Skip to contentSkip to conversation

s390x: KVM: CPU Topology

PATCH Superseded Pierre Morel 4 participants 25 messages Aug 3, 2021 – Sep 9, 2021

Subsystems (3)

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

v3 cover s390x: KVM: CPU Topology

Hi all,

This new series add the implementation of interpretation for the PTF instruction.

The series is devided in three parts:
1- handling of the STSI instruction forwarding the CPU topology 2- implementation of the interpretation of the PTF instruction 3- use of the PTF interpretation to optimize topology change callback

1- STSI
To provide Topology information to the guest through the STSI instruction, we need to forward STSI with Function Code 15 to QEMU which will take care to provide the right information to the guest.

To let the guest use both the PTF instruction to check if a topology change occured and sthe STSI_15.x.x instruction we add a new KVM capability to enable the topology facility.

2- PTF
To implement PTF interpretation we make the MTCR pending when the last CPU backed by the vCPU changed from one socket to another.

The PTF instruction will report a topology change if there is any change with a previous STSI_15_2 SYSIB.
Changes inside a STSI_15_2 SYSIB occur if CPU bits are set or clear inside the CPU Topology List Entry CPU mask field, which happens with changes in CPU polarization, dedication, CPU types and adding or removing CPUs in a socket.

The reporting to the guest is done using the Multiprocessor Topology-Change-Report (MTCR) bit of the utility entry of the guest's SCA which will be cleared during the interpretation of PTF.

To check if the topology has been modified we use a new field of the arch vCPU to save the previous real CPU ID at the end of a schedule and verify on next schedule that the CPU used is in the same socket.

We deliberatly ignore:
- polarization: only horizontal polarization is currently used in linux.
- CPU Type: only IFL Type are supported in Linux
- Dedication: we consider that only a complete dedicated CPU stack can take benefit of the CPU Topology and let the admin take care of that.

Regards,
Pierre

Pierre Morel (3):
s390x: KVM: accept STSI for CPU topology information
s390x: KVM: Implementation of Multiprocessor Topology-Change-Report
s390x: optimization of the check for CPU topology change

 arch/s390/include/asm/kvm_host.h | 14 +++++++---
 arch/s390/kernel/topology.c      |  3 ++
 arch/s390/kvm/kvm-s390.c         | 48 +++++++++++++++++++++++++++++++-
 arch/s390/kvm/priv.c             |  7 ++++-
 arch/s390/kvm/vsie.c             |  3 ++
 include/uapi/linux/kvm.h         |  1 +
 6 files changed, 70 insertions(+), 6 deletions(-)
Signature · 25 lines
2.25.1

Changelog:

from v2 to v3

- use PTF interpretation
  (Christian)

- optimize arch_update_cpu_topology using PTF
  (Pierre)

from v1 to v2:

- Add a KVM capability to let QEMU know we support PTF and STSI 15
  (David)

- check KVM facility 11 before accepting STSI fc 15
  (David)

- handle all we can in userland
  (David)

- add tracing to STSI fc 15
  (Connie)
#1 lore raw

PATCH v3 1/3 s390x: KVM: accept STSI for CPU topology information

STSI(15.1.x) gives information on the CPU configuration topology. Let's accept the interception of STSI with the function code 15 and let the userland part of the hypervisor handle it when userland support the CPU Topology facility.

---
 arch/s390/kvm/priv.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
arch/s390/kvm/priv.c +6 -1
index 9928f785c677..8581b6881212 100644
@@ -856,7 +856,8 @@ static int handle_stsi(struct kvm_vcpu *vcpu)
 	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
 		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 
-	if (fc > 3) {
+	if ((fc > 3 && fc != 15) ||
+	    (fc == 15 && !test_kvm_facility(vcpu->kvm, 11))) {
 		kvm_s390_set_psw_cc(vcpu, 3);
 		return 0;
 	}
@@ -893,6 +894,10 @@ static int handle_stsi(struct kvm_vcpu *vcpu)
 			goto out_no_data;
 		handle_stsi_3_2_2(vcpu, (void *) mem);
 		break;
+	case 15:
+		trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
+		insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
+		return -EREMOTE;
 	}
 	if (kvm_s390_pv_cpu_is_protected(vcpu)) {
 		memcpy((void *)sida_origin(vcpu->arch.sie_block), (void *)mem,
Signature · 1 line
2.25.1
#2 lore raw

Re: [PATCH v3 1/3] s390x: KVM: accept STSI for CPU topology information

Quoted 26 lines · Pierre Morel

STSI(15.1.x) gives information on the CPU configuration topology. Let's accept the interception of STSI with the function code 15 and let the userland part of the hypervisor handle it when userland support the CPU Topology facility.

---
  arch/s390/kvm/priv.c | 7 ++++++-
  1 file changed, 6 insertions(+), 1 deletion(-)
arch/s390/kvm/priv.c +2 -1
index 9928f785c677..8581b6881212 100644
@@ -856,7 +856,8 @@ static int handle_stsi(struct kvm_vcpu *vcpu)
  	if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
  		return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
 
-	if (fc > 3) {
+	if ((fc > 3 && fc != 15) ||
+	    (fc == 15 && !test_kvm_facility(vcpu->kvm, 11))) {
  		kvm_s390_set_psw_cc(vcpu, 3);
  		return 0;
  	}

arch/s390/kvm/priv.c static int handle_stsi(struct kvm_vcpu *vcpu) · line 903

Quoted 10 lines
arch/s390/kvm/priv.c +4 -0
@@ -893,6 +894,10 @@ static int handle_stsi(struct kvm_vcpu *vcpu)
  			goto out_no_data;
  		handle_stsi_3_2_2(vcpu, (void *) mem);
  		break;
+	case 15:
+		trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
+		insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
+		return -EREMOTE;
  	}
  	if (kvm_s390_pv_cpu_is_protected(vcpu)) {
  		memcpy((void *)sida_origin(vcpu->arch.sie_block), (void *)mem,

Sorry, I'm a bit rusty on s390x kvm facility handling.

For test_kvm_facility() to succeed, the facility has to be in both:

a) fac_mask: actually available on the HW and supported by KVM (kvm_s390_fac_base via FACILITIES_KVM, kvm_s390_fac_ext via FACILITIES_KVM_CPUMODEL)

b) fac_list: enabled for a VM

AFAIU, facility 11 is neither in FACILITIES_KVM nor FACILITIES_KVM_CPUMODEL, and I remember it's a hypervisor-managed bit.

So unless we unlock facility 11 in FACILITIES_KVM_CPUMODEL, will test_kvm_facility(vcpu->kvm, 11) ever successfully trigger here?

I'm pretty sure I am messing something up :)

Signature · 3 lines
Thanks,

David / dhildenb
#3 lore raw

Re: [PATCH v3 1/3] s390x: KVM: accept STSI for CPU topology information

Quoted 44 lines · David Hildenbrand
Quoted 25 lines · Pierre Morel

STSI(15.1.x) gives information on the CPU configuration topology. Let's accept the interception of STSI with the function code 15 and let the userland part of the hypervisor handle it when userland support the CPU Topology facility.

---
  arch/s390/kvm/priv.c | 7 ++++++-

  1 file changed, 6 insertions(+), 1 deletion(-)
arch/s390/kvm/priv.c +2 -1
index 9928f785c677..8581b6881212 100644
@@ -856,7 +856,8 @@ static int handle_stsi(struct kvm_vcpu *vcpu)
       if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
           return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
-    if (fc > 3) {
+    if ((fc > 3 && fc != 15) ||
+        (fc == 15 && !test_kvm_facility(vcpu->kvm, 11))) {
           kvm_s390_set_psw_cc(vcpu, 3);
           return 0;
       }

arch/s390/kvm/priv.c static int handle_stsi(struct kvm_vcpu *vcpu) · line 903

Quoted 10 lines
arch/s390/kvm/priv.c +4 -0
@@ -893,6 +894,10 @@ static int handle_stsi(struct kvm_vcpu *vcpu)
               goto out_no_data;
           handle_stsi_3_2_2(vcpu, (void *) mem);
           break;
+    case 15:
+        trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
+        insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
+        return -EREMOTE;
       }
       if (kvm_s390_pv_cpu_is_protected(vcpu)) {
           memcpy((void *)sida_origin(vcpu->arch.sie_block), (void *)mem,

Sorry, I'm a bit rusty on s390x kvm facility handling.

For test_kvm_facility() to succeed, the facility has to be in both:

a) fac_mask: actually available on the HW and supported by KVM (kvm_s390_fac_base via FACILITIES_KVM, kvm_s390_fac_ext via FACILITIES_KVM_CPUMODEL)

Quoted 9 lines

b) fac_list: enabled for a VM

AFAIU, facility 11 is neither in FACILITIES_KVM nor FACILITIES_KVM_CPUMODEL, and I remember it's a hypervisor-managed bit.

So unless we unlock facility 11 in FACILITIES_KVM_CPUMODEL, will test_kvm_facility(vcpu->kvm, 11) ever successfully trigger here?

I'm pretty sure I am messing something up :)

I think it is the same remark that Christian did as wanted me to use the arch/s390/tools/gen_facilities.c to activate the facility.

The point is that CONFIGURATION_TOPOLOGY, STFL, 11, is already defined inside QEMU since full_GEN10_GA1, so the test_kvm_facility() will succeed with the next patch setting the facility 11 in the mask when getting the KVM_CAP_S390_CPU_TOPOLOGY from userland.

But if we activate it in KVM via any of the FACILITIES_KVM_xxx in the gen_facilities.c we will activate it for the guest what ever userland hypervizor we have, including old QEMU which will generate an exception.

In this circumstances we have the choice between:

- use FACILITY_KVM and handle everything in kernel
- use FACILITY_KVM and use an extra CAPABILITY to handle part in kernel to avoid guest crash and part in userland
- use only the extra CAPABILITY and handle almost everything in userland

I want to avoid kernel code when not really necessary so I eliminated the first option.

The last two are not very different but I found a better integration using the last one, allowing to use standard test_[kvm_]facility()

Thanks for reviewing.

Regards,
Pierre

Signature · 2 lines
Pierre Morel
IBM Lab Boeblingen
#4 lore raw

Re: [PATCH v3 1/3] s390x: KVM: accept STSI for CPU topology information

Quoted 61 lines · Pierre Morel
Quoted 48 lines · David Hildenbrand
Quoted 25 lines · Pierre Morel

STSI(15.1.x) gives information on the CPU configuration topology. Let's accept the interception of STSI with the function code 15 and let the userland part of the hypervisor handle it when userland support the CPU Topology facility.

---
   arch/s390/kvm/priv.c | 7 ++++++-
   1 file changed, 6 insertions(+), 1 deletion(-)
arch/s390/kvm/priv.c +2 -1
index 9928f785c677..8581b6881212 100644
@@ -856,7 +856,8 @@ static int handle_stsi(struct kvm_vcpu *vcpu)
       if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
           return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
-    if (fc > 3) {
+    if ((fc > 3 && fc != 15) ||
+        (fc == 15 && !test_kvm_facility(vcpu->kvm, 11))) {
           kvm_s390_set_psw_cc(vcpu, 3);
           return 0;
       }

arch/s390/kvm/priv.c static int handle_stsi(struct kvm_vcpu *vcpu) · line 903

Quoted 10 lines
arch/s390/kvm/priv.c +4 -0
@@ -893,6 +894,10 @@ static int handle_stsi(struct kvm_vcpu *vcpu)
               goto out_no_data;
           handle_stsi_3_2_2(vcpu, (void *) mem);
           break;
+    case 15:
+        trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
+        insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
+        return -EREMOTE;
       }
       if (kvm_s390_pv_cpu_is_protected(vcpu)) {
           memcpy((void *)sida_origin(vcpu->arch.sie_block), (void *)mem,

Sorry, I'm a bit rusty on s390x kvm facility handling.

For test_kvm_facility() to succeed, the facility has to be in both:

a) fac_mask: actually available on the HW and supported by KVM (kvm_s390_fac_base via FACILITIES_KVM, kvm_s390_fac_ext via FACILITIES_KVM_CPUMODEL)

b) fac_list: enabled for a VM

Quoted 9 lines

AFAIU, facility 11 is neither in FACILITIES_KVM nor FACILITIES_KVM_CPUMODEL, and I remember it's a hypervisor-managed bit.

So unless we unlock facility 11 in FACILITIES_KVM_CPUMODEL, will test_kvm_facility(vcpu->kvm, 11) ever successfully trigger here?

I'm pretty sure I am messing something up :)

Quoted 7 lines

I think it is the same remark that Christian did as wanted me to use the arch/s390/tools/gen_facilities.c to activate the facility.

The point is that CONFIGURATION_TOPOLOGY, STFL, 11, is already defined inside QEMU since full_GEN10_GA1, so the test_kvm_facility() will succeed with the next patch setting the facility 11 in the mask when getting the KVM_CAP_S390_CPU_TOPOLOGY from userland.

Ok, I see ...

QEMU knows the facility and as soon as we present it to QEMU, QEMU will want to automatically enable it in the "host" model.

However, we'd like QEMU to join in and handle some part of it.

So indeed, handling it like KVM_CAP_S390_VECTOR_REGISTERS or KVM_CAP_S390_RI looks like a reasonable approach.

Quoted 11 lines

But if we activate it in KVM via any of the FACILITIES_KVM_xxx in the gen_facilities.c we will activate it for the guest what ever userland hypervizor we have, including old QEMU which will generate an exception.

In this circumstances we have the choice between:

- use FACILITY_KVM and handle everything in kernel
- use FACILITY_KVM and use an extra CAPABILITY to handle part in kernel to avoid guest crash and part in userland

This sounds quite nice to me. Implement minimal kernel support and indicate the facility via stfl to user space.

In addition, add a new capability that intercepts to user space instead.

... but I can understand that it might not be worth it.

This patch as it stands doesn't make any sense on its own. Either document how it's supposed to work and why it is currently dead code, or simply squash into the next patch (preferred IMHO).

Signature · 3 lines
Thanks,

David / dhildenb
#5 lore raw

Re: [PATCH v3 1/3] s390x: KVM: accept STSI for CPU topology information

Quoted 90 lines · David Hildenbrand
Quoted 61 lines · Pierre Morel
Quoted 48 lines · David Hildenbrand
Quoted 25 lines · Pierre Morel

STSI(15.1.x) gives information on the CPU configuration topology. Let's accept the interception of STSI with the function code 15 and let the userland part of the hypervisor handle it when userland support the CPU Topology facility.

