host.h b/arch/arm/include/asm/kvm_host.h
>> index 8fc7a59..6960ff2 100644
>> --- a/arch/arm/include/asm/kvm_host.h
>> +++ b/arch/arm/include/asm/kvm_host.h
>> @@ -40,6 +40,8 @@
>>  
>>  #define KVM_MAX_VCPUS VGIC_V2_MAX_CPUS
>>  
>> +#define kvm_guest_is32bit(vcpu)	true
> 
> This should be defined as an inline function, and placed in
> asm/kvm_emulate.h, probably renamed as kvm_guest_vcpu_is_32bit.

Will do, this header file should also resolve my problems in armv7.
> 
>> +
>>  /*
>>   * Reads the host FPEXC register, saves to vcpu context and enables the
>>   * FPEXC.
>> @@ -260,6 +262,7 @@ void kvm_mmu_wp_memory_region(struct kvm *kvm, int slot);
>>  
>>  struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr);
>>  void kvm_restore_host_vfp_state(struct kvm_vcpu *);
>> +static inline void kvm_save_guest_fpexc(struct kvm_vcpu *vcpu) {}
>>  
>>  static inline void kvm_arch_hardware_disable(void) {}
>>  static inline void kvm_arch_hardware_unsetup(void) {}
>> diff --git a/arch/arm/kvm/arm.c b/arch/arm/kvm/arm.c
>> index cfc348a..7a20530 100644
>> --- a/arch/arm/kvm/arm.c
>> +++ b/arch/arm/kvm/arm.c
>> @@ -292,8 +292,12 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
>>  
>>  	kvm_arm_set_running_vcpu(vcpu);
>>  
>> -	/* Save and enable FPEXC before we load guest context */
>> -	kvm_enable_fpexc(vcpu);
>> +	/*
>> +	 * For 32bit guest executing on arm64, enable fp/simd access in
>> +	 * EL2. On arm32 save host fpexc and then enable fp/simd access.
>> +	 */
>> +	if (kvm_guest_is32bit(vcpu))
>> +		kvm_enable_fpexc(vcpu);
>>  }
>>  
>>  void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
>> @@ -301,10 +305,18 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
>>  	/* If the fp/simd registers are dirty save guest, restore host. */
>>  	if (vcpu->arch.vfp_dirty) {
>>  		kvm_restore_host_vfp_state(vcpu);
>> +
>> +		/*
>> +		 * For 32bit guest on arm64 save the guest fpexc register
>> +		 * in EL2 mode.
>> +		 */
>> +		if (kvm_guest_is32bit(vcpu))
>> +			kvm_save_guest_fpexc(vcpu);
>> +
>>  		vcpu->arch.vfp_dirty = 0;
>>  	}
>>  
>> -	/* Restore host FPEXC trashed in vcpu_load */
>> +	/* For arm32 restore host FPEXC trashed in vcpu_load. */
>>  	kvm_restore_host_fpexc(vcpu);
>>  
>>  	/*
>> diff --git a/arch/arm64/include/asm/kvm_asm.h b/arch/arm64/include/asm/kvm_asm.h
>> index 5e37710..c589ca9 100644
>> --- a/arch/arm64/include/asm/kvm_asm.h
>> +++ b/arch/arm64/include/asm/kvm_asm.h
>> @@ -117,6 +117,8 @@ extern char __kvm_hyp_vector[];
>>  extern void __kvm_flush_vm_context(void);
>>  extern void __kvm_tlb_flush_vmid_ipa(struct kvm *kvm, phys_addr_t ipa);
>>  extern void __kvm_tlb_flush_vmid(struct kvm *kvm);
>> +extern void __kvm_enable_fpexc32(void);
>> +extern void __kvm_save_fpexc32(struct kvm_vcpu *vcpu);
>>  
>>  extern int __kvm_vcpu_run(struct kvm_vcpu *vcpu);
>>  
>> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
>> index 83e65dd..6e2d6b5 100644
>> --- a/arch/arm64/include/asm/kvm_host.h
>> +++ b/arch/arm64/include/asm/kvm_host.h
>> @@ -41,6 +41,8 @@
>>  
>>  #define KVM_VCPU_MAX_FEATURES 3
>>  
>> +#define kvm_guest_is32bit(vcpu)	(!(vcpu->arch.hcr_el2 & HCR_RW))
>> +
>>  int __attribute_const__ kvm_target_cpu(void);
