Skip to contentSkip to conversation

KVM: arm64: vgic-v3: Missing check for redist region above the VM IPA size

PATCH Superseded Ricardo Koller 4 participants 20 messages Sep 8, 2021 – Sep 20, 2021

Subsystems (4)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

cover KVM: arm64: vgic-v3: Missing check for redist region above the VM IPA size

KVM doesn't check for redist regions that extend partially above the VM-specified IPA (phys_size). This can happen when using the KVM_VGIC_V3_ADDR_TYPE_REDIST attribute to set a new region that extends partially above phys_size (with the base below phys_size). The issue is that vcpus can potentially run into a situation where some redistributors are addressable and others are not.

Patch 1 adds the missing check, and patch 2 adds a test into aarch64/vgic_init.

Ricardo Koller (2):
KVM: arm64: vgic: check redist region is not above the VM IPA size
KVM: arm64: selftests: test for vgic redist above the VM IPA size

 arch/arm64/kvm/vgic/vgic-v3.c                 |  4 ++
 .../testing/selftests/kvm/aarch64/vgic_init.c | 44 +++++++++++++++++++
 2 files changed, 48 insertions(+)
Signature · 6 lines
2.33.0.153.gba50c8fa24-goog

_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
#1 lore raw

PATCH 1/2 KVM: arm64: vgic: check redist region is not above the VM IPA size

Extend vgic_v3_check_base() to verify that the redistributor regions don't go above the VM-specified IPA size (phys_size). This can happen when using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST attribute with:

base + size > phys_size AND base < phys_size

vgic_v3_check_base() is used to check the redist regions bases when setting them (with the vcpus added so far) and when attempting the first vcpu-run.

---
 arch/arm64/kvm/vgic/vgic-v3.c | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm64/kvm/vgic/vgic-v3.c +4 -0
index 66004f61cd83..5afd9f6f68f6 100644
@@ -512,6 +512,10 @@ bool vgic_v3_check_base(struct kvm *kvm)
 		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) <
 			rdreg->base)
 			return false;
+
+		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) >
+			kvm_phys_size(kvm))
+			return false;
 	}
 
 	if (IS_VGIC_ADDR_UNDEF(d->vgic_dist_base))
Signature · 6 lines
2.33.0.153.gba50c8fa24-goog

_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
#2 lore raw

Re: [PATCH 1/2] KVM: arm64: vgic: check redist region is not above the VM IPA size

Hi Ricardo,

Quoted 20 lines · Ricardo Koller

Extend vgic_v3_check_base() to verify that the redistributor regions don't go above the VM-specified IPA size (phys_size). This can happen when using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST attribute with:

base + size > phys_size AND base < phys_size

vgic_v3_check_base() is used to check the redist regions bases when setting them (with the vcpus added so far) and when attempting the first vcpu-run.

---
 arch/arm64/kvm/vgic/vgic-v3.c | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
index 66004f61cd83..5afd9f6f68f6 100644

arch/arm64/kvm/vgic/vgic-v3.c bool vgic_v3_check_base(struct kvm *kvm) · line 514

Quoted 3 lines
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
@@ -512,6 +512,10 @@ bool vgic_v3_check_base(struct kvm *kvm)
 		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) <
 			rdreg->base)
 			return false;

Can we drop this check in favor of explicitly comparing rdreg->base with kvm_phys_size()? I believe that would be more readable.

Quoted 10 lines
arch/arm64/kvm/vgic/vgic-v3.c +4 -0
+
+		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) >
+			kvm_phys_size(kvm))
+			return false;
 	}
 
 	if (IS_VGIC_ADDR_UNDEF(d->vgic_dist_base))
Signature · 1 line
2.33.0.153.gba50c8fa24-goog

--
Thanks,
Oliver

#3 lore raw

Re: [PATCH 1/2] KVM: arm64: vgic: check redist region is not above the VM IPA size

Quoted 18 lines · Oliver Upton

Hi Ricardo,

Quoted 15 lines · Ricardo Koller

Extend vgic_v3_check_base() to verify that the redistributor regions don't go above the VM-specified IPA size (phys_size). This can happen when using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST attribute with:

base + size > phys_size AND base < phys_size

vgic_v3_check_base() is used to check the redist regions bases when setting them (with the vcpus added so far) and when attempting the first vcpu-run.