---
   arch/s390/kvm/priv.c | 7 ++++++-

   1 file changed, 6 insertions(+), 1 deletion(-)
arch/s390/kvm/priv.c +2 -1
index 9928f785c677..8581b6881212 100644
@@ -856,7 +856,8 @@ static int handle_stsi(struct kvm_vcpu *vcpu)
        if (vcpu->arch.sie_block->gpsw.mask & PSW_MASK_PSTATE)
            return kvm_s390_inject_program_int(vcpu, PGM_PRIVILEGED_OP);
-    if (fc > 3) {
+    if ((fc > 3 && fc != 15) ||
+        (fc == 15 && !test_kvm_facility(vcpu->kvm, 11))) {
            kvm_s390_set_psw_cc(vcpu, 3);
            return 0;
        }

arch/s390/kvm/priv.c static int handle_stsi(struct kvm_vcpu *vcpu) · line 903

Quoted 10 lines
arch/s390/kvm/priv.c +4 -0
@@ -893,6 +894,10 @@ static int handle_stsi(struct kvm_vcpu *vcpu)
                goto out_no_data;
            handle_stsi_3_2_2(vcpu, (void *) mem);
            break;
+    case 15:
+        trace_kvm_s390_handle_stsi(vcpu, fc, sel1, sel2, operand2);
+        insert_stsi_usr_data(vcpu, operand2, ar, fc, sel1, sel2);
+        return -EREMOTE;
        }
        if (kvm_s390_pv_cpu_is_protected(vcpu)) {
            memcpy((void *)sida_origin(vcpu->arch.sie_block), (void *)mem,

Sorry, I'm a bit rusty on s390x kvm facility handling.

For test_kvm_facility() to succeed, the facility has to be in both:

a) fac_mask: actually available on the HW and supported by KVM (kvm_s390_fac_base via FACILITIES_KVM, kvm_s390_fac_ext via FACILITIES_KVM_CPUMODEL)

b) fac_list: enabled for a VM

Quoted 9 lines

AFAIU, facility 11 is neither in FACILITIES_KVM nor FACILITIES_KVM_CPUMODEL, and I remember it's a hypervisor-managed bit.

So unless we unlock facility 11 in FACILITIES_KVM_CPUMODEL, will test_kvm_facility(vcpu->kvm, 11) ever successfully trigger here?

I'm pretty sure I am messing something up :)

Quoted 7 lines

I think it is the same remark that Christian did as wanted me to use the arch/s390/tools/gen_facilities.c to activate the facility.

The point is that CONFIGURATION_TOPOLOGY, STFL, 11, is already defined inside QEMU since full_GEN10_GA1, so the test_kvm_facility() will succeed with the next patch setting the facility 11 in the mask when getting the KVM_CAP_S390_CPU_TOPOLOGY from userland.

Ok, I see ...

QEMU knows the facility and as soon as we present it to QEMU, QEMU will want to automatically enable it in the "host" model.

However, we'd like QEMU to join in and handle some part of it.

So indeed, handling it like KVM_CAP_S390_VECTOR_REGISTERS or KVM_CAP_S390_RI looks like a reasonable approach.

Quoted 11 lines

But if we activate it in KVM via any of the FACILITIES_KVM_xxx in the gen_facilities.c we will activate it for the guest what ever userland hypervizor we have, including old QEMU which will generate an exception.

In this circumstances we have the choice between:

- use FACILITY_KVM and handle everything in kernel
- use FACILITY_KVM and use an extra CAPABILITY to handle part in kernel to avoid guest crash and part in userland

Quoted 6 lines

This sounds quite nice to me. Implement minimal kernel support and indicate the facility via stfl to user space.

In addition, add a new capability that intercepts to user space instead.

... but I can understand that it might not be worth it.

yes, since we need a CAPABILITY anyway I find it makes things more complicated.

Quoted 4 lines

This patch as it stands doesn't make any sense on its own. Either document how it's supposed to work and why it is currently dead code, or simply squash into the next patch (preferred IMHO).

Yes, you are right, I will squash it with the next patch.

Thanks,
Pierre

Signature · 2 lines
Pierre Morel
IBM Lab Boeblingen
#6 lore raw

PATCH v3 2/3 s390x: KVM: Implementation of Multiprocessor Topology-Change-Report

In reply to Pierre Morel

We let the userland hypervisor know if the machine support the CPU topology facility using a new KVM capability: KVM_CAP_S390_CPU_TOPOLOGY.

The PTF instruction will report a topology change if there is any change with a previous STSI_15_2 SYSIB.
Changes inside a STSI_15_2 SYSIB occur if CPU bits are set or clear inside the CPU Topology List Entry CPU mask field, which happens with changes in CPU polarization, dedication, CPU types and adding or removing CPUs in a socket.

The reporting to the guest is done using the Multiprocessor Topology-Change-Report (MTCR) bit of the utility entry of the guest's SCA which will be cleared during the interpretation of PTF.

To check if the topology has been modified we use a new field of the arch vCPU to save the previous real CPU ID at the end of a schedule and verify on next schedule that the CPU used is in the same socket.

We deliberatly ignore:
- polarization: only horizontal polarization is currently used in linux.
- CPU Type: only IFL Type are supported in Linux
- Dedication: we consider that only a complete dedicated CPU stack can take benefit of the CPU Topology.

---
 arch/s390/include/asm/kvm_host.h | 14 +++++++---
 arch/s390/kvm/kvm-s390.c         | 48 +++++++++++++++++++++++++++++++-
 arch/s390/kvm/vsie.c             |  3 ++
 include/uapi/linux/kvm.h         |  1 +
 4 files changed, 61 insertions(+), 5 deletions(-)
arch/s390/include/asm/kvm_host.h +10 -4
index 9b4473f76e56..b7effdc96a7a 100644
@@ -95,15 +95,19 @@ struct bsca_block {
 	union ipte_control ipte_control;
 	__u64	reserved[5];
 	__u64	mcn;
-	__u64	reserved2;
+#define ESCA_UTILITY_MTCR	0x8000
+	__u16	utility;
+	__u8	reserved2[6];
 	struct bsca_entry cpu[KVM_S390_BSCA_CPU_SLOTS];
 };
 
 struct esca_block {
 	union ipte_control ipte_control;
-	__u64   reserved1[7];
+	__u64   reserved1[6];
+	__u16	utility;
+	__u8	reserved2[6];
 	__u64   mcn[4];
-	__u64   reserved2[20];
+	__u64   reserved3[20];
 	struct esca_entry cpu[KVM_S390_ESCA_CPU_SLOTS];
 };
 
@@ -228,7 +232,7 @@ struct kvm_s390_sie_block {
 	__u8	icptcode;		/* 0x0050 */
 	__u8	icptstatus;		/* 0x0051 */
 	__u16	ihcpu;			/* 0x0052 */
-	__u8	reserved54;		/* 0x0054 */
+	__u8	mtcr;			/* 0x0054 */
 #define IICTL_CODE_NONE		 0x00
 #define IICTL_CODE_MCHK		 0x01
 #define IICTL_CODE_EXT		 0x02
@@ -246,6 +250,7 @@ struct kvm_s390_sie_block {
 #define ECB_TE		0x10
 #define ECB_SRSI	0x04
 #define ECB_HOSTPROTINT	0x02
+#define ECB_PTF		0x01
 	__u8	ecb;			/* 0x0061 */
 #define ECB2_CMMA	0x80
 #define ECB2_IEP	0x20
@@ -747,6 +752,7 @@ struct kvm_vcpu_arch {
 	bool skey_enabled;
 	struct kvm_s390_pv_vcpu pv;
 	union diag318_info diag318_info;
+	int prev_cpu;
 };
 
 struct kvm_vm_stat {
arch/s390/kvm/kvm-s390.c +47 -1

83 lines not on this page · raw message

arch/s390/kvm/vsie.c +3 -0
index 4002a24bc43a..50d67190bf65 100644
@@ -503,6 +503,9 @@ static int shadow_scb(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
 	/* Host-protection-interruption introduced with ESOP */
 	if (test_kvm_cpu_feat(vcpu->kvm, KVM_S390_VM_CPU_FEAT_ESOP))
 		scb_s->ecb |= scb_o->ecb & ECB_HOSTPROTINT;
+	/* CPU Topology */
+	if (test_kvm_facility(vcpu->kvm, 11))
+		scb_s->ecb |= scb_o->ecb & ECB_PTF;
 	/* transactional execution */
 	if (test_kvm_facility(vcpu->kvm, 73) && wants_tx) {
 		/* remap the prefix is tx is toggled on */
include/uapi/linux/kvm.h +1 -0
index d9e4aabcb31a..081ce0cd44b9 100644
@@ -1112,6 +1112,7 @@ struct kvm_ppc_resize_hpt {
 #define KVM_CAP_BINARY_STATS_FD 203
 #define KVM_CAP_EXIT_ON_EMULATION_FAILURE 204
 #define KVM_CAP_ARM_MTE 205
+#define KVM_CAP_S390_CPU_TOPOLOGY 206
 
 #ifdef KVM_CAP_IRQ_ROUTING
 
Signature · 1 line
2.25.1
#7 lore raw

Re: [PATCH v3 2/3] s390x: KVM: Implementation of Multiprocessor Topology-Change-Report

Quoted 164 lines · Pierre Morel

We let the userland hypervisor know if the machine support the CPU topology facility using a new KVM capability: KVM_CAP_S390_CPU_TOPOLOGY.

The PTF instruction will report a topology change if there is any change with a previous STSI_15_2 SYSIB.
Changes inside a STSI_15_2 SYSIB occur if CPU bits are set or clear inside the CPU Topology List Entry CPU mask field, which happens with changes in CPU polarization, dedication, CPU types and adding or removing CPUs in a socket.

The reporting to the guest is done using the Multiprocessor Topology-Change-Report (MTCR) bit of the utility entry of the guest's SCA which will be cleared during the interpretation of PTF.

To check if the topology has been modified we use a new field of the arch vCPU to save the previous real CPU ID at the end of a schedule and verify on next schedule that the CPU used is in the same socket.

We deliberatly ignore:
- polarization: only horizontal polarization is currently used in linux.
- CPU Type: only IFL Type are supported in Linux
- Dedication: we consider that only a complete dedicated CPU stack can take benefit of the CPU Topology.

---
  arch/s390/include/asm/kvm_host.h | 14 +++++++---
  arch/s390/kvm/kvm-s390.c         | 48 +++++++++++++++++++++++++++++++-
  arch/s390/kvm/vsie.c             |  3 ++
  include/uapi/linux/kvm.h         |  1 +
  4 files changed, 61 insertions(+), 5 deletions(-)
arch/s390/include/asm/kvm_host.h +10 -4
index 9b4473f76e56..b7effdc96a7a 100644
@@ -95,15 +95,19 @@ struct bsca_block {
  	union ipte_control ipte_control;
  	__u64	reserved[5];
  	__u64	mcn;
-	__u64	reserved2;
+#define ESCA_UTILITY_MTCR	0x8000
+	__u16	utility;
+	__u8	reserved2[6];
  	struct bsca_entry cpu[KVM_S390_BSCA_CPU_SLOTS];
  };
 
  struct esca_block {
  	union ipte_control ipte_control;
-	__u64   reserved1[7];
+	__u64   reserved1[6];
+	__u16	utility;
+	__u8	reserved2[6];
  	__u64   mcn[4];
-	__u64   reserved2[20];
+	__u64   reserved3[20];
  	struct esca_entry cpu[KVM_S390_ESCA_CPU_SLOTS];
  };
 
@@ -228,7 +232,7 @@ struct kvm_s390_sie_block {
  	__u8	icptcode;		/* 0x0050 */
  	__u8	icptstatus;		/* 0x0051 */
  	__u16	ihcpu;			/* 0x0052 */
-	__u8	reserved54;		/* 0x0054 */
+	__u8	mtcr;			/* 0x0054 */
  #define IICTL_CODE_NONE		 0x00
  #define IICTL_CODE_MCHK		 0x01
  #define IICTL_CODE_EXT		 0x02
@@ -246,6 +250,7 @@ struct kvm_s390_sie_block {
  #define ECB_TE		0x10
  #define ECB_SRSI	0x04
  #define ECB_HOSTPROTINT	0x02
+#define ECB_PTF		0x01
  	__u8	ecb;			/* 0x0061 */
  #define ECB2_CMMA	0x80
  #define ECB2_IEP	0x20
@@ -747,6 +752,7 @@ struct kvm_vcpu_arch {
  	bool skey_enabled;
  	struct kvm_s390_pv_vcpu pv;
  	union diag318_info diag318_info;
+	int prev_cpu;
  };
 
  struct kvm_vm_stat {
arch/s390/kvm/kvm-s390.c +42 -1
index b655a7d82bf0..ff6d8a2b511c 100644
@@ -568,6 +568,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
  	case KVM_CAP_S390_VCPU_RESETS:
  	case KVM_CAP_SET_GUEST_DEBUG:
  	case KVM_CAP_S390_DIAG318:
+	case KVM_CAP_S390_CPU_TOPOLOGY:
  		r = 1;
  		break;
  	case KVM_CAP_SET_GUEST_DEBUG2:
@@ -819,6 +820,23 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
  		icpt_operexc_on_all_vcpus(kvm);
  		r = 0;
  		break;
+	case KVM_CAP_S390_CPU_TOPOLOGY:
+		mutex_lock(&kvm->lock);
+		if (kvm->created_vcpus) {
+			r = -EBUSY;
+		} else {
+			set_kvm_facility(kvm->arch.model.fac_mask, 11);
+			set_kvm_facility(kvm->arch.model.fac_list, 11);
+			r = 0;
+		}
+		mutex_unlock(&kvm->lock);
+		VM_EVENT(kvm, 3, "ENABLE: CPU TOPOLOGY %s",
+			 r ? "(not available)" : "(success)");
+		break;
+
+		r = -EINVAL;
+		break;
+
  	default:
  		r = -EINVAL;
  		break;
@@ -3067,18 +3085,41 @@ __u64 kvm_s390_get_cpu_timer(struct kvm_vcpu *vcpu)
  	return value;
  }
 
-void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
+static void kvm_s390_set_mtcr(struct kvm_vcpu *vcpu)
  {
+	struct esca_block *esca = vcpu->kvm->arch.sca;
+
+	if (vcpu->arch.sie_block->ecb & ECB_PTF) {
+		ipte_lock(vcpu);
+		WRITE_ONCE(esca->utility, ESCA_UTILITY_MTCR);
+		ipte_unlock(vcpu);
+	}
+}
 
+void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
+{
  	gmap_enable(vcpu->arch.enabled_gmap);
  	kvm_s390_set_cpuflags(vcpu, CPUSTAT_RUNNING);
  	if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu))
  		__start_cpu_timer_accounting(vcpu);
  	vcpu->cpu = cpu;
+
+	/*
+	 * With PTF interpretation the guest will be aware of topology
+	 * change by the Multiprocessor Topology-Change-Report is pending.
+	 * Check for reasons to make the MTCR pending and make it pending.
+	 */
+	if ((vcpu->arch.sie_block->ecb & ECB_PTF) &&
+	    cpu != vcpu->arch.prev_cpu) {
+		if (cpu_topology[cpu].socket_id !=
+		    cpu_topology[vcpu->arch.prev_cpu].socket_id)
+			kvm_s390_set_mtcr(vcpu);
+	}
  }
 
  void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
  {
+	vcpu->arch.prev_cpu = vcpu->cpu;
  	vcpu->cpu = -1;
  	if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu))
  		__stop_cpu_timer_accounting(vcpu);

arch/s390/kvm/kvm-s390.c static int kvm_s390_vcpu_setup(struct kvm_vcpu *vcpu) · line 3245

Quoted 7 lines
arch/s390/kvm/kvm-s390.c +4 -0
@@ -3198,6 +3239,11 @@ static int kvm_s390_vcpu_setup(struct kvm_vcpu *vcpu)
  		vcpu->arch.sie_block->ecb |= ECB_HOSTPROTINT;
  	if (test_kvm_facility(vcpu->kvm, 9))
  		vcpu->arch.sie_block->ecb |= ECB_SRSI;
+
+	/* PTF needs both host and guest facilities to enable interpretation */
+	if (test_kvm_facility(vcpu->kvm, 11) && test_facility(11))
+		vcpu->arch.sie_block->ecb |= ECB_PTF;

Again, doesn't test_kvm_facility(vcpu->kvm, 11) imply that we have host support by checking fac_mask?

