Skip to contentSkip to conversation

KVM: s390x: add debug statement for diag 318 CPNC data

PATCH In review Collin Walling 3 participants 10 messages Oct 27, 2021 – Nov 8, 2021

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 KVM: s390x: add debug statement for diag 318 CPNC data

The diag 318 data contains values that denote information regarding the guest's environment. Currently, it is unecessarily difficult to observe this value (either manually-inserted debug statements, gdb stepping, mem dumping etc). It's useful to observe this information to obtain an at-a-glance view of the guest's environment, so lets add a simple VCPU event that prints the CPNC to the s390dbf logs.

---
 arch/s390/kvm/kvm-s390.c | 1 +
 1 file changed, 1 insertion(+)
arch/s390/kvm/kvm-s390.c +1 -0
index 6a6dd5e1daf6..da3ff24eabd0 100644
@@ -4254,6 +4254,7 @@ static void sync_regs_fmt2(struct kvm_vcpu *vcpu)
 	if (kvm_run->kvm_dirty_regs & KVM_SYNC_DIAG318) {
 		vcpu->arch.diag318_info.val = kvm_run->s.regs.diag318;
 		vcpu->arch.sie_block->cpnc = vcpu->arch.diag318_info.cpnc;
+		VCPU_EVENT(vcpu, 2, "setting cpnc to %d", vcpu->arch.diag318_info.cpnc);
 	}
 	/*
 	 * If userspace sets the riccb (e.g. after migration) to a valid state,
Signature · 1 line
2.31.1
Acked-by on v1 1/1
#1 lore raw

Am 27.10.21 um 04:54 schrieb Collin Walling:

Quoted 8 lines

The diag 318 data contains values that denote information regarding the guest's environment. Currently, it is unecessarily difficult to observe this value (either manually-inserted debug statements, gdb stepping, mem dumping etc). It's useful to observe this information to obtain an at-a-glance view of the guest's environment, so lets add a simple VCPU event that prints the CPNC to the s390dbf logs.

And it even finds a bug in QEMU. We clear the CPNC on local CPU resets. Can you have a look? I think we just have to move the cpnc in the env field from the normal/initial reset range to the full reset range.

Quoted 17 lines
---
  arch/s390/kvm/kvm-s390.c | 1 +
  1 file changed, 1 insertion(+)
arch/s390/kvm/kvm-s390.c +1 -0
index 6a6dd5e1daf6..da3ff24eabd0 100644
@@ -4254,6 +4254,7 @@ static void sync_regs_fmt2(struct kvm_vcpu *vcpu)
  	if (kvm_run->kvm_dirty_regs & KVM_SYNC_DIAG318) {
  		vcpu->arch.diag318_info.val = kvm_run->s.regs.diag318;
  		vcpu->arch.sie_block->cpnc = vcpu->arch.diag318_info.cpnc;
+		VCPU_EVENT(vcpu, 2, "setting cpnc to %d", vcpu->arch.diag318_info.cpnc);
  	}
  	/*
  	 * If userspace sets the riccb (e.g. after migration) to a valid state,
#2 lore raw
Quoted 8 lines · Christian Borntraeger

Am 27.10.21 um 04:54 schrieb Collin Walling:

Quoted 7 lines

The diag 318 data contains values that denote information regarding the guest's environment. Currently, it is unecessarily difficult to observe this value (either manually-inserted debug statements, gdb stepping, mem dumping etc). It's useful to observe this information to obtain an at-a-glance view of the guest's environment, so lets add a simple VCPU event that prints the CPNC to the s390dbf logs.

Quoted 7 lines
Quoted 1 line

And it even finds a bug in QEMU. We clear the CPNC on local CPU resets. Can you have a look? I think we just have to move the cpnc in the env field from the normal/initial reset range to the full reset range.

I'll take a look at this right away.

Quoted 19 lines
Quoted 19 lines

---
  arch/s390/kvm/kvm-s390.c | 1 +

  1 file changed, 1 insertion(+)
arch/s390/kvm/kvm-s390.c +1 -0
index 6a6dd5e1daf6..da3ff24eabd0 100644
@@ -4254,6 +4254,7 @@ static void sync_regs_fmt2(struct kvm_vcpu *vcpu)
       if (kvm_run->kvm_dirty_regs & KVM_SYNC_DIAG318) {
           vcpu->arch.diag318_info.val = kvm_run->s.regs.diag318;
           vcpu->arch.sie_block->cpnc = vcpu->arch.diag318_info.cpnc;
+        VCPU_EVENT(vcpu, 2, "setting cpnc to %d",
 vcpu->arch.diag318_info.cpnc);
       }
       /*