---
 arch/arm64/kvm/vgic/vgic-v3.c | 4 ++++
 1 file changed, 4 insertions(+)
Quoted 12 lines

arch/arm64/kvm/vgic/vgic-v3.c bool vgic_v3_check_base(struct kvm *kvm) · line 514

Quoted 8 lines
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
index 66004f61cd83..5afd9f6f68f6 100644
@@ -512,6 +512,10 @@ bool vgic_v3_check_base(struct kvm *kvm)
 		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) <
 			rdreg->base)
 			return false;

Can we drop this check in favor of explicitly comparing rdreg->base with kvm_phys_size()? I believe that would be more readable.

You mean the integer overflow check above? in that case, I would prefer to leave it, if you don't mind. It seems that this type of check is used in some other places in KVM (like kvm_assign_ioeventfd and vgic_v3_alloc_redist_region).

Quoted 14 lines
Quoted 10 lines
arch/arm64/kvm/vgic/vgic-v3.c +4 -0
+
+		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) >
+			kvm_phys_size(kvm))
+			return false;
 	}
 
 	if (IS_VGIC_ADDR_UNDEF(d->vgic_dist_base))
Signature · 1 line
2.33.0.153.gba50c8fa24-goog

--
Thanks,
Oliver

Reviewed-by
#4 lore raw

Re: [PATCH 1/2] KVM: arm64: vgic: check redist region is not above the VM IPA size

Quoted 33 lines · Ricardo Koller
Quoted 18 lines · Oliver Upton

Hi Ricardo,

Quoted 15 lines · Ricardo Koller

Extend vgic_v3_check_base() to verify that the redistributor regions don't go above the VM-specified IPA size (phys_size). This can happen when using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST attribute with:

base + size > phys_size AND base < phys_size

vgic_v3_check_base() is used to check the redist regions bases when setting them (with the vcpus added so far) and when attempting the first vcpu-run.

---
 arch/arm64/kvm/vgic/vgic-v3.c | 4 ++++
 1 file changed, 4 insertions(+)
Quoted 12 lines

arch/arm64/kvm/vgic/vgic-v3.c bool vgic_v3_check_base(struct kvm *kvm) · line 514

Quoted 8 lines
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
index 66004f61cd83..5afd9f6f68f6 100644
@@ -512,6 +512,10 @@ bool vgic_v3_check_base(struct kvm *kvm)
            if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) <
                    rdreg->base)
                    return false;

Can we drop this check in favor of explicitly comparing rdreg->base with kvm_phys_size()? I believe that would be more readable.

Quoted 4 lines

You mean the integer overflow check above? in that case, I would prefer to leave it, if you don't mind. It seems that this type of check is used in some other places in KVM (like kvm_assign_ioeventfd and vgic_v3_alloc_redist_region).

I would expect rdreg->base to exceed kvm_phys_size() before wrapping, but we do derive rdreg->count from what userspace gives us. In that case, your addition in combination with this makes sense, so no real objections here.

Quoted 6 lines
Quoted 6 lines
Quoted 6 lines
arch/arm64/kvm/vgic/vgic-v3.c +4 -0
+
+           if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) >
+                   kvm_phys_size(kvm))
+                   return false;
    }
 
Quoted 8 lines
Quoted 8 lines
Quoted 4 lines
arch/arm64/kvm/vgic/vgic-v3.c +0 -1
    if (IS_VGIC_ADDR_UNDEF(d->vgic_dist_base))
--

2.33.0.153.gba50c8fa24-goog

--
Thanks,
Oliver

#5 lore raw

Re: [PATCH 1/2] KVM: arm64: vgic: check redist region is not above the VM IPA size

In reply to Ricardo Koller

Hi Ricardo,

Quoted 20 lines · Ricardo Koller

Extend vgic_v3_check_base() to verify that the redistributor regions don't go above the VM-specified IPA size (phys_size). This can happen when using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST attribute with:

base + size > phys_size AND base < phys_size

vgic_v3_check_base() is used to check the redist regions bases when setting them (with the vcpus added so far) and when attempting the first vcpu-run.