Signature · 3 lines
Thanks,

David / dhildenb
#8 lore raw

Re: [PATCH v3 2/3] s390x: KVM: Implementation of Multiprocessor Topology-Change-Report

Quoted 1 line · David Hildenbrand

On 03.08.21 10:26, Pierre Morel wrote:

...snip...

Quoted 17 lines
Quoted 17 lines
arch/s390/kvm/kvm-s390.c +12 -0
@@ -819,6 +820,23 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm, 
 struct kvm_enable_cap *cap)
           icpt_operexc_on_all_vcpus(kvm);
           r = 0;
           break;
+    case KVM_CAP_S390_CPU_TOPOLOGY:
+        mutex_lock(&kvm->lock);
+        if (kvm->created_vcpus) {
+            r = -EBUSY;
+        } else {
+            set_kvm_facility(kvm->arch.model.fac_mask, 11);
+            set_kvm_facility(kvm->arch.model.fac_list, 11);
+            r = 0;
+        }
+        mutex_unlock(&kvm->lock);
+        VM_EVENT(kvm, 3, "ENABLE: CPU TOPOLOGY %s",
+             r ? "(not available)" : "(success)");
Quoted 8 lines
Quoted 8 lines
arch/s390/kvm/kvm-s390.c +5 -0
+        break;
+
+        r = -EINVAL;
+        break;
+

      default:
          r = -EINVAL;
          break;

This above enables the facility 11.

...snip...

Quoted 14 lines
Quoted 10 lines
arch/s390/kvm/kvm-s390.c +2 -0
@@ -3198,6 +3239,11 @@ static int kvm_s390_vcpu_setup(struct kvm_vcpu 
 *vcpu)
           vcpu->arch.sie_block->ecb |= ECB_HOSTPROTINT;
       if (test_kvm_facility(vcpu->kvm, 9))
           vcpu->arch.sie_block->ecb |= ECB_SRSI;
+
+    /* PTF needs both host and guest facilities to enable 

interpretation */
+    if (test_kvm_facility(vcpu->kvm, 11) && test_facility(11)) +        vcpu->arch.sie_block->ecb |= ECB_PTF;

Again, doesn't test_kvm_facility(vcpu->kvm, 11) imply that we have host support by checking fac_mask?

Signature · 2 lines
Pierre Morel
IBM Lab Boeblingen
#9 lore raw

Re: [PATCH v3 2/3] s390x: KVM: Implementation of Multiprocessor Topology-Change-Report

In reply to Pierre Morel
Quoted 37 lines · Pierre Morel

We let the userland hypervisor know if the machine support the CPU topology facility using a new KVM capability: KVM_CAP_S390_CPU_TOPOLOGY.

The PTF instruction will report a topology change if there is any change with a previous STSI_15_2 SYSIB.
Changes inside a STSI_15_2 SYSIB occur if CPU bits are set or clear inside the CPU Topology List Entry CPU mask field, which happens with changes in CPU polarization, dedication, CPU types and adding or removing CPUs in a socket.

The reporting to the guest is done using the Multiprocessor Topology-Change-Report (MTCR) bit of the utility entry of the guest's SCA which will be cleared during the interpretation of PTF.

To check if the topology has been modified we use a new field of the arch vCPU to save the previous real CPU ID at the end of a schedule and verify on next schedule that the CPU used is in the same socket.

We deliberatly ignore:
- polarization: only horizontal polarization is currently used in linux.
- CPU Type: only IFL Type are supported in Linux
- Dedication: we consider that only a complete dedicated CPU stack can take benefit of the CPU Topology.

arch/s390/include/asm/kvm_host.h +1 -1
@@ -228,7 +232,7 @@ struct kvm_s390_sie_block {
  	__u8	icptcode;		/* 0x0050 */
  	__u8	icptstatus;		/* 0x0051 */
  	__u16	ihcpu;			/* 0x0052 */
-	__u8	reserved54;		/* 0x0054 */
+	__u8	mtcr;			/* 0x0054 */
  #define IICTL_CODE_NONE		 0x00
  #define IICTL_CODE_MCHK		 0x01
  #define IICTL_CODE_EXT		 0x02

arch/s390/include/asm/kvm_host.h struct kvm_s390_sie_block { · line 253

Quoted 4 lines
arch/s390/include/asm/kvm_host.h +1 -0
@@ -246,6 +250,7 @@ struct kvm_s390_sie_block {
  #define ECB_TE		0x10
  #define ECB_SRSI	0x04
  #define ECB_HOSTPROTINT	0x02
+#define ECB_PTF		0x01

From below I understand, that ECB_PTF can be used with stfl(11) in the hypervisor.

What is to happen if the hypervisor doesn't support stfl(11) and we consequently cannot use ECB_PTF? Will QEMU be able to emulate PTF fully?

Quoted 16 lines
arch/s390/include/asm/kvm_host.h +1 -0
  	__u8	ecb;			/* 0x0061 */
  #define ECB2_CMMA	0x80
  #define ECB2_IEP	0x20
@@ -747,6 +752,7 @@ struct kvm_vcpu_arch {
  	bool skey_enabled;
  	struct kvm_s390_pv_vcpu pv;
  	union diag318_info diag318_info;
+	int prev_cpu;
  };
 
  struct kvm_vm_stat {
arch/s390/kvm/kvm-s390.c +0 -0
index b655a7d82bf0..ff6d8a2b511c 100644

arch/s390/kvm/kvm-s390.c int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext) · line 571

Quoted 4 lines
arch/s390/kvm/kvm-s390.c +1 -0
@@ -568,6 +568,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
  	case KVM_CAP_S390_VCPU_RESETS:
  	case KVM_CAP_SET_GUEST_DEBUG:
  	case KVM_CAP_S390_DIAG318:
+	case KVM_CAP_S390_CPU_TOPOLOGY:

I would have expected instead

r = test_facility(11);
break

...

Quoted 12 lines
arch/s390/kvm/kvm-s390.c +5 -0
  		r = 1;
  		break;
  	case KVM_CAP_SET_GUEST_DEBUG2:
@@ -819,6 +820,23 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
  		icpt_operexc_on_all_vcpus(kvm);
  		r = 0;
  		break;
+	case KVM_CAP_S390_CPU_TOPOLOGY:
+		mutex_lock(&kvm->lock);
+		if (kvm->created_vcpus) {
+			r = -EBUSY;
+		} else {

...

} else if (test_facility(11)) {
	set_kvm_facility(kvm->arch.model.fac_mask, 11);
	set_kvm_facility(kvm->arch.model.fac_list, 11);
	r = 0;
} else {
	r = -EINVAL;
}

similar to how we handle KVM_CAP_S390_VECTOR_REGISTERS.

But I assume you want to be able to support hosts without ECB_PTF, correct?

arch/s390/kvm/kvm-s390.c int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap) · line 838

Quoted 11 lines
arch/s390/kvm/kvm-s390.c +11 -0
+			set_kvm_facility(kvm->arch.model.fac_mask, 11);
+			set_kvm_facility(kvm->arch.model.fac_list, 11);
+			r = 0;
+		}
+		mutex_unlock(&kvm->lock);
+		VM_EVENT(kvm, 3, "ENABLE: CPU TOPOLOGY %s",
+			 r ? "(not available)" : "(success)");
+		break;
+
+		r = -EINVAL;
+		break;

^ dead code

[...]

Quoted 9 lines
  }
  
  void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
  {
arch/s390/kvm/kvm-s390.c +1 -0
+	vcpu->arch.prev_cpu = vcpu->cpu;
  	vcpu->cpu = -1;
  	if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu))
  		__stop_cpu_timer_accounting(vcpu);

arch/s390/kvm/kvm-s390.c static int kvm_s390_vcpu_setup(struct kvm_vcpu *vcpu) · line 3245

Quoted 7 lines
arch/s390/kvm/kvm-s390.c +4 -0
@@ -3198,6 +3239,11 @@ static int kvm_s390_vcpu_setup(struct kvm_vcpu *vcpu)
  		vcpu->arch.sie_block->ecb |= ECB_HOSTPROTINT;
  	if (test_kvm_facility(vcpu->kvm, 9))
  		vcpu->arch.sie_block->ecb |= ECB_SRSI;
+
+	/* PTF needs both host and guest facilities to enable interpretation */
+	if (test_kvm_facility(vcpu->kvm, 11) && test_facility(11))
+		vcpu->arch.sie_block->ecb |= ECB_PTF;

Here you say we need both ...

Quoted 9 lines
arch/s390/kvm/kvm-s390.c +1 -0
+
  	if (test_kvm_facility(vcpu->kvm, 73))
  		vcpu->arch.sie_block->ecb |= ECB_TE;
 
arch/s390/kvm/vsie.c +0 -0
index 4002a24bc43a..50d67190bf65 100644

arch/s390/kvm/vsie.c static int shadow_scb(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page) · line 508

Quoted 6 lines
arch/s390/kvm/vsie.c +3 -0
@@ -503,6 +503,9 @@ static int shadow_scb(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
  	/* Host-protection-interruption introduced with ESOP */
  	if (test_kvm_cpu_feat(vcpu->kvm, KVM_S390_VM_CPU_FEAT_ESOP))
  		scb_s->ecb |= scb_o->ecb & ECB_HOSTPROTINT;
+	/* CPU Topology */
+	if (test_kvm_facility(vcpu->kvm, 11))
+		scb_s->ecb |= scb_o->ecb & ECB_PTF;

but here you don't check?

include/uapi/linux/kvm.h struct kvm_ppc_resize_hpt { · line 1116

Quoted 13 lines
arch/s390/kvm/vsie.c +0 -0
  	/* transactional execution */
  	if (test_kvm_facility(vcpu->kvm, 73) && wants_tx) {
  		/* remap the prefix is tx is toggled on */
include/uapi/linux/kvm.h +1 -0
index d9e4aabcb31a..081ce0cd44b9 100644
@@ -1112,6 +1112,7 @@ struct kvm_ppc_resize_hpt {
  #define KVM_CAP_BINARY_STATS_FD 203
  #define KVM_CAP_EXIT_ON_EMULATION_FAILURE 204
  #define KVM_CAP_ARM_MTE 205
+#define KVM_CAP_S390_CPU_TOPOLOGY 206
 

We'll need a Documentation/virt/kvm/api.rst description.

I'm not completely confident that the way we're handling the capability+facility is the right approach. It all feels a bit suboptimal.

Except stfl(74) -- STHYI --, we never enable a facility via set_kvm_facility() that's not available in the host. And STHYI is special such that it is never implemented in hardware.

I'll think about what might be cleaner once I get some more details about the interaction with stfl(11) in the hypervisor.

Signature · 3 lines
Thanks,

David / dhildenb
#10 lore raw

Re: [PATCH v3 2/3] s390x: KVM: Implementation of Multiprocessor Topology-Change-Report

Quoted 46 lines · David Hildenbrand
Quoted 37 lines · Pierre Morel

We let the userland hypervisor know if the machine support the CPU topology facility using a new KVM capability: KVM_CAP_S390_CPU_TOPOLOGY.

The PTF instruction will report a topology change if there is any change with a previous STSI_15_2 SYSIB.
Changes inside a STSI_15_2 SYSIB occur if CPU bits are set or clear inside the CPU Topology List Entry CPU mask field, which happens with changes in CPU polarization, dedication, CPU types and adding or removing CPUs in a socket.

The reporting to the guest is done using the Multiprocessor Topology-Change-Report (MTCR) bit of the utility entry of the guest's SCA which will be cleared during the interpretation of PTF.

To check if the topology has been modified we use a new field of the arch vCPU to save the previous real CPU ID at the end of a schedule and verify on next schedule that the CPU used is in the same socket.

We deliberatly ignore:
- polarization: only horizontal polarization is currently used in linux.
- CPU Type: only IFL Type are supported in Linux
- Dedication: we consider that only a complete dedicated CPU stack can take benefit of the CPU Topology.

(quoted diff) +0 -0
@@ -228,7 +232,7 @@ struct kvm_s390_sie_block {

      __u8    icptcode;        /* 0x0050 */
      __u8    icptstatus;        /* 0x0051 */
      __u16    ihcpu;            /* 0x0052 */
-    __u8    reserved54;        /* 0x0054 */
+    __u8    mtcr;            /* 0x0054 */
  #define IICTL_CODE_NONE         0x00
  #define IICTL_CODE_MCHK         0x01
  #define IICTL_CODE_EXT         0x02

arch/s390/include/asm/kvm_host.h struct kvm_s390_sie_block { · line 253

Quoted 4 lines
arch/s390/include/asm/kvm_host.h +1 -0
@@ -246,6 +250,7 @@ struct kvm_s390_sie_block {
   #define ECB_TE        0x10
   #define ECB_SRSI    0x04
   #define ECB_HOSTPROTINT    0x02
+#define ECB_PTF        0x01

From below I understand, that ECB_PTF can be used with stfl(11) in the
hypervisor.

Quoted 1 line

What is to happen if the hypervisor doesn't support stfl(11) and we consequently cannot use ECB_PTF? Will QEMU be able to emulate PTF fully?

Yes.

Quoted 25 lines
Quoted 18 lines

      __u8    ecb;            /* 0x0061 */
  #define ECB2_CMMA    0x80
  #define ECB2_IEP    0x20

arch/s390/include/asm/kvm_host.h +1 -0
@@ -747,6 +752,7 @@ struct kvm_vcpu_arch {
       bool skey_enabled;
       struct kvm_s390_pv_vcpu pv;
       union diag318_info diag318_info;
+    int prev_cpu;
   };
   struct kvm_vm_stat {
arch/s390/kvm/kvm-s390.c +0 -0
index b655a7d82bf0..ff6d8a2b511c 100644
@@ -568,6 +568,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, 
 long ext)
Quoted 4 lines

      case KVM_CAP_S390_VCPU_RESETS:
      case KVM_CAP_SET_GUEST_DEBUG:
      case KVM_CAP_S390_DIAG318:
+    case KVM_CAP_S390_CPU_TOPOLOGY:

I would have expected instead

Quoted 2 lines

r = test_facility(11);
break

The idea is that QEMU will emulate both PTF and SYSIB_15 in this case.