       * If userspace sets the riccb (e.g. after migration) to a valid state,

Signature · 4 lines
Regards,
Collin

Stay safe and stay healthy
#3 lore raw
In reply to Collin Walling

Am 27.10.21 um 04:54 schrieb Collin Walling:

Quoted 8 lines

The diag 318 data contains values that denote information regarding the guest's environment. Currently, it is unecessarily difficult to observe this value (either manually-inserted debug statements, gdb stepping, mem dumping etc). It's useful to observe this information to obtain an at-a-glance view of the guest's environment, so lets add a simple VCPU event that prints the CPNC to the s390dbf logs.

applied and queued

arch/s390/kvm/kvm-s390.c static void sync_regs_fmt2(struct kvm_vcpu *vcpu) · line 4257

Quoted 13 lines
---
  arch/s390/kvm/kvm-s390.c | 1 +
  1 file changed, 1 insertion(+)
arch/s390/kvm/kvm-s390.c +1 -0
index 6a6dd5e1daf6..da3ff24eabd0 100644
@@ -4254,6 +4254,7 @@ static void sync_regs_fmt2(struct kvm_vcpu *vcpu)
  	if (kvm_run->kvm_dirty_regs & KVM_SYNC_DIAG318) {
  		vcpu->arch.diag318_info.val = kvm_run->s.regs.diag318;
  		vcpu->arch.sie_block->cpnc = vcpu->arch.diag318_info.cpnc;
+		VCPU_EVENT(vcpu, 2, "setting cpnc to %d", vcpu->arch.diag318_info.cpnc);

After comparing this with the other events I think level==3 is better. Changed when applying.

Quoted 4 lines
arch/s390/kvm/kvm-s390.c +0 -0
  	}
  	/*
  	 * If userspace sets the riccb (e.g. after migration) to a valid state,
#4 lore raw
Quoted 15 lines · Christian Borntraeger

Am 27.10.21 um 04:54 schrieb Collin Walling:

Quoted 8 lines

The diag 318 data contains values that denote information regarding the guest's environment. Currently, it is unecessarily difficult to observe this value (either manually-inserted debug statements, gdb stepping, mem dumping etc). It's useful to observe this information to obtain an at-a-glance view of the guest's environment, so lets add a simple VCPU event that prints the CPNC to the s390dbf logs.

applied and queued

Quoted 4 lines

---
  arch/s390/kvm/kvm-s390.c | 1 +

  1 file changed, 1 insertion(+)
Quoted 18 lines

arch/s390/kvm/kvm-s390.c static void sync_regs_fmt2(struct kvm_vcpu *vcpu) · line 4258

Quoted 10 lines
arch/s390/kvm/kvm-s390.c +1 -0
index 6a6dd5e1daf6..da3ff24eabd0 100644
@@ -4254,6 +4254,7 @@ static void sync_regs_fmt2(struct kvm_vcpu *vcpu)
       if (kvm_run->kvm_dirty_regs & KVM_SYNC_DIAG318) {
           vcpu->arch.diag318_info.val = kvm_run->s.regs.diag318;
           vcpu->arch.sie_block->cpnc = vcpu->arch.diag318_info.cpnc;
+        VCPU_EVENT(vcpu, 2, "setting cpnc to %d",
 vcpu->arch.diag318_info.cpnc);

After comparing this with the other events I think level==3 is better. Changed when applying.

Quoted 5 lines
arch/s390/kvm/kvm-s390.c +0 -0
       }
       /*

       * If userspace sets the riccb (e.g. after migration) to a valid state,

Thanks!

Signature · 4 lines
Regards,
Collin

Stay safe and stay healthy
maintainer
#5 lore raw
In reply to Collin Walling
Quoted 18 lines · Collin Walling

The diag 318 data contains values that denote information regarding the guest's environment. Currently, it is unecessarily difficult to observe this value (either manually-inserted debug statements, gdb stepping, mem dumping etc). It's useful to observe this information to obtain an at-a-glance view of the guest's environment, so lets add a simple VCPU event that prints the CPNC to the s390dbf logs.

