Re: [PATCH 5/6] KVM: s390: wire up seb feature

[Christian Borntraeger]

On 01/17/2018 12:33 PM, David Hildenbrand wrote:
> 
>>  #define ECB_GS		0x40
>>  #define ECB_TE		0x10
>>  #define ECB_SRSI	0x04
>> diff --git a/arch/s390/include/uapi/asm/kvm.h b/arch/s390/include/uapi/asm/kvm.h
>> index 38535a57..20b9e9f 100644
>> --- a/arch/s390/include/uapi/asm/kvm.h
>> +++ b/arch/s390/include/uapi/asm/kvm.h
>> @@ -224,6 +224,7 @@ struct kvm_guest_debug_arch {
>>  #define KVM_SYNC_RICCB  (1UL << 7)
>>  #define KVM_SYNC_FPRS   (1UL << 8)
>>  #define KVM_SYNC_GSCB   (1UL << 9)
>> +#define KVM_SYNC_SEBC   (1UL << 10)
>>  /* length and alignment of the sdnx as a power of two */
>>  #define SDNXC 8
>>  #define SDNXL (1UL << SDNXC)
>> @@ -247,7 +248,8 @@ struct kvm_sync_regs {
>>  	};
>>  	__u8  reserved[512];	/* for future vector expansion */
>>  	__u32 fpc;		/* valid on KVM_SYNC_VRS or KVM_SYNC_FPRS */
>> -	__u8 padding1[52];	/* riccb needs to be 64byte aligned */
>> +	__u8 sebc:1;		/* spec blocking */
> 
> do you want to define the unused bits as reserved? Nicer to read IMHO

I certainly want to have these bits for future use. So maybe a 
__u8 reserved : 7; 
after that makes a lot of sense. Also the sebc : 1; (spaces)

(FWIW, I will rename that to bpbc for other reasons).


> 
> (especially also using spaces "sebc : 1")
> 
>> +	__u8 padding1[51];	/* riccb needs to be 64byte aligned */
>>  	__u8 riccb[64];		/* runtime instrumentation controls block */
>>  	__u8 padding2[192];	/* sdnx needs to be 256byte aligned */
>>  	union {
>> diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
>> index 2c93cbb..0c18f73 100644
>> --- a/arch/s390/kvm/kvm-s390.c
>> +++ b/arch/s390/kvm/kvm-s390.c
>> @@ -421,6 +421,9 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
>>  	case KVM_CAP_S390_GS:
>>  		r = test_facility(133);
>>  		break;
>> +	case KVM_CAP_S390_SEB:
>> +		r = test_facility(82);
>> +		break;
>>  	default:
>>  		r = 0;
>>  	}
>> @@ -2198,6 +2201,8 @@ int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
>>  	kvm_s390_set_prefix(vcpu, 0);
>>  	if (test_kvm_facility(vcpu->kvm, 64))
>>  		vcpu->run->kvm_valid_regs |= KVM_SYNC_RICCB;
>> +	if (test_kvm_facility(vcpu->kvm, 82))
>> +		vcpu->run->kvm_valid_regs |= KVM_SYNC_SEBC;
>>  	if (test_kvm_facility(vcpu->kvm, 133))
>>  		vcpu->run->kvm_valid_regs |= KVM_SYNC_GSCB;
>>  	/* fprs can be synchronized via vrs, even if the guest has no vx. With
>> @@ -2339,6 +2344,7 @@ static void kvm_s390_vcpu_initial_reset(struct kvm_vcpu *vcpu)
>>  	current->thread.fpu.fpc = 0;
>>  	vcpu->arch.sie_block->gbea = 1;
>>  	vcpu->arch.sie_block->pp = 0;
>> +	vcpu->arch.sie_block->fpf &= ~FPF_SEBC;
>>  	vcpu->arch.pfault_token = KVM_S390_PFAULT_TOKEN_INVALID;
>>  	kvm_clear_async_pf_completion_queue(vcpu);
>>  	if (!kvm_s390_user_cpu_state_ctrl(vcpu->kvm))
>> @@ -3298,6 +3304,10 @@ static void sync_regs(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
>>  		vcpu->arch.sie_block->ecd |= ECD_HOSTREGMGMT;
>>  		vcpu->arch.gs_enabled = 1;
>>  	}
>> +	if (kvm_run->kvm_dirty_regs & KVM_SYNC_SEBC) {
> 
> We should test for test_facility(82). Otherwise user space can enable
> undefined bits in the SCB on machines with !facility 82.

Agreed, will fix.

> 
>> +		vcpu->arch.sie_block->fpf &= ~FPF_SEBC;
>> +		vcpu->arch.sie_block->fpf |= kvm_run->s.regs.sebc ? FPF_SEBC : 0;
>> +	}
>>  	save_access_regs(vcpu->arch.host_acrs);
>>  	restore_access_regs(vcpu->run->s.regs.acrs);
>>  	/* save host (userspace) fprs/vrs */
>> @@ -3344,6 +3354,7 @@ static void store_regs(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
>>  	kvm_run->s.regs.pft = vcpu->arch.pfault_token;
>>  	kvm_run->s.regs.pfs = vcpu->arch.pfault_select;
>>  	kvm_run->s.regs.pfc = vcpu->arch.pfault_compare;
>> +	kvm_run->s.regs.sebc = (vcpu->arch.sie_block->fpf & FPF_SEBC) == FPF_SEBC;
>>  	save_access_regs(vcpu->run->s.regs.acrs);
>>  	restore_access_regs(vcpu->arch.host_acrs);
>>  	/* Save guest register state */
>> diff --git a/arch/s390/kvm/vsie.c b/arch/s390/kvm/vsie.c
>> index 5d6ae03..10ea208 100644
>> --- a/arch/s390/kvm/vsie.c
>> +++ b/arch/s390/kvm/vsie.c
>> @@ -223,6 +223,10 @@ static void unshadow_scb(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
>>  	memcpy(scb_o->gcr, scb_s->gcr, 128);
>>  	scb_o->pp = scb_s->pp;
>>  
>> +	/* speculative blocking */
> 
> This field should only be written back with test_kvm_facility(vcpu->kvm, 82)

Agreed. 

> 
> (no public documentation, this looks like the SIE can modify this field?
> Triggered by which instruction?)

The instruction from patch 3.
> 
>> +	scb_o->fpf &= ~FPF_SEBC;
>> +	scb_o->fpf |= scb_s->fpf & FPF_SEBC;
>> +
>>  	/* interrupt intercept */
>>  	switch (scb_s->icptcode) {
>>  	case ICPT_PROGI:
>> @@ -265,6 +269,7 @@ static int shadow_scb(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
>>  	scb_s->ecb3 = 0;
>>  	scb_s->ecd = 0;
>>  	scb_s->fac = 0;
>> +	scb_s->fpf = 0;
>>  
>>  	rc = prepare_cpuflags(vcpu, vsie_page);
>>  	if (rc)
>> @@ -324,6 +329,9 @@ static int shadow_scb(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
>>  			prefix_unmapped(vsie_page);
>>  		scb_s->ecb |= scb_o->ecb & ECB_TE;
>>  	}
>> +	/* speculative blocking */
>> +	if (test_kvm_facility(vcpu->kvm, 82))
>> +		scb_s->fpf |= scb_o->fpf & FPF_SEBC;
>>  	/* SIMD */
>>  	if (test_kvm_facility(vcpu->kvm, 129)) {
>>  		scb_s->eca |= scb_o->eca & ECA_VX;
> 

Thanks for the quick review.