Quoted 26 lines

...

arch/s390/kvm/kvm-s390.c int kvm_vm_ioctl_enable_cap(struct kvm *kvm, · line 828

Quoted 13 lines

          r = 1;
          break;
      case KVM_CAP_SET_GUEST_DEBUG2:

arch/s390/kvm/kvm-s390.c +5 -0
@@ -819,6 +820,23 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm, 
 struct kvm_enable_cap *cap)
           icpt_operexc_on_all_vcpus(kvm);
           r = 0;
           break;
+    case KVM_CAP_S390_CPU_TOPOLOGY:
+        mutex_lock(&kvm->lock);
+        if (kvm->created_vcpus) {
+            r = -EBUSY;
+        } else {

...
} else if (test_facility(11)) {
    set_kvm_facility(kvm->arch.model.fac_mask, 11); set_kvm_facility(kvm->arch.model.fac_list, 11); r = 0;
} else {
    r = -EINVAL;
}

Quoted 3 lines

similar to how we handle KVM_CAP_S390_VECTOR_REGISTERS.

But I assume you want to be able to support hosts without ECB_PTF, correct?

yes, this was the idea.

Quoted 14 lines

arch/s390/kvm/kvm-s390.c int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap) · line 838

Quoted 13 lines
arch/s390/kvm/kvm-s390.c +11 -0
+            set_kvm_facility(kvm->arch.model.fac_mask, 11);
+            set_kvm_facility(kvm->arch.model.fac_list, 11);
+            r = 0;
+        }
+        mutex_unlock(&kvm->lock);
+        VM_EVENT(kvm, 3, "ENABLE: CPU TOPOLOGY %s",
+             r ? "(not available)" : "(success)");
+        break;
+
+        r = -EINVAL;
+        break;
Quoted 2 lines

^ dead code

:) indeed , sorry.

Quoted 13 lines

[...]

Quoted 11 lines

  }
  void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
  {
+    vcpu->arch.prev_cpu = vcpu->cpu;
      vcpu->cpu = -1;
      if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu)) __stop_cpu_timer_accounting(vcpu);

arch/s390/kvm/kvm-s390.c +0 -0
@@ -3198,6 +3239,11 @@ static int kvm_s390_vcpu_setup(struct kvm_vcpu 
 *vcpu)
           vcpu->arch.sie_block->ecb |= ECB_HOSTPROTINT;
       if (test_kvm_facility(vcpu->kvm, 9))
Quoted 8 lines
Quoted 6 lines
arch/s390/kvm/kvm-s390.c +2 -0
           vcpu->arch.sie_block->ecb |= ECB_SRSI;
+
+    /* PTF needs both host and guest facilities to enable 

interpretation */
+    if (test_kvm_facility(vcpu->kvm, 11) && test_facility(11)) +        vcpu->arch.sie_block->ecb |= ECB_PTF;

Here you say we need both ...

Yes because for interpretation we need both.
But if PTF is not interpreted we will emulate it in QEMU.

Quoted 10 lines
Quoted 10 lines

+
      if (test_kvm_facility(vcpu->kvm, 73))
          vcpu->arch.sie_block->ecb |= ECB_TE;

arch/s390/kvm/vsie.c +0 -0
index 4002a24bc43a..50d67190bf65 100644
@@ -503,6 +503,9 @@ static int shadow_scb(struct kvm_vcpu *vcpu, 
 struct vsie_page *vsie_page)
Quoted 8 lines

arch/s390/kvm/vsie.c static int shadow_scb(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page) · line 508

Quoted 6 lines
arch/s390/kvm/vsie.c +3 -0
       /* Host-protection-interruption introduced with ESOP */
       if (test_kvm_cpu_feat(vcpu->kvm, KVM_S390_VM_CPU_FEAT_ESOP))
           scb_s->ecb |= scb_o->ecb & ECB_HOSTPROTINT;
+    /* CPU Topology */
+    if (test_kvm_facility(vcpu->kvm, 11))
+        scb_s->ecb |= scb_o->ecb & ECB_PTF;

but here you don't check?

Arrrg, yes, this is false, we must check both here too.

Quoted 16 lines

include/uapi/linux/kvm.h struct kvm_ppc_resize_hpt { · line 1115

Quoted 13 lines
arch/s390/kvm/vsie.c +0 -0
       /* transactional execution */
       if (test_kvm_facility(vcpu->kvm, 73) && wants_tx) {

          /* remap the prefix is tx is toggled on */

include/uapi/linux/kvm.h +1 -0
index d9e4aabcb31a..081ce0cd44b9 100644
@@ -1112,6 +1112,7 @@ struct kvm_ppc_resize_hpt {
   #define KVM_CAP_BINARY_STATS_FD 203
   #define KVM_CAP_EXIT_ON_EMULATION_FAILURE 204
   #define KVM_CAP_ARM_MTE 205
+#define KVM_CAP_S390_CPU_TOPOLOGY 206

We'll need a Documentation/virt/kvm/api.rst description.

Quoted 3 lines

I'm not completely confident that the way we're handling the capability+facility is the right approach. It all feels a bit suboptimal.

Except stfl(74) -- STHYI --, we never enable a facility via set_kvm_facility() that's not available in the host. And STHYI is special such that it is never implemented in hardware.

Then we can fall back to KVM_facility + in kernel emulation but if for PTF it will be quite simple, for STSI_15 it will be much bigger.

Quoted 3 lines

I'll think about what might be cleaner once I get some more details about the interaction with stfl(11) in the hypervisor.

And I just saw I for an unknown reason forgot two patches in the QEMU series:

s390x: kvm: make topology change report pending
s390x: kvm: enable CPU Topology Function

So I will publish a new QEMU series this afternoon with the comments from Thomas.

thanks,
Pierre

Signature · 2 lines
Pierre Morel
IBM Lab Boeblingen
#11 lore raw

Re: [PATCH v3 2/3] s390x: KVM: Implementation of Multiprocessor Topology-Change-Report

Quoted 48 lines · Pierre Morel
Quoted 45 lines · David Hildenbrand
Quoted 37 lines · Pierre Morel

We let the userland hypervisor know if the machine support the CPU topology facility using a new KVM capability: KVM_CAP_S390_CPU_TOPOLOGY.

The PTF instruction will report a topology change if there is any change with a previous STSI_15_2 SYSIB.
Changes inside a STSI_15_2 SYSIB occur if CPU bits are set or clear inside the CPU Topology List Entry CPU mask field, which happens with changes in CPU polarization, dedication, CPU types and adding or removing CPUs in a socket.

The reporting to the guest is done using the Multiprocessor Topology-Change-Report (MTCR) bit of the utility entry of the guest's SCA which will be cleared during the interpretation of PTF.

To check if the topology has been modified we use a new field of the arch vCPU to save the previous real CPU ID at the end of a schedule and verify on next schedule that the CPU used is in the same socket.

We deliberatly ignore:
- polarization: only horizontal polarization is currently used in linux.
- CPU Type: only IFL Type are supported in Linux
- Dedication: we consider that only a complete dedicated CPU stack can take benefit of the CPU Topology.

(quoted diff) +0 -0
@@ -228,7 +232,7 @@ struct kvm_s390_sie_block {

      __u8    icptcode;        /* 0x0050 */
      __u8    icptstatus;        /* 0x0051 */
      __u16    ihcpu;            /* 0x0052 */
-    __u8    reserved54;        /* 0x0054 */
+    __u8    mtcr;            /* 0x0054 */
  #define IICTL_CODE_NONE         0x00
  #define IICTL_CODE_MCHK         0x01
  #define IICTL_CODE_EXT         0x02

arch/s390/include/asm/kvm_host.h struct kvm_s390_sie_block { · line 253

Quoted 4 lines
arch/s390/include/asm/kvm_host.h +1 -0
@@ -246,6 +250,7 @@ struct kvm_s390_sie_block {
   #define ECB_TE        0x10
   #define ECB_SRSI    0x04
   #define ECB_HOSTPROTINT    0x02
+#define ECB_PTF        0x01

 From below I understand, that ECB_PTF can be used with stfl(11) in the hypervisor.

Quoted 3 lines
Quoted 1 line

What is to happen if the hypervisor doesn't support stfl(11) and we consequently cannot use ECB_PTF? Will QEMU be able to emulate PTF fully?

Yes.

Do we want that? I do not think so. Other OSes (like zOS) do use PTF in there low level interrupt handler, so PTF must be really fast. I think I would prefer that in that case the guest will simply not see stfle(11).
So the user can still specify the topology but the guest will have no interface to query it.

#12 lore raw

Re: [PATCH v3 2/3] s390x: KVM: Implementation of Multiprocessor Topology-Change-Report

Quoted 52 lines · Christian Borntraeger
Quoted 49 lines · Pierre Morel
Quoted 46 lines · David Hildenbrand
Quoted 38 lines · Pierre Morel

We let the userland hypervisor know if the machine support the CPU topology facility using a new KVM capability: KVM_CAP_S390_CPU_TOPOLOGY.

The PTF instruction will report a topology change if there is any change with a previous STSI_15_2 SYSIB.
Changes inside a STSI_15_2 SYSIB occur if CPU bits are set or clear inside the CPU Topology List Entry CPU mask field, which happens with changes in CPU polarization, dedication, CPU types and adding or removing CPUs in a socket.

The reporting to the guest is done using the Multiprocessor Topology-Change-Report (MTCR) bit of the utility entry of the guest's SCA which will be cleared during the interpretation of PTF.

To check if the topology has been modified we use a new field of the arch vCPU to save the previous real CPU ID at the end of a schedule and verify on next schedule that the CPU used is in the same socket.

We deliberatly ignore:
- polarization: only horizontal polarization is currently used in linux.
- CPU Type: only IFL Type are supported in Linux
- Dedication: we consider that only a complete dedicated CPU stack can take benefit of the CPU Topology.

(quoted diff) +0 -0
@@ -228,7 +232,7 @@ struct kvm_s390_sie_block {

      __u8    icptcode;        /* 0x0050 */
      __u8    icptstatus;        /* 0x0051 */
      __u16    ihcpu;            /* 0x0052 */
-    __u8    reserved54;        /* 0x0054 */
+    __u8    mtcr;            /* 0x0054 */
  #define IICTL_CODE_NONE         0x00
  #define IICTL_CODE_MCHK         0x01
  #define IICTL_CODE_EXT         0x02

arch/s390/include/asm/kvm_host.h struct kvm_s390_sie_block { · line 253

Quoted 4 lines
arch/s390/include/asm/kvm_host.h +1 -0
@@ -246,6 +250,7 @@ struct kvm_s390_sie_block {
   #define ECB_TE        0x10
   #define ECB_SRSI    0x04
   #define ECB_HOSTPROTINT    0x02
+#define ECB_PTF        0x01

 From below I understand, that ECB_PTF can be used with stfl(11) in the hypervisor.

Quoted 7 lines
Quoted 3 lines
Quoted 1 line

What is to happen if the hypervisor doesn't support stfl(11) and we consequently cannot use ECB_PTF? Will QEMU be able to emulate PTF fully?

Yes.

Do we want that? I do not think so. Other OSes (like zOS) do use PTF in there low level interrupt handler, so PTF must be really fast. I think I would prefer that in that case the guest will simply not see stfle(11).
So the user can still specify the topology but the guest will have no interface to query it.

I do not understand.
If the host support stfle(11) we interpret PTF.

The proposition was to emulate only in the case it is not supported, what you propose is to not advertise stfl(11) if the host does not support it, and consequently to never emulate is it right?

In this case, as STSI_15 is linked to stfl(11) too, the guest will not be aware of the topology.

OK for me.

Signature · 2 lines
Pierre Morel
IBM Lab Boeblingen
#13 lore raw

Re: [PATCH v3 2/3] s390x: KVM: Implementation of Multiprocessor Topology-Change-Report

Quoted 58 lines · Pierre Morel
Quoted 55 lines · Christian Borntraeger
Quoted 51 lines · Pierre Morel
Quoted 45 lines · David Hildenbrand
Quoted 37 lines · Pierre Morel

We let the userland hypervisor know if the machine support the CPU topology facility using a new KVM capability: KVM_CAP_S390_CPU_TOPOLOGY.

The PTF instruction will report a topology change if there is any change with a previous STSI_15_2 SYSIB.
Changes inside a STSI_15_2 SYSIB occur if CPU bits are set or clear inside the CPU Topology List Entry CPU mask field, which happens with changes in CPU polarization, dedication, CPU types and adding or removing CPUs in a socket.

The reporting to the guest is done using the Multiprocessor Topology-Change-Report (MTCR) bit of the utility entry of the guest's SCA which will be cleared during the interpretation of PTF.

To check if the topology has been modified we use a new field of the arch vCPU to save the previous real CPU ID at the end of a schedule and verify on next schedule that the CPU used is in the same socket.

We deliberatly ignore:
- polarization: only horizontal polarization is currently used in linux.
- CPU Type: only IFL Type are supported in Linux
- Dedication: we consider that only a complete dedicated CPU stack can take benefit of the CPU Topology.

(quoted diff) +0 -0
@@ -228,7 +232,7 @@ struct kvm_s390_sie_block {

      __u8    icptcode;        /* 0x0050 */
      __u8    icptstatus;        /* 0x0051 */
      __u16    ihcpu;            /* 0x0052 */
-    __u8    reserved54;        /* 0x0054 */
+    __u8    mtcr;            /* 0x0054 */
  #define IICTL_CODE_NONE         0x00
  #define IICTL_CODE_MCHK         0x01
  #define IICTL_CODE_EXT         0x02

arch/s390/include/asm/kvm_host.h struct kvm_s390_sie_block { · line 253

Quoted 4 lines
arch/s390/include/asm/kvm_host.h +1 -0
@@ -246,6 +250,7 @@ struct kvm_s390_sie_block {
   #define ECB_TE        0x10
   #define ECB_SRSI    0x04
   #define ECB_HOSTPROTINT    0x02
+#define ECB_PTF        0x01

 From below I understand, that ECB_PTF can be used with stfl(11) in the hypervisor.

Quoted 1 line

What is to happen if the hypervisor doesn't support stfl(11) and we consequently cannot use ECB_PTF? Will QEMU be able to emulate PTF fully?

Yes.

Quoted 8 lines
Quoted 3 lines

Do we want that? I do not think so. Other OSes (like zOS) do use PTF in there low level interrupt handler, so PTF must be really fast. I think I would prefer that in that case the guest will simply not see stfle(11).
So the user can still specify the topology but the guest will have no interface to query it.