---
 arch/arm64/kvm/vgic/vgic-v3.c | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
index 66004f61cd83..5afd9f6f68f6 100644

arch/arm64/kvm/vgic/vgic-v3.c bool vgic_v3_check_base(struct kvm *kvm) · line 518

Quoted 7 lines
arch/arm64/kvm/vgic/vgic-v3.c +4 -0
@@ -512,6 +512,10 @@ bool vgic_v3_check_base(struct kvm *kvm)
 		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) <
 			rdreg->base)
 			return false;
+
+		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) >
+			kvm_phys_size(kvm))
+			return false;

Looks to me like this same check (and the overflow one before it) is done when adding a new Redistributor region in kvm_vgic_addr() -> vgic_v3_set_redist_base() -> vgic_v3_alloc_redist_region() -> vgic_check_ioaddr(). As far as I can tell, kvm_vgic_addr() handles both ways of setting the Redistributor address.

Without this patch, did you manage to set a base address such that base + size > kvm_phys_size()?

Thanks,

Alex

Quoted 3 lines
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
 	}
 
 	if (IS_VGIC_ADDR_UNDEF(d->vgic_dist_base))
#6 lore raw

Re: [PATCH 1/2] KVM: arm64: vgic: check redist region is not above the VM IPA size

Hi,

Quoted 18 lines · Alexandru Elisei

Hi Ricardo,

Quoted 15 lines · Ricardo Koller

Extend vgic_v3_check_base() to verify that the redistributor regions don't go above the VM-specified IPA size (phys_size). This can happen when using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST attribute with:

base + size > phys_size AND base < phys_size

vgic_v3_check_base() is used to check the redist regions bases when setting them (with the vcpus added so far) and when attempting the first vcpu-run.

---
 arch/arm64/kvm/vgic/vgic-v3.c | 4 ++++
 1 file changed, 4 insertions(+)
Quoted 16 lines

arch/arm64/kvm/vgic/vgic-v3.c bool vgic_v3_check_base(struct kvm *kvm) · line 518

Quoted 12 lines
arch/arm64/kvm/vgic/vgic-v3.c +4 -0
index 66004f61cd83..5afd9f6f68f6 100644
@@ -512,6 +512,10 @@ bool vgic_v3_check_base(struct kvm *kvm)
 		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) <
 			rdreg->base)
 			return false;
+
+		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) >
+			kvm_phys_size(kvm))
+			return false;

Looks to me like this same check (and the overflow one before it) is done when adding a new Redistributor region in kvm_vgic_addr() -> vgic_v3_set_redist_base() -> vgic_v3_alloc_redist_region() -> vgic_check_ioaddr(). As far as I can tell, kvm_vgic_addr() handles both ways of setting the Redistributor address.

To me vgic_check_ioaddr() does check the base addr but not the end addr. So looks this fix is needed.
As I commented on the selftest patch, I think you should double check your fix also handles the KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION case.

In vgic_v3_alloc_redist_region(), in this later case, we know the number of redistributors in the region (count), so it would be easy to check the end addr. But I think this would be a duplicate of your new check as vgic_v3_check_base() also gets called in vgic_register_redist_iodev(). But better to check it ;-)

Thanks

Eric

Quoted 11 lines

Without this patch, did you manage to set a base address such that base + size > kvm_phys_size()?

Thanks,

Alex

Quoted 3 lines
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
 	}
 
 	if (IS_VGIC_ADDR_UNDEF(d->vgic_dist_base))

_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

#7 lore raw

Re: [PATCH 1/2] KVM: arm64: vgic: check redist region is not above the VM IPA size

In reply to Alexandru Elisei
Quoted 36 lines · Alexandru Elisei

Hi Ricardo,

Quoted 20 lines · Ricardo Koller

Extend vgic_v3_check_base() to verify that the redistributor regions don't go above the VM-specified IPA size (phys_size). This can happen when using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST attribute with:

base + size > phys_size AND base < phys_size

vgic_v3_check_base() is used to check the redist regions bases when setting them (with the vcpus added so far) and when attempting the first vcpu-run.