---
  arch/s390/kvm/kvm-s390.c | 1 +
  1 file changed, 1 insertion(+)
arch/s390/kvm/kvm-s390.c +0 -0
index 6a6dd5e1daf6..da3ff24eabd0 100644

arch/s390/kvm/kvm-s390.c static void sync_regs_fmt2(struct kvm_vcpu *vcpu) · line 4260

Quoted 7 lines
arch/s390/kvm/kvm-s390.c +1 -0
@@ -4254,6 +4254,7 @@ static void sync_regs_fmt2(struct kvm_vcpu *vcpu)
  	if (kvm_run->kvm_dirty_regs & KVM_SYNC_DIAG318) {
  		vcpu->arch.diag318_info.val = kvm_run->s.regs.diag318;
  		vcpu->arch.sie_block->cpnc = vcpu->arch.diag318_info.cpnc;
+		VCPU_EVENT(vcpu, 2, "setting cpnc to %d", vcpu->arch.diag318_info.cpnc);
  	}
  	/*
  	 * If userspace sets the riccb (e.g. after migration) to a valid state,

Won't that turn up for every vcpu and spam the log?

#6 lore raw

Am 08.11.21 um 12:12 schrieb Janosch Frank:

Quoted 13 lines · Collin Walling
Quoted 12 lines

The diag 318 data contains values that denote information regarding the guest's environment. Currently, it is unecessarily difficult to observe this value (either manually-inserted debug statements, gdb stepping, mem dumping etc). It's useful to observe this information to obtain an at-a-glance view of the guest's environment, so lets add a simple VCPU event that prints the CPNC to the s390dbf logs.

---
  arch/s390/kvm/kvm-s390.c | 1 +

  1 file changed, 1 insertion(+)
Quoted 15 lines

arch/s390/kvm/kvm-s390.c static void sync_regs_fmt2(struct kvm_vcpu *vcpu) · line 4260

Quoted 13 lines
arch/s390/kvm/kvm-s390.c +1 -0
index 6a6dd5e1daf6..da3ff24eabd0 100644
@@ -4254,6 +4254,7 @@ static void sync_regs_fmt2(struct kvm_vcpu *vcpu)
       if (kvm_run->kvm_dirty_regs & KVM_SYNC_DIAG318) {
           vcpu->arch.diag318_info.val = kvm_run->s.regs.diag318;
           vcpu->arch.sie_block->cpnc = vcpu->arch.diag318_info.cpnc;
+        VCPU_EVENT(vcpu, 2, "setting cpnc to %d", vcpu->arch.diag318_info.cpnc);
       }
       /*
        * If userspace sets the riccb (e.g. after migration) to a valid state,

Won't that turn up for every vcpu and spam the log?

only if the userspace always sets the dirty bit (which it should not).

maintainer
#7 lore raw
Quoted 32 lines · Christian Borntraeger

Am 08.11.21 um 12:12 schrieb Janosch Frank:

Quoted 13 lines · Collin Walling
Quoted 12 lines

The diag 318 data contains values that denote information regarding the guest's environment. Currently, it is unecessarily difficult to observe this value (either manually-inserted debug statements, gdb stepping, mem dumping etc). It's useful to observe this information to obtain an at-a-glance view of the guest's environment, so lets add a simple VCPU event that prints the CPNC to the s390dbf logs.

---
   arch/s390/kvm/kvm-s390.c | 1 +
   1 file changed, 1 insertion(+)
Quoted 15 lines

arch/s390/kvm/kvm-s390.c static void sync_regs_fmt2(struct kvm_vcpu *vcpu) · line 4260

Quoted 13 lines
arch/s390/kvm/kvm-s390.c +1 -0
index 6a6dd5e1daf6..da3ff24eabd0 100644
@@ -4254,6 +4254,7 @@ static void sync_regs_fmt2(struct kvm_vcpu *vcpu)
       if (kvm_run->kvm_dirty_regs & KVM_SYNC_DIAG318) {
           vcpu->arch.diag318_info.val = kvm_run->s.regs.diag318;
           vcpu->arch.sie_block->cpnc = vcpu->arch.diag318_info.cpnc;
+        VCPU_EVENT(vcpu, 2, "setting cpnc to %d", vcpu->arch.diag318_info.cpnc);
       }
       /*
        * If userspace sets the riccb (e.g. after migration) to a valid state,

Won't that turn up for every vcpu and spam the log?

Quoted 2 lines

only if the userspace always sets the dirty bit (which it should not).

But that's exactly what it does, no?
We do a loop over all vcpus and call kvm_s390_set_diag318() which sets the info in kvm_run and sets the diag318 bit in the kvm_dirty_regs.

@Collin: Could you check that please?