I do not understand.
If the host support stfle(11) we interpret PTF.

The proposition was to emulate only in the case it is not supported, what you propose is to not advertise stfl(11) if the host does not support it, and consequently to never emulate is it right?

Yes, exactly. My idea is to provide it to guests if we can do it fast, but do not provide it if it would add a performance issue.

Quoted 5 lines

In this case, as STSI_15 is linked to stfl(11) too, the guest will not be aware of the topology.

OK for me.

#14 lore raw

Re: [PATCH v3 2/3] s390x: KVM: Implementation of Multiprocessor Topology-Change-Report

Quoted 67 lines · Christian Borntraeger
Quoted 64 lines · Pierre Morel
Quoted 60 lines · Christian Borntraeger
Quoted 50 lines · Pierre Morel
Quoted 47 lines · David Hildenbrand
Quoted 39 lines · Pierre Morel

We let the userland hypervisor know if the machine support the CPU topology facility using a new KVM capability: KVM_CAP_S390_CPU_TOPOLOGY.

The PTF instruction will report a topology change if there is any change with a previous STSI_15_2 SYSIB.
Changes inside a STSI_15_2 SYSIB occur if CPU bits are set or clear inside the CPU Topology List Entry CPU mask field, which happens with changes in CPU polarization, dedication, CPU types and adding or removing CPUs in a socket.

The reporting to the guest is done using the Multiprocessor Topology-Change-Report (MTCR) bit of the utility entry of the guest's SCA which will be cleared during the interpretation of PTF.

To check if the topology has been modified we use a new field of the arch vCPU to save the previous real CPU ID at the end of a schedule and verify on next schedule that the CPU used is in the same socket.

We deliberatly ignore:
- polarization: only horizontal polarization is currently used in linux.
- CPU Type: only IFL Type are supported in Linux
- Dedication: we consider that only a complete dedicated CPU stack can
   take benefit of the CPU Topology.

(quoted diff) +0 -0
@@ -228,7 +232,7 @@ struct kvm_s390_sie_block {

      __u8    icptcode;        /* 0x0050 */
      __u8    icptstatus;        /* 0x0051 */
      __u16    ihcpu;            /* 0x0052 */
-    __u8    reserved54;        /* 0x0054 */
+    __u8    mtcr;            /* 0x0054 */
  #define IICTL_CODE_NONE         0x00
  #define IICTL_CODE_MCHK         0x01
  #define IICTL_CODE_EXT         0x02

arch/s390/include/asm/kvm_host.h struct kvm_s390_sie_block { · line 253

Quoted 4 lines
arch/s390/include/asm/kvm_host.h +1 -0
@@ -246,6 +250,7 @@ struct kvm_s390_sie_block {
   #define ECB_TE        0x10
   #define ECB_SRSI    0x04
   #define ECB_HOSTPROTINT    0x02
+#define ECB_PTF        0x01

 From below I understand, that ECB_PTF can be used with stfl(11) in the hypervisor.

Quoted 3 lines
Quoted 1 line

What is to happen if the hypervisor doesn't support stfl(11) and we consequently cannot use ECB_PTF? Will QEMU be able to emulate PTF fully?

Yes.

Do we want that? I do not think so. Other OSes (like zOS) do use PTF in there low level interrupt handler, so PTF must be really fast. I think I would prefer that in that case the guest will simply not see stfle(11).
So the user can still specify the topology but the guest will have no interface to query it.

Quoted 6 lines
Quoted 4 lines

I do not understand.
If the host support stfle(11) we interpret PTF.

The proposition was to emulate only in the case it is not supported, what you propose is to not advertise stfl(11) if the host does not support it, and consequently to never emulate is it right?

Yes, exactly. My idea is to provide it to guests if we can do it fast, but do not provide it if it would add a performance issue.

OK, understood, I will update this and the QEMU part too as we do not need emulation there anymore.

Thanks,
Pierre

Signature · 2 lines
Pierre Morel
IBM Lab Boeblingen
#15 lore raw

Re: [PATCH v3 2/3] s390x: KVM: Implementation of Multiprocessor Topology-Change-Report

In reply to David Hildenbrand
Quoted 174 lines · David Hildenbrand
Quoted 37 lines · Pierre Morel

We let the userland hypervisor know if the machine support the CPU topology facility using a new KVM capability: KVM_CAP_S390_CPU_TOPOLOGY.

The PTF instruction will report a topology change if there is any change with a previous STSI_15_2 SYSIB.
Changes inside a STSI_15_2 SYSIB occur if CPU bits are set or clear inside the CPU Topology List Entry CPU mask field, which happens with changes in CPU polarization, dedication, CPU types and adding or removing CPUs in a socket.

The reporting to the guest is done using the Multiprocessor Topology-Change-Report (MTCR) bit of the utility entry of the guest's SCA which will be cleared during the interpretation of PTF.

To check if the topology has been modified we use a new field of the arch vCPU to save the previous real CPU ID at the end of a schedule and verify on next schedule that the CPU used is in the same socket.

We deliberatly ignore:
- polarization: only horizontal polarization is currently used in linux.
- CPU Type: only IFL Type are supported in Linux
- Dedication: we consider that only a complete dedicated CPU stack can take benefit of the CPU Topology.

(quoted diff) +0 -0
@@ -228,7 +232,7 @@ struct kvm_s390_sie_block {

      __u8    icptcode;        /* 0x0050 */
      __u8    icptstatus;        /* 0x0051 */
      __u16    ihcpu;            /* 0x0052 */
-    __u8    reserved54;        /* 0x0054 */
+    __u8    mtcr;            /* 0x0054 */
  #define IICTL_CODE_NONE         0x00
  #define IICTL_CODE_MCHK         0x01
  #define IICTL_CODE_EXT         0x02

arch/s390/include/asm/kvm_host.h struct kvm_s390_sie_block { · line 253

Quoted 4 lines
arch/s390/include/asm/kvm_host.h +1 -0
@@ -246,6 +250,7 @@ struct kvm_s390_sie_block {
   #define ECB_TE        0x10
   #define ECB_SRSI    0x04
   #define ECB_HOSTPROTINT    0x02
+#define ECB_PTF        0x01

From below I understand, that ECB_PTF can be used with stfl(11) in the
hypervisor.

What is to happen if the hypervisor doesn't support stfl(11) and we consequently cannot use ECB_PTF? Will QEMU be able to emulate PTF fully?

Quoted 16 lines

      __u8    ecb;            /* 0x0061 */
  #define ECB2_CMMA    0x80
  #define ECB2_IEP    0x20

arch/s390/include/asm/kvm_host.h +1 -0
@@ -747,6 +752,7 @@ struct kvm_vcpu_arch {
       bool skey_enabled;
       struct kvm_s390_pv_vcpu pv;
       union diag318_info diag318_info;
+    int prev_cpu;
   };
   struct kvm_vm_stat {
arch/s390/kvm/kvm-s390.c +0 -0
index b655a7d82bf0..ff6d8a2b511c 100644
@@ -568,6 +568,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, 
 long ext)
Quoted 4 lines

      case KVM_CAP_S390_VCPU_RESETS:
      case KVM_CAP_SET_GUEST_DEBUG:
      case KVM_CAP_S390_DIAG318:
+    case KVM_CAP_S390_CPU_TOPOLOGY:

I would have expected instead

r = test_facility(11);
break

...

arch/s390/kvm/kvm-s390.c int kvm_vm_ioctl_enable_cap(struct kvm *kvm, · line 828

Quoted 13 lines

          r = 1;
          break;
      case KVM_CAP_SET_GUEST_DEBUG2:

arch/s390/kvm/kvm-s390.c +5 -0
@@ -819,6 +820,23 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm, 
 struct kvm_enable_cap *cap)
           icpt_operexc_on_all_vcpus(kvm);
           r = 0;
           break;
+    case KVM_CAP_S390_CPU_TOPOLOGY:
+        mutex_lock(&kvm->lock);
+        if (kvm->created_vcpus) {
+            r = -EBUSY;
+        } else {

...
} else if (test_facility(11)) {
    set_kvm_facility(kvm->arch.model.fac_mask, 11); set_kvm_facility(kvm->arch.model.fac_list, 11); r = 0;
} else {
    r = -EINVAL;
}

similar to how we handle KVM_CAP_S390_VECTOR_REGISTERS.

But I assume you want to be able to support hosts without ECB_PTF, correct?

arch/s390/kvm/kvm-s390.c int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap) · line 838

Quoted 11 lines
arch/s390/kvm/kvm-s390.c +11 -0
+            set_kvm_facility(kvm->arch.model.fac_mask, 11);
+            set_kvm_facility(kvm->arch.model.fac_list, 11);
+            r = 0;
+        }
+        mutex_unlock(&kvm->lock);
+        VM_EVENT(kvm, 3, "ENABLE: CPU TOPOLOGY %s",
+             r ? "(not available)" : "(success)");
+        break;
+
+        r = -EINVAL;
+        break;

^ dead code

[...]

Quoted 14 lines
arch/s390/kvm/kvm-s390.c +0 -0
   }
   void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)

  {
+    vcpu->arch.prev_cpu = vcpu->cpu;
      vcpu->cpu = -1;
      if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu)) __stop_cpu_timer_accounting(vcpu);

arch/s390/kvm/kvm-s390.c +2 -0
@@ -3198,6 +3239,11 @@ static int kvm_s390_vcpu_setup(struct kvm_vcpu 
 *vcpu)
           vcpu->arch.sie_block->ecb |= ECB_HOSTPROTINT;
       if (test_kvm_facility(vcpu->kvm, 9))
           vcpu->arch.sie_block->ecb |= ECB_SRSI;
+
+    /* PTF needs both host and guest facilities to enable 
Quoted 3 lines

interpretation */
+    if (test_kvm_facility(vcpu->kvm, 11) && test_facility(11)) +        vcpu->arch.sie_block->ecb |= ECB_PTF;

Here you say we need both ...

Quoted 8 lines

+
      if (test_kvm_facility(vcpu->kvm, 73))
          vcpu->arch.sie_block->ecb |= ECB_TE;

arch/s390/kvm/vsie.c +0 -0
index 4002a24bc43a..50d67190bf65 100644

arch/s390/kvm/vsie.c static int shadow_scb(struct kvm_vcpu *vcpu, · line 509

Quoted 7 lines
arch/s390/kvm/vsie.c +3 -0
@@ -503,6 +503,9 @@ static int shadow_scb(struct kvm_vcpu *vcpu, 
 struct vsie_page *vsie_page)
       /* Host-protection-interruption introduced with ESOP */
       if (test_kvm_cpu_feat(vcpu->kvm, KVM_S390_VM_CPU_FEAT_ESOP))
           scb_s->ecb |= scb_o->ecb & ECB_HOSTPROTINT;
+    /* CPU Topology */
+    if (test_kvm_facility(vcpu->kvm, 11))
+        scb_s->ecb |= scb_o->ecb & ECB_PTF;

but here you don't check?

include/uapi/linux/kvm.h struct kvm_ppc_resize_hpt { · line 1115

Quoted 12 lines
arch/s390/kvm/vsie.c +0 -0
       /* transactional execution */
       if (test_kvm_facility(vcpu->kvm, 73) && wants_tx) {

          /* remap the prefix is tx is toggled on */

include/uapi/linux/kvm.h +1 -0
index d9e4aabcb31a..081ce0cd44b9 100644
@@ -1112,6 +1112,7 @@ struct kvm_ppc_resize_hpt {
   #define KVM_CAP_BINARY_STATS_FD 203
   #define KVM_CAP_EXIT_ON_EMULATION_FAILURE 204
   #define KVM_CAP_ARM_MTE 205
+#define KVM_CAP_S390_CPU_TOPOLOGY 206

We'll need a Documentation/virt/kvm/api.rst description.

Quoted 6 lines

I'm not completely confident that the way we're handling the capability+facility is the right approach. It all feels a bit suboptimal.

Except stfl(74) -- STHYI --, we never enable a facility via set_kvm_facility() that's not available in the host. And STHYI is special such that it is never implemented in hardware.

I'll think about what might be cleaner once I get some more details about the interaction with stfl(11) in the hypervisor.

OK, may be we do not need to handle the case stfl(11) is not present in the host, these are pre GA10...

Signature · 2 lines
Pierre Morel
IBM Lab Boeblingen
#16 lore raw

Re: [PATCH v3 2/3] s390x: KVM: Implementation of Multiprocessor Topology-Change-Report

Quoted 171 lines · Pierre Morel
Quoted 168 lines · David Hildenbrand
Quoted 37 lines · Pierre Morel

We let the userland hypervisor know if the machine support the CPU topology facility using a new KVM capability: KVM_CAP_S390_CPU_TOPOLOGY.

The PTF instruction will report a topology change if there is any change with a previous STSI_15_2 SYSIB.
Changes inside a STSI_15_2 SYSIB occur if CPU bits are set or clear inside the CPU Topology List Entry CPU mask field, which happens with changes in CPU polarization, dedication, CPU types and adding or removing CPUs in a socket.

The reporting to the guest is done using the Multiprocessor Topology-Change-Report (MTCR) bit of the utility entry of the guest's SCA which will be cleared during the interpretation of PTF.

To check if the topology has been modified we use a new field of the arch vCPU to save the previous real CPU ID at the end of a schedule and verify on next schedule that the CPU used is in the same socket.

We deliberatly ignore:
- polarization: only horizontal polarization is currently used in linux.
- CPU Type: only IFL Type are supported in Linux
- Dedication: we consider that only a complete dedicated CPU stack can take benefit of the CPU Topology.

(quoted diff) +0 -0
@@ -228,7 +232,7 @@ struct kvm_s390_sie_block {

      __u8    icptcode;        /* 0x0050 */
      __u8    icptstatus;        /* 0x0051 */
      __u16    ihcpu;            /* 0x0052 */
-    __u8    reserved54;        /* 0x0054 */
+    __u8    mtcr;            /* 0x0054 */
  #define IICTL_CODE_NONE         0x00
  #define IICTL_CODE_MCHK         0x01
  #define IICTL_CODE_EXT         0x02

arch/s390/include/asm/kvm_host.h struct kvm_s390_sie_block { · line 253

Quoted 4 lines
arch/s390/include/asm/kvm_host.h +1 -0
@@ -246,6 +250,7 @@ struct kvm_s390_sie_block {
   #define ECB_TE        0x10
   #define ECB_SRSI    0x04
   #define ECB_HOSTPROTINT    0x02
+#define ECB_PTF        0x01

 From below I understand, that ECB_PTF can be used with stfl(11) in the hypervisor.

What is to happen if the hypervisor doesn't support stfl(11) and we consequently cannot use ECB_PTF? Will QEMU be able to emulate PTF fully?