>>  int kvm_reset_vcpu(struct kvm_vcpu *vcpu);
>>  int kvm_arch_dev_ioctl_check_extension(long ext);
>> @@ -251,9 +253,20 @@ static inline void kvm_arch_hardware_unsetup(void) {}
>>  static inline void kvm_arch_sync_events(struct kvm *kvm) {}
>>  static inline void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu) {}
>>  static inline void kvm_arch_sched_in(struct kvm_vcpu *vcpu, int cpu) {}
>> -static inline void kvm_enable_fpexc(struct kvm_vcpu *vcpu) {}
>> -static inline void kvm_restore_host_vfp_state(struct kvm_vcpu *vcpu) {}
>> +
>> +static inline void kvm_enable_fpexc(struct kvm_vcpu *vcpu)
>> +{
>> +	/* Enable FP/SIMD access from EL2 mode*/
>> +	kvm_call_hyp(__kvm_enable_fpexc32);
>> +}
>> +
>> +static inline void kvm_save_guest_fpexc(struct kvm_vcpu *vcpu)
>> +{
>> +	/* Save FPEXEC32_EL2 in EL2 mode */
>> +	kvm_call_hyp(__kvm_save_fpexc32, vcpu);
>> +}
>>  static inline void kvm_restore_host_fpexc(struct kvm_vcpu *vcpu) {}
>> +void kvm_restore_host_vfp_state(struct kvm_vcpu *vcpu);
>>  
>>  void kvm_arm_init_debug(void);
>>  void kvm_arm_setup_debug(struct kvm_vcpu *vcpu);
>> diff --git a/arch/arm64/kernel/asm-offsets.c b/arch/arm64/kernel/asm-offsets.c
>> index 8d89cf8..c9c5242 100644
>> --- a/arch/arm64/kernel/asm-offsets.c
>> +++ b/arch/arm64/kernel/asm-offsets.c
>> @@ -124,6 +124,7 @@ int main(void)
>>    DEFINE(VCPU_HCR_EL2,		offsetof(struct kvm_vcpu, arch.hcr_el2));
>>    DEFINE(VCPU_MDCR_EL2,	offsetof(struct kvm_vcpu, arch.mdcr_el2));
>>    DEFINE(VCPU_IRQ_LINES,	offsetof(struct kvm_vcpu, arch.irq_lines));
>> +  DEFINE(VCPU_VFP_DIRTY,	offsetof(struct kvm_vcpu, arch.vfp_dirty));
>>    DEFINE(VCPU_HOST_CONTEXT,	offsetof(struct kvm_vcpu, arch.host_cpu_context));
>>    DEFINE(VCPU_HOST_DEBUG_STATE, offsetof(struct kvm_vcpu, arch.host_debug_state));
>>    DEFINE(VCPU_TIMER_CNTV_CTL,	offsetof(struct kvm_vcpu, arch.timer_cpu.cntv_ctl));
>> diff --git a/arch/arm64/kvm/hyp.S b/arch/arm64/kvm/hyp.S
>> index e583613..06efb93 100644
>> --- a/arch/arm64/kvm/hyp.S
>> +++ b/arch/arm64/kvm/hyp.S
>> @@ -36,6 +36,23 @@
>>  #define CPU_SYSREG_OFFSET(x)	(CPU_SYSREGS + 8*x)
>>  
>>  	.text
>> +/**
>> + * void kvm_restore_host_vfp_state(struct vcpu *vcpu) -
>> + *	This function saves the guest, restores host, called from host.
>> + */
>> +ENTRY(kvm_restore_host_vfp_state)
>> +	push	xzr, lr
>> +
>> +	add	x2, x0, #VCPU_CONTEXT
>> +	bl __save_fpsimd
>> +
>> +	ldr	x2, [x0, #VCPU_HOST_CONTEXT]
>> +	bl __restore_fpsimd
>> +
>> +	pop	xzr, lr
>> +	ret
>> +ENDPROC(kvm_restore_host_vfp_state)
>> +
> 
> Same here. Don't mix HYP and non-HYP in the same file.
Yep got it.

Thnaks.
> 
>>  	.pushsection	.hyp.text, "ax"
>>  	.align	PAGE_SHIFT
>>  
>> @@ -385,14 +402,6 @@
>>  	tbz	\tmp, #KVM_ARM64_DEBUG_DIRTY_SHIFT, \target
>>  .endm
>>  
>> -/*
>> - * Branch to target if CPTR_EL2.TFP bit is set (VFP/SIMD trapping enabled)
>> - */
>> -.macro skip_fpsimd_state tmp, target
>> -	mrs	\tmp, cptr_el2
>> -	tbnz	\tmp, #CPTR_EL2_TFP_SHIFT, \target
>> -.endm
>> -
>>  .macro compute_debug_state target
>>  	// Compute debug state: If any of KDE, MDE or KVM_ARM64_DEBUG_DIRTY
>>  	// is set, we do a full save/restore cycle and disable trapping.