---
 arch/arm64/kvm/vgic/vgic-v3.c | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
index 66004f61cd83..5afd9f6f68f6 100644

arch/arm64/kvm/vgic/vgic-v3.c bool vgic_v3_check_base(struct kvm *kvm) · line 518

Quoted 7 lines
arch/arm64/kvm/vgic/vgic-v3.c +4 -0
@@ -512,6 +512,10 @@ bool vgic_v3_check_base(struct kvm *kvm)
 		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) <
 			rdreg->base)
 			return false;
+
+		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) >
+			kvm_phys_size(kvm))
+			return false;

Looks to me like this same check (and the overflow one before it) is done when adding a new Redistributor region in kvm_vgic_addr() -> vgic_v3_set_redist_base() -> vgic_v3_alloc_redist_region() -> vgic_check_ioaddr(). As far as I can tell, kvm_vgic_addr() handles both ways of setting the Redistributor address.

Quoted 3 lines

Without this patch, did you manage to set a base address such that base + size > kvm_phys_size()?

Yes, with the KVM_VGIC_V3_ADDR_TYPE_REDIST legacy API. The easiest way to get to this situation is with the selftest in patch 2. I then tried an extra experiment: map the first redistributor, run the first vcpu, and access the redist from inside the guest. KVM didn't complain in any of these steps.

Thanks,
Ricardo

Quoted 7 lines

Thanks,

Alex

Quoted 3 lines
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
 	}
 
 	if (IS_VGIC_ADDR_UNDEF(d->vgic_dist_base))
#8 lore raw

Re: [PATCH 1/2] KVM: arm64: vgic: check redist region is not above the VM IPA size

Hi Ricardo,

Quoted 36 lines · Ricardo Koller
Quoted 35 lines · Alexandru Elisei

Hi Ricardo,

Quoted 20 lines · Ricardo Koller

Extend vgic_v3_check_base() to verify that the redistributor regions don't go above the VM-specified IPA size (phys_size). This can happen when using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST attribute with:

base + size > phys_size AND base < phys_size

vgic_v3_check_base() is used to check the redist regions bases when setting them (with the vcpus added so far) and when attempting the first vcpu-run.

---
 arch/arm64/kvm/vgic/vgic-v3.c | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
index 66004f61cd83..5afd9f6f68f6 100644

arch/arm64/kvm/vgic/vgic-v3.c bool vgic_v3_check_base(struct kvm *kvm) · line 518

Quoted 7 lines
arch/arm64/kvm/vgic/vgic-v3.c +4 -0
@@ -512,6 +512,10 @@ bool vgic_v3_check_base(struct kvm *kvm)
 		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) <
 			rdreg->base)
 			return false;
+
+		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) >
+			kvm_phys_size(kvm))
+			return false;

Looks to me like this same check (and the overflow one before it) is done when adding a new Redistributor region in kvm_vgic_addr() -> vgic_v3_set_redist_base() -> vgic_v3_alloc_redist_region() -> vgic_check_ioaddr(). As far as I can tell, kvm_vgic_addr() handles both ways of setting the Redistributor address.

Quoted 8 lines
Quoted 3 lines

Without this patch, did you manage to set a base address such that base + size > kvm_phys_size()?

Yes, with the KVM_VGIC_V3_ADDR_TYPE_REDIST legacy API. The easiest way to get to this situation is with the selftest in patch 2. I then tried an extra experiment: map the first redistributor, run the first vcpu, and access the redist from inside the guest. KVM didn't complain in any of these steps.

Yes, Eric pointed out that I was mistaken and there is no check being done for base + size > kvm_phys_size().

What I was trying to say is that this check is better done when the user creates a Redistributor region, not when a VCPU is first run. We have everything we need to make the check when a region is created, why wait until the VCPU is run?

For example, vgic_v3_insert_redist_region() is called each time the adds a new Redistributor region (via either of the two APIs), and already has a check for the upper limit overflowing (identical to the check in vgic_v3_check_base()). I would add the check against the maximum IPA size there.