Quoted 15 lines

      __u8    ecb;            /* 0x0061 */
  #define ECB2_CMMA    0x80
  #define ECB2_IEP    0x20

arch/s390/include/asm/kvm_host.h +1 -0
@@ -747,6 +752,7 @@ struct kvm_vcpu_arch {
       bool skey_enabled;
       struct kvm_s390_pv_vcpu pv;
       union diag318_info diag318_info;
+    int prev_cpu;
   };
   struct kvm_vm_stat {
arch/s390/kvm/kvm-s390.c +0 -0
index b655a7d82bf0..ff6d8a2b511c 100644
@@ -568,6 +568,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
Quoted 4 lines

      case KVM_CAP_S390_VCPU_RESETS:
      case KVM_CAP_SET_GUEST_DEBUG:
      case KVM_CAP_S390_DIAG318:
+    case KVM_CAP_S390_CPU_TOPOLOGY:

I would have expected instead

r = test_facility(11);
break

...

arch/s390/kvm/kvm-s390.c int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap) · line 827

Quoted 12 lines

          r = 1;
          break;
      case KVM_CAP_SET_GUEST_DEBUG2:

arch/s390/kvm/kvm-s390.c +5 -0
@@ -819,6 +820,23 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
           icpt_operexc_on_all_vcpus(kvm);
           r = 0;
           break;
+    case KVM_CAP_S390_CPU_TOPOLOGY:
+        mutex_lock(&kvm->lock);
+        if (kvm->created_vcpus) {
+            r = -EBUSY;
+        } else {

...
} else if (test_facility(11)) { set_kvm_facility(kvm->arch.model.fac_mask, 11); set_kvm_facility(kvm->arch.model.fac_list, 11); r = 0;
} else {
     r = -EINVAL;
}

similar to how we handle KVM_CAP_S390_VECTOR_REGISTERS.

But I assume you want to be able to support hosts without ECB_PTF, correct?

arch/s390/kvm/kvm-s390.c int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap) · line 838

Quoted 11 lines
arch/s390/kvm/kvm-s390.c +11 -0
+            set_kvm_facility(kvm->arch.model.fac_mask, 11);
+            set_kvm_facility(kvm->arch.model.fac_list, 11);
+            r = 0;
+        }
+        mutex_unlock(&kvm->lock);
+        VM_EVENT(kvm, 3, "ENABLE: CPU TOPOLOGY %s",
+             r ? "(not available)" : "(success)");
+        break;
+
+        r = -EINVAL;
+        break;

^ dead code

[...]

Quoted 8 lines
arch/s390/kvm/kvm-s390.c +1 -0
   }
   void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
   {
+    vcpu->arch.prev_cpu = vcpu->cpu;

      vcpu->cpu = -1;
      if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu)) __stop_cpu_timer_accounting(vcpu);

arch/s390/kvm/kvm-s390.c static int kvm_s390_vcpu_setup(struct kvm_vcpu *vcpu) · line 3245

Quoted 7 lines
arch/s390/kvm/kvm-s390.c +4 -0
@@ -3198,6 +3239,11 @@ static int kvm_s390_vcpu_setup(struct kvm_vcpu *vcpu)
           vcpu->arch.sie_block->ecb |= ECB_HOSTPROTINT;
       if (test_kvm_facility(vcpu->kvm, 9))
           vcpu->arch.sie_block->ecb |= ECB_SRSI;
+
+    /* PTF needs both host and guest facilities to enable interpretation */
+    if (test_kvm_facility(vcpu->kvm, 11) && test_facility(11))
+        vcpu->arch.sie_block->ecb |= ECB_PTF;

Here you say we need both ...

Quoted 8 lines
arch/s390/kvm/kvm-s390.c +1 -0
+
       if (test_kvm_facility(vcpu->kvm, 73))
           vcpu->arch.sie_block->ecb |= ECB_TE;
arch/s390/kvm/vsie.c +0 -0
index 4002a24bc43a..50d67190bf65 100644

arch/s390/kvm/vsie.c static int shadow_scb(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page) · line 508

Quoted 6 lines
arch/s390/kvm/vsie.c +3 -0
@@ -503,6 +503,9 @@ static int shadow_scb(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
       /* Host-protection-interruption introduced with ESOP */
       if (test_kvm_cpu_feat(vcpu->kvm, KVM_S390_VM_CPU_FEAT_ESOP))
           scb_s->ecb |= scb_o->ecb & ECB_HOSTPROTINT;
+    /* CPU Topology */
+    if (test_kvm_facility(vcpu->kvm, 11))
+        scb_s->ecb |= scb_o->ecb & ECB_PTF;

but here you don't check?

include/uapi/linux/kvm.h struct kvm_ppc_resize_hpt { · line 1115

Quoted 12 lines
arch/s390/kvm/vsie.c +0 -0
       /* transactional execution */
       if (test_kvm_facility(vcpu->kvm, 73) && wants_tx) {
           /* remap the prefix is tx is toggled on */
include/uapi/linux/kvm.h +1 -0
index d9e4aabcb31a..081ce0cd44b9 100644
@@ -1112,6 +1112,7 @@ struct kvm_ppc_resize_hpt {
   #define KVM_CAP_BINARY_STATS_FD 203
   #define KVM_CAP_EXIT_ON_EMULATION_FAILURE 204
   #define KVM_CAP_ARM_MTE 205
+#define KVM_CAP_S390_CPU_TOPOLOGY 206

We'll need a Documentation/virt/kvm/api.rst description.

Quoted 8 lines
Quoted 6 lines

I'm not completely confident that the way we're handling the capability+facility is the right approach. It all feels a bit suboptimal.

Except stfl(74) -- STHYI --, we never enable a facility via set_kvm_facility() that's not available in the host. And STHYI is special such that it is never implemented in hardware.

I'll think about what might be cleaner once I get some more details about the interaction with stfl(11) in the hypervisor.

OK, may be we do not need to handle the case stfl(11) is not present in the host, these are pre GA10...

What about VSIE? For all existing KVM guests, stfl11 is off.

#17 lore raw

Re: [PATCH v3 2/3] s390x: KVM: Implementation of Multiprocessor Topology-Change-Report

Quoted 182 lines · Christian Borntraeger
Quoted 179 lines · Pierre Morel
Quoted 176 lines · David Hildenbrand
Quoted 38 lines · Pierre Morel

We let the userland hypervisor know if the machine support the CPU topology facility using a new KVM capability: KVM_CAP_S390_CPU_TOPOLOGY.

The PTF instruction will report a topology change if there is any change with a previous STSI_15_2 SYSIB.
Changes inside a STSI_15_2 SYSIB occur if CPU bits are set or clear inside the CPU Topology List Entry CPU mask field, which happens with changes in CPU polarization, dedication, CPU types and adding or removing CPUs in a socket.

The reporting to the guest is done using the Multiprocessor Topology-Change-Report (MTCR) bit of the utility entry of the guest's SCA which will be cleared during the interpretation of PTF.

To check if the topology has been modified we use a new field of the arch vCPU to save the previous real CPU ID at the end of a schedule and verify on next schedule that the CPU used is in the same socket.

We deliberatly ignore:
- polarization: only horizontal polarization is currently used in linux.
- CPU Type: only IFL Type are supported in Linux
- Dedication: we consider that only a complete dedicated CPU stack can take benefit of the CPU Topology.

(quoted diff) +0 -0
@@ -228,7 +232,7 @@ struct kvm_s390_sie_block {

      __u8    icptcode;        /* 0x0050 */
      __u8    icptstatus;        /* 0x0051 */
      __u16    ihcpu;            /* 0x0052 */
-    __u8    reserved54;        /* 0x0054 */
+    __u8    mtcr;            /* 0x0054 */
  #define IICTL_CODE_NONE         0x00
  #define IICTL_CODE_MCHK         0x01
  #define IICTL_CODE_EXT         0x02

arch/s390/include/asm/kvm_host.h struct kvm_s390_sie_block { · line 253

Quoted 4 lines
arch/s390/include/asm/kvm_host.h +1 -0
@@ -246,6 +250,7 @@ struct kvm_s390_sie_block {
   #define ECB_TE        0x10
   #define ECB_SRSI    0x04
   #define ECB_HOSTPROTINT    0x02
+#define ECB_PTF        0x01

 From below I understand, that ECB_PTF can be used with stfl(11) in the hypervisor.

What is to happen if the hypervisor doesn't support stfl(11) and we consequently cannot use ECB_PTF? Will QEMU be able to emulate PTF fully?

Quoted 16 lines

      __u8    ecb;            /* 0x0061 */
  #define ECB2_CMMA    0x80
  #define ECB2_IEP    0x20

arch/s390/include/asm/kvm_host.h +1 -0
@@ -747,6 +752,7 @@ struct kvm_vcpu_arch {
       bool skey_enabled;
       struct kvm_s390_pv_vcpu pv;
       union diag318_info diag318_info;
+    int prev_cpu;
   };
   struct kvm_vm_stat {
arch/s390/kvm/kvm-s390.c +0 -0
index b655a7d82bf0..ff6d8a2b511c 100644
@@ -568,6 +568,7 @@ int kvm_vm_ioctl_check_extension(struct kvm 
 *kvm, long ext)
Quoted 4 lines

      case KVM_CAP_S390_VCPU_RESETS:
      case KVM_CAP_SET_GUEST_DEBUG:
      case KVM_CAP_S390_DIAG318:
+    case KVM_CAP_S390_CPU_TOPOLOGY:

I would have expected instead

r = test_facility(11);
break

...

arch/s390/kvm/kvm-s390.c int kvm_vm_ioctl_enable_cap(struct kvm *kvm, · line 828

Quoted 13 lines

          r = 1;
          break;
      case KVM_CAP_SET_GUEST_DEBUG2:

arch/s390/kvm/kvm-s390.c +5 -0
@@ -819,6 +820,23 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm, 
 struct kvm_enable_cap *cap)
           icpt_operexc_on_all_vcpus(kvm);
           r = 0;
           break;
+    case KVM_CAP_S390_CPU_TOPOLOGY:
+        mutex_lock(&kvm->lock);
+        if (kvm->created_vcpus) {
+            r = -EBUSY;
+        } else {

...
} else if (test_facility(11)) { set_kvm_facility(kvm->arch.model.fac_mask, 11); set_kvm_facility(kvm->arch.model.fac_list, 11); r = 0;
} else {
     r = -EINVAL;
}

similar to how we handle KVM_CAP_S390_VECTOR_REGISTERS.

But I assume you want to be able to support hosts without ECB_PTF, correct?

arch/s390/kvm/kvm-s390.c int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap) · line 838

Quoted 11 lines
arch/s390/kvm/kvm-s390.c +11 -0
+            set_kvm_facility(kvm->arch.model.fac_mask, 11);
+            set_kvm_facility(kvm->arch.model.fac_list, 11);
+            r = 0;
+        }
+        mutex_unlock(&kvm->lock);
+        VM_EVENT(kvm, 3, "ENABLE: CPU TOPOLOGY %s",
+             r ? "(not available)" : "(success)");
+        break;
+
+        r = -EINVAL;
+        break;

^ dead code

[...]

Quoted 14 lines
arch/s390/kvm/kvm-s390.c +0 -0
   }
   void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)

  {
+    vcpu->arch.prev_cpu = vcpu->cpu;
      vcpu->cpu = -1;
      if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu)) __stop_cpu_timer_accounting(vcpu);

arch/s390/kvm/kvm-s390.c +2 -0
@@ -3198,6 +3239,11 @@ static int kvm_s390_vcpu_setup(struct 
 kvm_vcpu *vcpu)
           vcpu->arch.sie_block->ecb |= ECB_HOSTPROTINT;
       if (test_kvm_facility(vcpu->kvm, 9))
           vcpu->arch.sie_block->ecb |= ECB_SRSI;
+
+    /* PTF needs both host and guest facilities to enable 
Quoted 3 lines

interpretation */
+    if (test_kvm_facility(vcpu->kvm, 11) && test_facility(11)) +        vcpu->arch.sie_block->ecb |= ECB_PTF;

Here you say we need both ...

Quoted 8 lines

+
      if (test_kvm_facility(vcpu->kvm, 73))
          vcpu->arch.sie_block->ecb |= ECB_TE;

arch/s390/kvm/vsie.c +0 -0
index 4002a24bc43a..50d67190bf65 100644

arch/s390/kvm/vsie.c static int shadow_scb(struct kvm_vcpu *vcpu, · line 509

Quoted 7 lines
arch/s390/kvm/vsie.c +3 -0
@@ -503,6 +503,9 @@ static int shadow_scb(struct kvm_vcpu *vcpu, 
 struct vsie_page *vsie_page)
       /* Host-protection-interruption introduced with ESOP */
       if (test_kvm_cpu_feat(vcpu->kvm, KVM_S390_VM_CPU_FEAT_ESOP))
           scb_s->ecb |= scb_o->ecb & ECB_HOSTPROTINT;
+    /* CPU Topology */
+    if (test_kvm_facility(vcpu->kvm, 11))
+        scb_s->ecb |= scb_o->ecb & ECB_PTF;

but here you don't check?

include/uapi/linux/kvm.h struct kvm_ppc_resize_hpt { · line 1115

Quoted 12 lines
arch/s390/kvm/vsie.c +0 -0
       /* transactional execution */
       if (test_kvm_facility(vcpu->kvm, 73) && wants_tx) {

          /* remap the prefix is tx is toggled on */

include/uapi/linux/kvm.h +1 -0
index d9e4aabcb31a..081ce0cd44b9 100644
@@ -1112,6 +1112,7 @@ struct kvm_ppc_resize_hpt {
   #define KVM_CAP_BINARY_STATS_FD 203
   #define KVM_CAP_EXIT_ON_EMULATION_FAILURE 204
   #define KVM_CAP_ARM_MTE 205
+#define KVM_CAP_S390_CPU_TOPOLOGY 206

We'll need a Documentation/virt/kvm/api.rst description.

I'm not completely confident that the way we're handling the capability+facility is the right approach. It all feels a bit suboptimal.

Quoted 8 lines
Quoted 6 lines
Quoted 4 lines

Except stfl(74) -- STHYI --, we never enable a facility via set_kvm_facility() that's not available in the host. And STHYI is special such that it is never implemented in hardware.

I'll think about what might be cleaner once I get some more details about the interaction with stfl(11) in the hypervisor.

OK, may be we do not need to handle the case stfl(11) is not present in the host, these are pre GA10...

What about VSIE? For all existing KVM guests, stfl11 is off.

In VSIE the patch activates stfl(11) only if the host has stfl(11).

I do not see any problem to activate the interpretation in VSIE with ECB_PTF (ECB.7) when the host has stfl(11) and QEMU asks to enable it for the guest using the CAPABILITY as it is done in this patch.

if any intermediary hypervizor decide to not advertize stfl(11) for the guest like an old QEMU not having the CAPABILITY, or a QEMU with ctop=off, KVM will not set ECB_PTF and the PTF instruction will trigger a program check as before.