#8 lore raw

Am 08.11.21 um 13:48 schrieb Janosch Frank:

Quoted 33 lines · Christian Borntraeger
Quoted 32 lines

Am 08.11.21 um 12:12 schrieb Janosch Frank:

Quoted 28 lines · Collin Walling
Quoted 18 lines

The diag 318 data contains values that denote information regarding the guest's environment. Currently, it is unecessarily difficult to observe this value (either manually-inserted debug statements, gdb stepping, mem dumping etc). It's useful to observe this information to obtain an at-a-glance view of the guest's environment, so lets add a simple VCPU event that prints the CPNC to the s390dbf logs.

---
   arch/s390/kvm/kvm-s390.c | 1 +

   1 file changed, 1 insertion(+)
arch/s390/kvm/kvm-s390.c +0 -0
index 6a6dd5e1daf6..da3ff24eabd0 100644

arch/s390/kvm/kvm-s390.c static void sync_regs_fmt2(struct kvm_vcpu *vcpu) · line 4260

Quoted 7 lines
arch/s390/kvm/kvm-s390.c +1 -0
@@ -4254,6 +4254,7 @@ static void sync_regs_fmt2(struct kvm_vcpu *vcpu)
        if (kvm_run->kvm_dirty_regs & KVM_SYNC_DIAG318) {
            vcpu->arch.diag318_info.val = kvm_run->s.regs.diag318;
            vcpu->arch.sie_block->cpnc = vcpu->arch.diag318_info.cpnc;
+        VCPU_EVENT(vcpu, 2, "setting cpnc to %d", vcpu->arch.diag318_info.cpnc);
        }
        /*
         * If userspace sets the riccb (e.g. after migration) to a valid state,

Won't that turn up for every vcpu and spam the log?

Quoted 5 lines
Quoted 2 lines

only if the userspace always sets the dirty bit (which it should not).

But that's exactly what it does, no?
We do a loop over all vcpus and call kvm_s390_set_diag318() which sets the info in kvm_run and sets the diag318 bit in the kvm_dirty_regs.

Yes, ONCE per CPU. And this is exactly what I want to see. (and it did show a bug in qemu that we only set it for one cpu to the correct value).

Quoted 2 lines

@Collin: Could you check that please?

maintainer
#9 lore raw
Quoted 36 lines · Christian Borntraeger

Am 08.11.21 um 13:48 schrieb Janosch Frank:

Quoted 33 lines · Christian Borntraeger
Quoted 32 lines

Am 08.11.21 um 12:12 schrieb Janosch Frank:

Quoted 28 lines · Collin Walling
Quoted 18 lines

The diag 318 data contains values that denote information regarding the guest's environment. Currently, it is unecessarily difficult to observe this value (either manually-inserted debug statements, gdb stepping, mem dumping etc). It's useful to observe this information to obtain an at-a-glance view of the guest's environment, so lets add a simple VCPU event that prints the CPNC to the s390dbf logs.

---
    arch/s390/kvm/kvm-s390.c | 1 +
    1 file changed, 1 insertion(+)
arch/s390/kvm/kvm-s390.c +0 -0
index 6a6dd5e1daf6..da3ff24eabd0 100644

arch/s390/kvm/kvm-s390.c static void sync_regs_fmt2(struct kvm_vcpu *vcpu) · line 4260

Quoted 7 lines
arch/s390/kvm/kvm-s390.c +1 -0
@@ -4254,6 +4254,7 @@ static void sync_regs_fmt2(struct kvm_vcpu *vcpu)
        if (kvm_run->kvm_dirty_regs & KVM_SYNC_DIAG318) {
            vcpu->arch.diag318_info.val = kvm_run->s.regs.diag318;
            vcpu->arch.sie_block->cpnc = vcpu->arch.diag318_info.cpnc;
+        VCPU_EVENT(vcpu, 2, "setting cpnc to %d", vcpu->arch.diag318_info.cpnc);
        }
        /*
         * If userspace sets the riccb (e.g. after migration) to a valid state,

Won't that turn up for every vcpu and spam the log?

Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

only if the userspace always sets the dirty bit (which it should not).

But that's exactly what it does, no?
We do a loop over all vcpus and call kvm_s390_set_diag318() which sets the info in kvm_run and sets the diag318 bit in the kvm_dirty_regs.

Yes, ONCE per CPU. And this is exactly what I want to see. (and it did show a bug in qemu that we only set it for one cpu to the correct value).

Ok.
I didn't really want to have n entries in the log hence I asked.

318 is a bit weird as it's a per VM value we need to put into all sie blocks.

Quoted 3 lines
Quoted 3 lines

@Collin: Could you check that please?