Also, because vgic_v3_insert_redist_region() already checks for overflow, I believe the overflow check in vgic_v3_check_base() is redundant.

As far as I can tell, vgic_v3_check_base() is there to make sure that the Distributor doesn't overlap with any of the Redistributors, and because the Redistributors and the Distributor can be created in any order, we defer the check until the first VCPU is run. I might be wrong about this, someone please correct me if I'm wrong.

Also, did you verify that KVM is also doing this check for GICv2? KVM does something similar and calls vgic_v2_check_base() when mapping the GIC resources, and I don't see a check for the maximum IPA size in that function either.

Thanks,

Alex

Quoted 11 lines

Thanks,
Ricardo

Quoted 7 lines

Thanks,

Alex

Quoted 3 lines
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
 	}
 
 	if (IS_VGIC_ADDR_UNDEF(d->vgic_dist_base))
#9 lore raw

Re: [PATCH 1/2] KVM: arm64: vgic: check redist region is not above the VM IPA size

Hi Alexandru,

Quoted 50 lines · Alexandru Elisei

Hi Ricardo,

Quoted 36 lines · Ricardo Koller
Quoted 35 lines · Alexandru Elisei

Hi Ricardo,

Quoted 20 lines · Ricardo Koller

Extend vgic_v3_check_base() to verify that the redistributor regions don't go above the VM-specified IPA size (phys_size). This can happen when using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST attribute with:

base + size > phys_size AND base < phys_size

vgic_v3_check_base() is used to check the redist regions bases when setting them (with the vcpus added so far) and when attempting the first vcpu-run.

---
 arch/arm64/kvm/vgic/vgic-v3.c | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
index 66004f61cd83..5afd9f6f68f6 100644

arch/arm64/kvm/vgic/vgic-v3.c bool vgic_v3_check_base(struct kvm *kvm) · line 518

Quoted 7 lines
arch/arm64/kvm/vgic/vgic-v3.c +4 -0
@@ -512,6 +512,10 @@ bool vgic_v3_check_base(struct kvm *kvm)
 		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) <
 			rdreg->base)
 			return false;
+
+		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) >
+			kvm_phys_size(kvm))
+			return false;

Looks to me like this same check (and the overflow one before it) is done when adding a new Redistributor region in kvm_vgic_addr() -> vgic_v3_set_redist_base() -> vgic_v3_alloc_redist_region() -> vgic_check_ioaddr(). As far as I can tell, kvm_vgic_addr() handles both ways of setting the Redistributor address.

Quoted 8 lines
Quoted 3 lines

Without this patch, did you manage to set a base address such that base + size > kvm_phys_size()?

Yes, with the KVM_VGIC_V3_ADDR_TYPE_REDIST legacy API. The easiest way to get to this situation is with the selftest in patch 2. I then tried an extra experiment: map the first redistributor, run the first vcpu, and access the redist from inside the guest. KVM didn't complain in any of these steps.

Yes, Eric pointed out that I was mistaken and there is no check being done for base + size > kvm_phys_size().

Quoted 8 lines

What I was trying to say is that this check is better done when the user creates a Redistributor region, not when a VCPU is first run. We have everything we need to make the check when a region is created, why wait until the VCPU is run?

For example, vgic_v3_insert_redist_region() is called each time the adds a new Redistributor region (via either of the two APIs), and already has a check for the upper limit overflowing (identical to the check in vgic_v3_check_base()). I would add the check against the maximum IPA size there.

you seem to refer to an old kernel as vgic_v3_insert_redist_region was renamed into  vgic_v3_alloc_redist_region in
e5a35635464b kvm: arm64: vgic-v3: Introduce vgic_v3_free_redist_region()

I think in case you use the old rdist API you do not know yet the size of the redist region at this point (count=0), hence Ricardo's choice to do the check latter.

Quoted 10 lines

Also, because vgic_v3_insert_redist_region() already checks for overflow, I believe the overflow check in vgic_v3_check_base() is redundant.