Is it OK or did I missed something?

Signature · 2 lines
Pierre Morel
IBM Lab Boeblingen
#18 lore raw

Re: [PATCH v3 2/3] s390x: KVM: Implementation of Multiprocessor Topology-Change-Report

Quoted 188 lines · Pierre Morel
Quoted 176 lines · Christian Borntraeger
Quoted 173 lines · Pierre Morel
Quoted 170 lines · David Hildenbrand
Quoted 37 lines · Pierre Morel

We let the userland hypervisor know if the machine support the CPU topology facility using a new KVM capability: KVM_CAP_S390_CPU_TOPOLOGY.

The PTF instruction will report a topology change if there is any change with a previous STSI_15_2 SYSIB.
Changes inside a STSI_15_2 SYSIB occur if CPU bits are set or clear inside the CPU Topology List Entry CPU mask field, which happens with changes in CPU polarization, dedication, CPU types and adding or removing CPUs in a socket.

The reporting to the guest is done using the Multiprocessor Topology-Change-Report (MTCR) bit of the utility entry of the guest's SCA which will be cleared during the interpretation of PTF.

To check if the topology has been modified we use a new field of the arch vCPU to save the previous real CPU ID at the end of a schedule and verify on next schedule that the CPU used is in the same socket.

We deliberatly ignore:
- polarization: only horizontal polarization is currently used in linux.
- CPU Type: only IFL Type are supported in Linux
- Dedication: we consider that only a complete dedicated CPU stack can take benefit of the CPU Topology.

(quoted diff) +0 -0
@@ -228,7 +232,7 @@ struct kvm_s390_sie_block {

      __u8    icptcode;        /* 0x0050 */
      __u8    icptstatus;        /* 0x0051 */
      __u16    ihcpu;            /* 0x0052 */
-    __u8    reserved54;        /* 0x0054 */
+    __u8    mtcr;            /* 0x0054 */
  #define IICTL_CODE_NONE         0x00
  #define IICTL_CODE_MCHK         0x01
  #define IICTL_CODE_EXT         0x02

arch/s390/include/asm/kvm_host.h struct kvm_s390_sie_block { · line 253

Quoted 4 lines
arch/s390/include/asm/kvm_host.h +1 -0
@@ -246,6 +250,7 @@ struct kvm_s390_sie_block {
   #define ECB_TE        0x10
   #define ECB_SRSI    0x04
   #define ECB_HOSTPROTINT    0x02
+#define ECB_PTF        0x01

 From below I understand, that ECB_PTF can be used with stfl(11) in the hypervisor.

What is to happen if the hypervisor doesn't support stfl(11) and we consequently cannot use ECB_PTF? Will QEMU be able to emulate PTF fully?

Quoted 15 lines

      __u8    ecb;            /* 0x0061 */
  #define ECB2_CMMA    0x80
  #define ECB2_IEP    0x20

arch/s390/include/asm/kvm_host.h +1 -0
@@ -747,6 +752,7 @@ struct kvm_vcpu_arch {
       bool skey_enabled;
       struct kvm_s390_pv_vcpu pv;
       union diag318_info diag318_info;
+    int prev_cpu;
   };
   struct kvm_vm_stat {
arch/s390/kvm/kvm-s390.c +0 -0
index b655a7d82bf0..ff6d8a2b511c 100644
@@ -568,6 +568,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
Quoted 4 lines

      case KVM_CAP_S390_VCPU_RESETS:
      case KVM_CAP_SET_GUEST_DEBUG:
      case KVM_CAP_S390_DIAG318:
+    case KVM_CAP_S390_CPU_TOPOLOGY:

I would have expected instead

r = test_facility(11);
break

...

arch/s390/kvm/kvm-s390.c int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap) · line 827

Quoted 12 lines

          r = 1;
          break;
      case KVM_CAP_SET_GUEST_DEBUG2:

arch/s390/kvm/kvm-s390.c +5 -0
@@ -819,6 +820,23 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap)
           icpt_operexc_on_all_vcpus(kvm);
           r = 0;
           break;
+    case KVM_CAP_S390_CPU_TOPOLOGY:
+        mutex_lock(&kvm->lock);
+        if (kvm->created_vcpus) {
+            r = -EBUSY;
+        } else {

...
} else if (test_facility(11)) { set_kvm_facility(kvm->arch.model.fac_mask, 11); set_kvm_facility(kvm->arch.model.fac_list, 11); r = 0;
} else {
     r = -EINVAL;
}

similar to how we handle KVM_CAP_S390_VECTOR_REGISTERS.

But I assume you want to be able to support hosts without ECB_PTF, correct?

arch/s390/kvm/kvm-s390.c int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap) · line 838

Quoted 11 lines
arch/s390/kvm/kvm-s390.c +11 -0
+            set_kvm_facility(kvm->arch.model.fac_mask, 11);
+            set_kvm_facility(kvm->arch.model.fac_list, 11);
+            r = 0;
+        }
+        mutex_unlock(&kvm->lock);
+        VM_EVENT(kvm, 3, "ENABLE: CPU TOPOLOGY %s",
+             r ? "(not available)" : "(success)");
+        break;
+
+        r = -EINVAL;
+        break;

^ dead code

[...]

Quoted 8 lines
arch/s390/kvm/kvm-s390.c +1 -0
   }
   void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
   {
+    vcpu->arch.prev_cpu = vcpu->cpu;

      vcpu->cpu = -1;
      if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu)) __stop_cpu_timer_accounting(vcpu);

arch/s390/kvm/kvm-s390.c static int kvm_s390_vcpu_setup(struct kvm_vcpu *vcpu) · line 3245

Quoted 7 lines
arch/s390/kvm/kvm-s390.c +4 -0
@@ -3198,6 +3239,11 @@ static int kvm_s390_vcpu_setup(struct kvm_vcpu *vcpu)
           vcpu->arch.sie_block->ecb |= ECB_HOSTPROTINT;
       if (test_kvm_facility(vcpu->kvm, 9))
           vcpu->arch.sie_block->ecb |= ECB_SRSI;
+
+    /* PTF needs both host and guest facilities to enable interpretation */
+    if (test_kvm_facility(vcpu->kvm, 11) && test_facility(11))
+        vcpu->arch.sie_block->ecb |= ECB_PTF;

Here you say we need both ...

Quoted 8 lines
arch/s390/kvm/kvm-s390.c +1 -0
+
       if (test_kvm_facility(vcpu->kvm, 73))
           vcpu->arch.sie_block->ecb |= ECB_TE;
arch/s390/kvm/vsie.c +0 -0
index 4002a24bc43a..50d67190bf65 100644

arch/s390/kvm/vsie.c static int shadow_scb(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page) · line 508

Quoted 6 lines
arch/s390/kvm/vsie.c +3 -0
@@ -503,6 +503,9 @@ static int shadow_scb(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page)
       /* Host-protection-interruption introduced with ESOP */
       if (test_kvm_cpu_feat(vcpu->kvm, KVM_S390_VM_CPU_FEAT_ESOP))
           scb_s->ecb |= scb_o->ecb & ECB_HOSTPROTINT;
+    /* CPU Topology */
+    if (test_kvm_facility(vcpu->kvm, 11))
+        scb_s->ecb |= scb_o->ecb & ECB_PTF;

but here you don't check?

include/uapi/linux/kvm.h struct kvm_ppc_resize_hpt { · line 1115

Quoted 12 lines
arch/s390/kvm/vsie.c +0 -0
       /* transactional execution */
       if (test_kvm_facility(vcpu->kvm, 73) && wants_tx) {
           /* remap the prefix is tx is toggled on */
include/uapi/linux/kvm.h +1 -0
index d9e4aabcb31a..081ce0cd44b9 100644
@@ -1112,6 +1112,7 @@ struct kvm_ppc_resize_hpt {
   #define KVM_CAP_BINARY_STATS_FD 203
   #define KVM_CAP_EXIT_ON_EMULATION_FAILURE 204
   #define KVM_CAP_ARM_MTE 205
+#define KVM_CAP_S390_CPU_TOPOLOGY 206

We'll need a Documentation/virt/kvm/api.rst description.

I'm not completely confident that the way we're handling the capability+facility is the right approach. It all feels a bit suboptimal.

Quoted 8 lines
Quoted 6 lines
Quoted 4 lines

Except stfl(74) -- STHYI --, we never enable a facility via set_kvm_facility() that's not available in the host. And STHYI is special such that it is never implemented in hardware.

I'll think about what might be cleaner once I get some more details about the interaction with stfl(11) in the hypervisor.

OK, may be we do not need to handle the case stfl(11) is not present in the host, these are pre GA10...

What about VSIE? For all existing KVM guests, stfl11 is off.

Quoted 7 lines

In VSIE the patch activates stfl(11) only if the host has stfl(11).

I do not see any problem to activate the interpretation in VSIE with ECB_PTF (ECB.7) when the host has stfl(11) and QEMU asks to enable it for the guest using the CAPABILITY as it is done in this patch.

if any intermediary hypervizor decide to not advertize stfl(11) for the guest like an old QEMU not having the CAPABILITY, or a QEMU with ctop=off, KVM will not set ECB_PTF and the PTF instruction will trigger a program check as before.

Is it OK or did I missed something?

Yes, sure.
My point was regarding the pre z10 statement. We will see hosts without stfl(e)11 when running nested on z14, z15 and co.

#19 lore raw

Re: [PATCH v3 2/3] s390x: KVM: Implementation of Multiprocessor Topology-Change-Report

Quoted 200 lines · Christian Borntraeger
Quoted 197 lines · Pierre Morel
Quoted 183 lines · Christian Borntraeger
Quoted 180 lines · Pierre Morel
Quoted 177 lines · David Hildenbrand
Quoted 39 lines · Pierre Morel

We let the userland hypervisor know if the machine support the CPU topology facility using a new KVM capability: KVM_CAP_S390_CPU_TOPOLOGY.

The PTF instruction will report a topology change if there is any change with a previous STSI_15_2 SYSIB.
Changes inside a STSI_15_2 SYSIB occur if CPU bits are set or clear inside the CPU Topology List Entry CPU mask field, which happens with changes in CPU polarization, dedication, CPU types and adding or removing CPUs in a socket.

The reporting to the guest is done using the Multiprocessor Topology-Change-Report (MTCR) bit of the utility entry of the guest's SCA which will be cleared during the interpretation of PTF.

To check if the topology has been modified we use a new field of the arch vCPU to save the previous real CPU ID at the end of a schedule and verify on next schedule that the CPU used is in the same socket.

We deliberatly ignore:
- polarization: only horizontal polarization is currently used in linux.
- CPU Type: only IFL Type are supported in Linux
- Dedication: we consider that only a complete dedicated CPU stack can
   take benefit of the CPU Topology.

(quoted diff) +0 -0
@@ -228,7 +232,7 @@ struct kvm_s390_sie_block {

      __u8    icptcode;        /* 0x0050 */
      __u8    icptstatus;        /* 0x0051 */
      __u16    ihcpu;            /* 0x0052 */
-    __u8    reserved54;        /* 0x0054 */
+    __u8    mtcr;            /* 0x0054 */
  #define IICTL_CODE_NONE         0x00
  #define IICTL_CODE_MCHK         0x01
  #define IICTL_CODE_EXT         0x02

arch/s390/include/asm/kvm_host.h struct kvm_s390_sie_block { · line 253

Quoted 4 lines
arch/s390/include/asm/kvm_host.h +1 -0
@@ -246,6 +250,7 @@ struct kvm_s390_sie_block {
   #define ECB_TE        0x10
   #define ECB_SRSI    0x04
   #define ECB_HOSTPROTINT    0x02
+#define ECB_PTF        0x01

 From below I understand, that ECB_PTF can be used with stfl(11) in the hypervisor.

What is to happen if the hypervisor doesn't support stfl(11) and we consequently cannot use ECB_PTF? Will QEMU be able to emulate PTF fully?

Quoted 16 lines

      __u8    ecb;            /* 0x0061 */
  #define ECB2_CMMA    0x80
  #define ECB2_IEP    0x20

arch/s390/include/asm/kvm_host.h +1 -0
@@ -747,6 +752,7 @@ struct kvm_vcpu_arch {
       bool skey_enabled;
       struct kvm_s390_pv_vcpu pv;
       union diag318_info diag318_info;
+    int prev_cpu;
   };
   struct kvm_vm_stat {
arch/s390/kvm/kvm-s390.c +0 -0
index b655a7d82bf0..ff6d8a2b511c 100644
@@ -568,6 +568,7 @@ int kvm_vm_ioctl_check_extension(struct kvm 
 *kvm, long ext)
Quoted 4 lines

      case KVM_CAP_S390_VCPU_RESETS:
      case KVM_CAP_SET_GUEST_DEBUG:
      case KVM_CAP_S390_DIAG318:
+    case KVM_CAP_S390_CPU_TOPOLOGY:

I would have expected instead

r = test_facility(11);
break

...

arch/s390/kvm/kvm-s390.c int kvm_vm_ioctl_enable_cap(struct kvm *kvm, · line 828

Quoted 13 lines

          r = 1;
          break;
      case KVM_CAP_SET_GUEST_DEBUG2:

arch/s390/kvm/kvm-s390.c +5 -0
@@ -819,6 +820,23 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm, 
 struct kvm_enable_cap *cap)
           icpt_operexc_on_all_vcpus(kvm);
           r = 0;
           break;
+    case KVM_CAP_S390_CPU_TOPOLOGY:
+        mutex_lock(&kvm->lock);
+        if (kvm->created_vcpus) {
+            r = -EBUSY;
+        } else {

...
} else if (test_facility(11)) { set_kvm_facility(kvm->arch.model.fac_mask, 11); set_kvm_facility(kvm->arch.model.fac_list, 11); r = 0;
} else {
     r = -EINVAL;
}

similar to how we handle KVM_CAP_S390_VECTOR_REGISTERS.

But I assume you want to be able to support hosts without ECB_PTF, correct?

arch/s390/kvm/kvm-s390.c int kvm_vm_ioctl_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap) · line 838

Quoted 11 lines
arch/s390/kvm/kvm-s390.c +11 -0
+            set_kvm_facility(kvm->arch.model.fac_mask, 11);
+            set_kvm_facility(kvm->arch.model.fac_list, 11);
+            r = 0;
+        }
+        mutex_unlock(&kvm->lock);
+        VM_EVENT(kvm, 3, "ENABLE: CPU TOPOLOGY %s",
+             r ? "(not available)" : "(success)");
+        break;
+
+        r = -EINVAL;
+        break;

^ dead code

[...]