>> @@ -433,10 +442,6 @@
>>  	mrs	x5, ifsr32_el2
>>  	stp	x4, x5, [x3]
>>  
>> -	skip_fpsimd_state x8, 2f
>> -	mrs	x6, fpexc32_el2
>> -	str	x6, [x3, #16]
>> -2:
>>  	skip_debug_state x8, 1f
>>  	mrs	x7, dbgvcr32_el2
>>  	str	x7, [x3, #24]
>> @@ -468,21 +473,13 @@
>>  .macro activate_traps
>>  	ldr     x2, [x0, #VCPU_HCR_EL2]
>>  
>> -	/*
>> -	 * We are about to set CPTR_EL2.TFP to trap all floating point
>> -	 * register accesses to EL2, however, the ARM ARM clearly states that
>> -	 * traps are only taken to EL2 if the operation would not otherwise
>> -	 * trap to EL1.  Therefore, always make sure that for 32-bit guests,
>> -	 * we set FPEXC.EN to prevent traps to EL1, when setting the TFP bit.
>> -	 */
>> -	tbnz	x2, #HCR_RW_SHIFT, 99f // open code skip_32bit_state
>> -	mov	x3, #(1 << 30)
>> -	msr	fpexc32_el2, x3
>> -	isb
>> -99:
>>  	msr     hcr_el2, x2
>>  	mov	x2, #CPTR_EL2_TTA
>> +
>> +	ldrb    w3, [x0, #VCPU_VFP_DIRTY]
>> +	tbnz    w3, #0, 98f
>>  	orr     x2, x2, #CPTR_EL2_TFP
>> +98:
>>  	msr	cptr_el2, x2
>>  
>>  	mov	x2, #(1 << 15)	// Trap CP15 Cr=15
>> @@ -669,14 +666,12 @@ __restore_debug:
>>  	ret
>>  
>>  __save_fpsimd:
>> -	skip_fpsimd_state x3, 1f
>>  	save_fpsimd
>> -1:	ret
>> +	ret
>>  
>>  __restore_fpsimd:
>> -	skip_fpsimd_state x3, 1f
>>  	restore_fpsimd
>> -1:	ret
>> +	ret
>>  
>>  switch_to_guest_fpsimd:
>>  	push	x4, lr
>> @@ -688,6 +683,9 @@ switch_to_guest_fpsimd:
>>  
>>  	mrs	x0, tpidr_el2
>>  
>> +	mov	w2, #1
>> +	strb    w2, [x0, #VCPU_VFP_DIRTY]
>> +
>>  	ldr	x2, [x0, #VCPU_HOST_CONTEXT]
>>  	kern_hyp_va x2
>>  	bl __save_fpsimd
>> @@ -763,7 +761,6 @@ __kvm_vcpu_return:
>>  	add	x2, x0, #VCPU_CONTEXT
>>  
>>  	save_guest_regs
>> -	bl __save_fpsimd
>>  	bl __save_sysregs
>>  
>>  	skip_debug_state x3, 1f
>> @@ -784,7 +781,6 @@ __kvm_vcpu_return:
>>  	kern_hyp_va x2
>>  
>>  	bl __restore_sysregs
>> -	bl __restore_fpsimd
>>  	/* Clear FPSIMD and Trace trapping */
>>  	msr     cptr_el2, xzr
>>  
>> @@ -863,6 +859,33 @@ ENTRY(__kvm_flush_vm_context)
>>  	ret
>>  ENDPROC(__kvm_flush_vm_context)
>>  
>> +/**
>> + * void __kvm_enable_fpexc32(void) -
>> + *	We may be entering the guest and set CPTR_EL2.TFP to trap all floating
>> + *	point register accesses to EL2, however, the ARM manual clearly states
>> + *	that traps are only taken to EL2 if the operation would not otherwise
>> + *	trap to EL1.  Therefore, always make sure that for 32-bit guests,
>> + *	we set FPEXC.EN to prevent traps to EL1, when setting the TFP bit.
>> + */
>> +ENTRY(__kvm_enable_fpexc32)
>> +	mov	x3, #(1 << 30)
>> +	msr	fpexc32_el2, x3
>> +	ret
>> +ENDPROC(__kvm_enable_fpexc32)
>> +
>> +/**
>> + * void __kvm_save_fpexc32(void) -
>> + *	This function restores guest FPEXC to its vcpu context, we call this
>> + *	function from vcpu_put.
>> + */
>> +ENTRY(__kvm_save_fpexc32)
>> +	kern_hyp_va x0
>> +	add     x2, x0, #VCPU_CONTEXT
>> +	mrs     x1, fpexc32_el2
>> +	str     x1, [x2, #CPU_SYSREG_OFFSET(FPEXC32_EL2)]
>> +	ret
>> +ENDPROC(__kvm_save_fpexc32)
>> +
>>  __kvm_hyp_panic:
>>  	// Guess the context by looking at VTTBR:
>>  	// If zero, then we're already a host.
>>
> 
> Thanks,
> 
> 	M.
>