As far as I can tell, vgic_v3_check_base() is there to make sure that the Distributor doesn't overlap with any of the Redistributors, and because the Redistributors and the Distributor can be created in any order, we defer the check until the first VCPU is run. I might be wrong about this, someone please correct me if I'm wrong.

Quoted 3 lines

Also, did you verify that KVM is also doing this check for GICv2? KVM does something similar and calls vgic_v2_check_base() when mapping the GIC resources, and I don't see a check for the maximum IPA size in that function either.

I think vgic_check_ioaddr() called in kvm_vgic_addr() does the job (it checks the base @)

Thanks

Eric

Quoted 15 lines

Thanks,

Alex

Quoted 10 lines

Thanks,
Ricardo

Quoted 7 lines

Thanks,

Alex

Quoted 3 lines
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
 	}
 
 	if (IS_VGIC_ADDR_UNDEF(d->vgic_dist_base))
#10 lore raw

Re: [PATCH 1/2] KVM: arm64: vgic: check redist region is not above the VM IPA size

Hi Alexandru and Eric,

Quoted 61 lines · Eric Auger

Hi Alexandru,

Quoted 50 lines · Alexandru Elisei

Hi Ricardo,

Quoted 36 lines · Ricardo Koller
Quoted 35 lines · Alexandru Elisei

Hi Ricardo,

Quoted 20 lines · Ricardo Koller

Extend vgic_v3_check_base() to verify that the redistributor regions don't go above the VM-specified IPA size (phys_size). This can happen when using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST attribute with:

base + size > phys_size AND base < phys_size

vgic_v3_check_base() is used to check the redist regions bases when setting them (with the vcpus added so far) and when attempting the first vcpu-run.

---
 arch/arm64/kvm/vgic/vgic-v3.c | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
index 66004f61cd83..5afd9f6f68f6 100644

arch/arm64/kvm/vgic/vgic-v3.c bool vgic_v3_check_base(struct kvm *kvm) · line 518

Quoted 7 lines
arch/arm64/kvm/vgic/vgic-v3.c +4 -0
@@ -512,6 +512,10 @@ bool vgic_v3_check_base(struct kvm *kvm)
 		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) <
 			rdreg->base)
 			return false;
+
+		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) >
+			kvm_phys_size(kvm))
+			return false;

Looks to me like this same check (and the overflow one before it) is done when adding a new Redistributor region in kvm_vgic_addr() -> vgic_v3_set_redist_base() -> vgic_v3_alloc_redist_region() -> vgic_check_ioaddr(). As far as I can tell, kvm_vgic_addr() handles both ways of setting the Redistributor address.

Quoted 8 lines
Quoted 3 lines

Without this patch, did you manage to set a base address such that base + size > kvm_phys_size()?

Yes, with the KVM_VGIC_V3_ADDR_TYPE_REDIST legacy API. The easiest way to get to this situation is with the selftest in patch 2. I then tried an extra experiment: map the first redistributor, run the first vcpu, and access the redist from inside the guest. KVM didn't complain in any of these steps.

Yes, Eric pointed out that I was mistaken and there is no check being done for base + size > kvm_phys_size().

Quoted 8 lines

What I was trying to say is that this check is better done when the user creates a Redistributor region, not when a VCPU is first run. We have everything we need to make the check when a region is created, why wait until the VCPU is run?

For example, vgic_v3_insert_redist_region() is called each time the adds a new Redistributor region (via either of the two APIs), and already has a check for the upper limit overflowing (identical to the check in vgic_v3_check_base()). I would add the check against the maximum IPA size there.

Quoted 7 lines

you seem to refer to an old kernel as vgic_v3_insert_redist_region was renamed into  vgic_v3_alloc_redist_region in
e5a35635464b kvm: arm64: vgic-v3: Introduce vgic_v3_free_redist_region()

I think in case you use the old rdist API you do not know yet the size of the redist region at this point (count=0), hence Ricardo's choice to do the check latter.

Just wanted to add one more detail. vgic_v3_check_base() is also called when creating the redistributor region (via vgic_v3_set_redist_base -> vgic_register_redist_iodev). This patch reuses that check for the old redist API to also check for "base + size > kvm_phys_size()" with a size calculated using the vcpus added so far.