Quoted 14 lines
arch/s390/kvm/kvm-s390.c +0 -0
   }
   void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)

  {
+    vcpu->arch.prev_cpu = vcpu->cpu;
      vcpu->cpu = -1;
      if (vcpu->arch.cputm_enabled && !is_vcpu_idle(vcpu)) __stop_cpu_timer_accounting(vcpu);

arch/s390/kvm/kvm-s390.c +2 -0
@@ -3198,6 +3239,11 @@ static int kvm_s390_vcpu_setup(struct 
 kvm_vcpu *vcpu)
           vcpu->arch.sie_block->ecb |= ECB_HOSTPROTINT;
       if (test_kvm_facility(vcpu->kvm, 9))
           vcpu->arch.sie_block->ecb |= ECB_SRSI;
+
+    /* PTF needs both host and guest facilities to enable 
Quoted 3 lines

interpretation */
+    if (test_kvm_facility(vcpu->kvm, 11) && test_facility(11)) +        vcpu->arch.sie_block->ecb |= ECB_PTF;

Here you say we need both ...

Quoted 8 lines

+
      if (test_kvm_facility(vcpu->kvm, 73))
          vcpu->arch.sie_block->ecb |= ECB_TE;

arch/s390/kvm/vsie.c +0 -0
index 4002a24bc43a..50d67190bf65 100644

arch/s390/kvm/vsie.c static int shadow_scb(struct kvm_vcpu *vcpu, · line 509

Quoted 7 lines
arch/s390/kvm/vsie.c +3 -0
@@ -503,6 +503,9 @@ static int shadow_scb(struct kvm_vcpu *vcpu, 
 struct vsie_page *vsie_page)
       /* Host-protection-interruption introduced with ESOP */
       if (test_kvm_cpu_feat(vcpu->kvm, KVM_S390_VM_CPU_FEAT_ESOP))
           scb_s->ecb |= scb_o->ecb & ECB_HOSTPROTINT;
+    /* CPU Topology */
+    if (test_kvm_facility(vcpu->kvm, 11))
+        scb_s->ecb |= scb_o->ecb & ECB_PTF;

but here you don't check?

include/uapi/linux/kvm.h struct kvm_ppc_resize_hpt { · line 1115

Quoted 12 lines
arch/s390/kvm/vsie.c +0 -0
       /* transactional execution */
       if (test_kvm_facility(vcpu->kvm, 73) && wants_tx) {

          /* remap the prefix is tx is toggled on */

include/uapi/linux/kvm.h +1 -0
index d9e4aabcb31a..081ce0cd44b9 100644
@@ -1112,6 +1112,7 @@ struct kvm_ppc_resize_hpt {
   #define KVM_CAP_BINARY_STATS_FD 203
   #define KVM_CAP_EXIT_ON_EMULATION_FAILURE 204
   #define KVM_CAP_ARM_MTE 205
+#define KVM_CAP_S390_CPU_TOPOLOGY 206

We'll need a Documentation/virt/kvm/api.rst description.

I'm not completely confident that the way we're handling the capability+facility is the right approach. It all feels a bit suboptimal.

Quoted 8 lines
Quoted 6 lines
Quoted 4 lines

Except stfl(74) -- STHYI --, we never enable a facility via set_kvm_facility() that's not available in the host. And STHYI is special such that it is never implemented in hardware.

I'll think about what might be cleaner once I get some more details about the interaction with stfl(11) in the hypervisor.

OK, may be we do not need to handle the case stfl(11) is not present in the host, these are pre GA10...

What about VSIE? For all existing KVM guests, stfl11 is off.

In VSIE the patch activates stfl(11) only if the host has stfl(11).

Quoted 8 lines
Quoted 5 lines

I do not see any problem to activate the interpretation in VSIE with ECB_PTF (ECB.7) when the host has stfl(11) and QEMU asks to enable it for the guest using the CAPABILITY as it is done in this patch.

if any intermediary hypervizor decide to not advertize stfl(11) for the guest like an old QEMU not having the CAPABILITY, or a QEMU with ctop=off, KVM will not set ECB_PTF and the PTF instruction will trigger a program check as before.

Is it OK or did I missed something?

Yes, sure.
My point was regarding the pre z10 statement.  We will see hosts without stfl(e)11 when running nested on z14, z15 and co.

Ah OK, yes.
understood.

Thanks,
Pierre

Signature · 2 lines
Pierre Morel
IBM Lab Boeblingen
#20 lore raw

Re: [PATCH v3 2/3] s390x: KVM: Implementation of Multiprocessor Topology-Change-Report

In reply to David Hildenbrand
Quoted 72 lines · David Hildenbrand
Quoted 37 lines · Pierre Morel

We let the userland hypervisor know if the machine support the CPU topology facility using a new KVM capability: KVM_CAP_S390_CPU_TOPOLOGY.

The PTF instruction will report a topology change if there is any change with a previous STSI_15_2 SYSIB.
Changes inside a STSI_15_2 SYSIB occur if CPU bits are set or clear inside the CPU Topology List Entry CPU mask field, which happens with changes in CPU polarization, dedication, CPU types and adding or removing CPUs in a socket.

The reporting to the guest is done using the Multiprocessor Topology-Change-Report (MTCR) bit of the utility entry of the guest's SCA which will be cleared during the interpretation of PTF.

To check if the topology has been modified we use a new field of the arch vCPU to save the previous real CPU ID at the end of a schedule and verify on next schedule that the CPU used is in the same socket.

We deliberatly ignore:
- polarization: only horizontal polarization is currently used in linux.
- CPU Type: only IFL Type are supported in Linux
- Dedication: we consider that only a complete dedicated CPU stack can take benefit of the CPU Topology.

(quoted diff) +0 -0
@@ -228,7 +232,7 @@ struct kvm_s390_sie_block {

      __u8    icptcode;        /* 0x0050 */
      __u8    icptstatus;        /* 0x0051 */
      __u16    ihcpu;            /* 0x0052 */
-    __u8    reserved54;        /* 0x0054 */
+    __u8    mtcr;            /* 0x0054 */
  #define IICTL_CODE_NONE         0x00
  #define IICTL_CODE_MCHK         0x01
  #define IICTL_CODE_EXT         0x02

arch/s390/include/asm/kvm_host.h struct kvm_s390_sie_block { · line 253

Quoted 4 lines
arch/s390/include/asm/kvm_host.h +1 -0
@@ -246,6 +250,7 @@ struct kvm_s390_sie_block {
   #define ECB_TE        0x10
   #define ECB_SRSI    0x04
   #define ECB_HOSTPROTINT    0x02
+#define ECB_PTF        0x01

From below I understand, that ECB_PTF can be used with stfl(11) in the
hypervisor.

What is to happen if the hypervisor doesn't support stfl(11) and we consequently cannot use ECB_PTF? Will QEMU be able to emulate PTF fully?

Quoted 16 lines

      __u8    ecb;            /* 0x0061 */
  #define ECB2_CMMA    0x80
  #define ECB2_IEP    0x20

arch/s390/include/asm/kvm_host.h +1 -0
@@ -747,6 +752,7 @@ struct kvm_vcpu_arch {
       bool skey_enabled;
       struct kvm_s390_pv_vcpu pv;
       union diag318_info diag318_info;
+    int prev_cpu;
   };
   struct kvm_vm_stat {
arch/s390/kvm/kvm-s390.c +0 -0
index b655a7d82bf0..ff6d8a2b511c 100644
@@ -568,6 +568,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, 
 long ext)
Quoted 4 lines

      case KVM_CAP_S390_VCPU_RESETS:
      case KVM_CAP_SET_GUEST_DEBUG:
      case KVM_CAP_S390_DIAG318:
+    case KVM_CAP_S390_CPU_TOPOLOGY:

I would have expected instead

Quoted 2 lines

r = test_facility(11);
break

I will change to this as we decided not to support emulation if the hist does not support facility 11.

Quoted 26 lines

...

arch/s390/kvm/kvm-s390.c int kvm_vm_ioctl_enable_cap(struct kvm *kvm, · line 828

Quoted 13 lines

          r = 1;
          break;
      case KVM_CAP_SET_GUEST_DEBUG2:

arch/s390/kvm/kvm-s390.c +5 -0
@@ -819,6 +820,23 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm, 
 struct kvm_enable_cap *cap)
           icpt_operexc_on_all_vcpus(kvm);
           r = 0;
           break;
+    case KVM_CAP_S390_CPU_TOPOLOGY:
+        mutex_lock(&kvm->lock);
+        if (kvm->created_vcpus) {
+            r = -EBUSY;
+        } else {

...
} else if (test_facility(11)) {
    set_kvm_facility(kvm->arch.model.fac_mask, 11); set_kvm_facility(kvm->arch.model.fac_list, 11); r = 0;
} else {
    r = -EINVAL;
}

Quoted 3 lines

similar to how we handle KVM_CAP_S390_VECTOR_REGISTERS.

But I assume you want to be able to support hosts without ECB_PTF, correct?

No more, after Christian comments we do not want to support emulation at all.

Quoted 2 lines

...snip...

Quoted 17 lines
Quoted 5 lines

+
+    /* PTF needs both host and guest facilities to enable interpretation */
+    if (test_kvm_facility(vcpu->kvm, 11) && test_facility(11)) +        vcpu->arch.sie_block->ecb |= ECB_PTF;

Here you say we need both ...

Quoted 9 lines

+
      if (test_kvm_facility(vcpu->kvm, 73))
          vcpu->arch.sie_block->ecb |= ECB_TE;

arch/s390/kvm/vsie.c +0 -0
index 4002a24bc43a..50d67190bf65 100644
@@ -503,6 +503,9 @@ static int shadow_scb(struct kvm_vcpu *vcpu, 
 struct vsie_page *vsie_page)
Quoted 8 lines

arch/s390/kvm/vsie.c static int shadow_scb(struct kvm_vcpu *vcpu, struct vsie_page *vsie_page) · line 508

Quoted 6 lines
arch/s390/kvm/vsie.c +3 -0
       /* Host-protection-interruption introduced with ESOP */
       if (test_kvm_cpu_feat(vcpu->kvm, KVM_S390_VM_CPU_FEAT_ESOP))
           scb_s->ecb |= scb_o->ecb & ECB_HOSTPROTINT;
+    /* CPU Topology */
+    if (test_kvm_facility(vcpu->kvm, 11))
+        scb_s->ecb |= scb_o->ecb & ECB_PTF;

but here you don't check?

do we really need to check at all, even for test_kvm_facility() ?
as facilities do not change during a guest session and we checked for setting it at first time.

Regards,
Pierre

Signature · 2 lines
Pierre Morel
IBM Lab Boeblingen
#21 lore raw

PATCH v3 3/3 s390x: optimization of the check for CPU topology change

In reply to Pierre Morel

Now that the PTF instruction is interpreted by the SIE we can optimize the arch_update_cpu_topology callback to check if there is a real need to update the topology by using the PTF instruction.

---
 arch/s390/kernel/topology.c | 3 +++
 1 file changed, 3 insertions(+)
arch/s390/kernel/topology.c +3 -0
index 26aa2614ee35..741cb447e78e 100644
@@ -322,6 +322,9 @@ int arch_update_cpu_topology(void)
 	struct device *dev;
 	int cpu, rc;
 
+	if (!ptf(PTF_CHECK))
+		return 0;
+
 	rc = __arch_update_cpu_topology();
 	on_each_cpu(__arch_update_dedicated_flag, NULL, 0);
 	for_each_online_cpu(cpu) {
Signature · 1 line
2.25.1
maintainer
#22 lore raw

Re: [PATCH v3 3/3] s390x: optimization of the check for CPU topology change

Quoted 14 lines · Pierre Morel

Now that the PTF instruction is interpreted by the SIE we can optimize the arch_update_cpu_topology callback to check if there is a real need to update the topology by using the PTF instruction.

---
 arch/s390/kernel/topology.c | 3 +++
 1 file changed, 3 insertions(+)
arch/s390/kernel/topology.c +0 -0
index 26aa2614ee35..741cb447e78e 100644

arch/s390/kernel/topology.c int arch_update_cpu_topology(void) · line 327

Quoted 6 lines
arch/s390/kernel/topology.c +3 -0
@@ -322,6 +322,9 @@ int arch_update_cpu_topology(void)
 	struct device *dev;
 	int cpu, rc;
 
+	if (!ptf(PTF_CHECK))
+		return 0;
+

We have a timer which checks if topology changed and then triggers a call to arch_update_cpu_topology() via rebuild_sched_domains(). With this change topology changes would get lost.

#23 lore raw

Re: [PATCH v3 3/3] s390x: optimization of the check for CPU topology change

Quoted 18 lines · Heiko Carstens
Quoted 17 lines · Pierre Morel

Now that the PTF instruction is interpreted by the SIE we can optimize the arch_update_cpu_topology callback to check if there is a real need to update the topology by using the PTF instruction.

---
  arch/s390/kernel/topology.c | 3 +++
  1 file changed, 3 insertions(+)
arch/s390/kernel/topology.c +0 -0
index 26aa2614ee35..741cb447e78e 100644
@@ -322,6 +322,9 @@ int arch_update_cpu_topology(void)
  	struct device *dev;
  	int cpu, rc;
 
Quoted 7 lines

arch/s390/kernel/topology.c int arch_update_cpu_topology(void) · line 327

Quoted 3 lines
arch/s390/kernel/topology.c +3 -0
+	if (!ptf(PTF_CHECK))
+		return 0;
+

We have a timer which checks if topology changed and then triggers a call to arch_update_cpu_topology() via rebuild_sched_domains(). With this change topology changes would get lost.

For my understanding, if PTF check return 0 it means that there are no topology changes. So they could not get lost.

What did I miss?

Signature · 2 lines
Pierre Morel
IBM Lab Boeblingen
#24 lore raw

Re: [PATCH v3 3/3] s390x: optimization of the check for CPU topology change

Quoted 28 lines · Pierre Morel
Quoted 10 lines · Heiko Carstens
Quoted 9 lines · Pierre Morel

Now that the PTF instruction is interpreted by the SIE we can optimize the arch_update_cpu_topology callback to check if there is a real need to update the topology by using the PTF instruction.

---
  arch/s390/kernel/topology.c | 3 +++

  1 file changed, 3 insertions(+)
Quoted 14 lines

arch/s390/kernel/topology.c int arch_update_cpu_topology(void) · line 326

Quoted 10 lines
arch/s390/kernel/topology.c +3 -0
index 26aa2614ee35..741cb447e78e 100644
@@ -322,6 +322,9 @@ int arch_update_cpu_topology(void)
       struct device *dev;
       int cpu, rc;
+    if (!ptf(PTF_CHECK))
+        return 0;
+

We have a timer which checks if topology changed and then triggers a call to arch_update_cpu_topology() via rebuild_sched_domains(). With this change topology changes would get lost.

Quoted 6 lines

For my understanding, if PTF check return 0 it means that there are no topology changes. So they could not get lost.

What did I miss?

I missed that PTF clears the MCTR... and only one of the two calls will return 1 while we need both to return 1...

Signature · 2 lines
Pierre Morel
IBM Lab Boeblingen