Quoted 4 lines
Quoted 4 lines

Also, because vgic_v3_insert_redist_region() already checks for overflow, I believe the overflow check in vgic_v3_check_base() is redundant.

It's redundant for the new redist API, but still needed for the old redist API.

Quoted 6 lines
Quoted 6 lines

As far as I can tell, vgic_v3_check_base() is there to make sure that the Distributor doesn't overlap with any of the Redistributors, and because the Redistributors and the Distributor can be created in any order, we defer the check until the first VCPU is run. I might be wrong about this, someone please correct me if I'm wrong.

Quoted 7 lines
Quoted 3 lines

Also, did you verify that KVM is also doing this check for GICv2? KVM does something similar and calls vgic_v2_check_base() when mapping the GIC resources, and I don't see a check for the maximum IPA size in that function either.

I think vgic_check_ioaddr() called in kvm_vgic_addr() does the job (it checks the base @)

It seems that GICv2 suffers from the same problem. The cpu interface base is checked but the end can extend above IPA size. Note that the cpu interface is 8KBs and vgic_check_ioaddr() is only checking that its base is 4KB aligned and below IPA size. The distributor region is 4KB so vgic_check_ioaddr() is enough in that case.

What about the following?

I can work on the next version of this patch (v2 has the GICv2 issue) which adds vgic_check_range(), which is like vgic_check_ioaddr() but with a size arg. kvm_vgic_addr() can then call vgic_check_range() and do all the checks for GICv2 and GICv3. Note that for GICv2, there's no need to wait until first vcpu run to do the check. Also note that I will have to keep the change in vgic_v3_check_base() to check for the old v3 redist API at first vcpu run.

Thanks,
Ricardo

Quoted 19 lines

Thanks

Eric

Quoted 15 lines

Thanks,

Alex

Quoted 10 lines

Thanks,
Ricardo

Quoted 7 lines

Thanks,

Alex

Quoted 3 lines
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
 	}
 
 	if (IS_VGIC_ADDR_UNDEF(d->vgic_dist_base))
#11 lore raw

Re: [PATCH 1/2] KVM: arm64: vgic: check redist region is not above the VM IPA size

Hi Ricardo,

Quoted 92 lines · Ricardo Koller

Hi Alexandru and Eric,

Quoted 61 lines · Eric Auger

Hi Alexandru,

Quoted 50 lines · Alexandru Elisei

Hi Ricardo,

Quoted 36 lines · Ricardo Koller
Quoted 35 lines · Alexandru Elisei

Hi Ricardo,

Quoted 20 lines · Ricardo Koller

Extend vgic_v3_check_base() to verify that the redistributor regions don't go above the VM-specified IPA size (phys_size). This can happen when using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST attribute with:

base + size > phys_size AND base < phys_size

vgic_v3_check_base() is used to check the redist regions bases when setting them (with the vcpus added so far) and when attempting the first vcpu-run.

---
 arch/arm64/kvm/vgic/vgic-v3.c | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
index 66004f61cd83..5afd9f6f68f6 100644

arch/arm64/kvm/vgic/vgic-v3.c bool vgic_v3_check_base(struct kvm *kvm) · line 518

Quoted 7 lines
arch/arm64/kvm/vgic/vgic-v3.c +4 -0
@@ -512,6 +512,10 @@ bool vgic_v3_check_base(struct kvm *kvm)
 		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) <
 			rdreg->base)
 			return false;
+
+		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) >
+			kvm_phys_size(kvm))
+			return false;

Looks to me like this same check (and the overflow one before it) is done when adding a new Redistributor region in kvm_vgic_addr() -> vgic_v3_set_redist_base() -> vgic_v3_alloc_redist_region() -> vgic_check_ioaddr(). As far as I can tell, kvm_vgic_addr() handles both ways of setting the Redistributor address.

Quoted 8 lines
Quoted 3 lines

Without this patch, did you manage to set a base address such that base + size > kvm_phys_size()?

Yes, with the KVM_VGIC_V3_ADDR_TYPE_REDIST legacy API. The easiest way to get to this situation is with the selftest in patch 2. I then tried an extra experiment: map the first redistributor, run the first vcpu, and access the redist from inside the guest. KVM didn't complain in any of these steps.

Yes, Eric pointed out that I was mistaken and there is no check being done for base + size > kvm_phys_size().

Quoted 8 lines

What I was trying to say is that this check is better done when the user creates a Redistributor region, not when a VCPU is first run. We have everything we need to make the check when a region is created, why wait until the VCPU is run?

For example, vgic_v3_insert_redist_region() is called each time the adds a new Redistributor region (via either of the two APIs), and already has a check for the upper limit overflowing (identical to the check in vgic_v3_check_base()). I would add the check against the maximum IPA size there.

Quoted 7 lines

you seem to refer to an old kernel as vgic_v3_insert_redist_region was renamed into  vgic_v3_alloc_redist_region in
e5a35635464b kvm: arm64: vgic-v3: Introduce vgic_v3_free_redist_region()

I think in case you use the old rdist API you do not know yet the size of the redist region at this point (count=0), hence Ricardo's choice to do the check latter.

Just wanted to add one more detail. vgic_v3_check_base() is also called when creating the redistributor region (via vgic_v3_set_redist_base -> vgic_register_redist_iodev). This patch reuses that check for the old redist API to also check for "base + size > kvm_phys_size()" with a size calculated using the vcpus added so far.

Quoted 3 lines
Quoted 3 lines

Also, because vgic_v3_insert_redist_region() already checks for overflow, I believe the overflow check in vgic_v3_check_base() is redundant.

It's redundant for the new redist API, but still needed for the old redist API.

Quoted 9 lines
Quoted 9 lines

As far as I can tell, vgic_v3_check_base() is there to make sure that the Distributor doesn't overlap with any of the Redistributors, and because the Redistributors and the Distributor can be created in any order, we defer the check until the first VCPU is run. I might be wrong about this, someone please correct me if I'm wrong.

Also, did you verify that KVM is also doing this check for GICv2? KVM does something similar and calls vgic_v2_check_base() when mapping the GIC resources, and I don't see a check for the maximum IPA size in that function either.

Quoted 6 lines
Quoted 3 lines

I think vgic_check_ioaddr() called in kvm_vgic_addr() does the job (it checks the base @)

It seems that GICv2 suffers from the same problem. The cpu interface base is checked but the end can extend above IPA size. Note that the cpu interface is 8KBs and vgic_check_ioaddr() is only checking that its base

I missed this 8kB thingy.

Quoted 5 lines

is 4KB aligned and below IPA size. The distributor region is 4KB so vgic_check_ioaddr() is enough in that case.

What about the following?

Quoted 7 lines

I can work on the next version of this patch (v2 has the GICv2 issue) which adds vgic_check_range(), which is like vgic_check_ioaddr() but with a size arg. kvm_vgic_addr() can then call vgic_check_range() and do all the checks for GICv2 and GICv3. Note that for GICv2, there's no need to wait until first vcpu run to do the check. Also note that I will have to keep the change in vgic_v3_check_base() to check for the old v3 redist API at first vcpu run.

Looking forward to reviewing it.

Thanks

Eric

Quoted 21 lines

Thanks,
Ricardo

Quoted 17 lines

Thanks

Eric

Quoted 14 lines

Thanks,

Alex

Quoted 10 lines

Thanks,
Ricardo

Quoted 7 lines

Thanks,

Alex

Quoted 3 lines
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
 	}
 
 	if (IS_VGIC_ADDR_UNDEF(d->vgic_dist_base))

_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

#12 lore raw

Re: [PATCH 1/2] KVM: arm64: vgic: check redist region is not above the VM IPA size

In reply to Ricardo Koller

Hi Eric, Ricardo,

Quoted 68 lines · Ricardo Koller

Hi Alexandru and Eric,

Quoted 65 lines · Eric Auger

Hi Alexandru,

Quoted 50 lines · Alexandru Elisei

Hi Ricardo,

Quoted 36 lines · Ricardo Koller
Quoted 35 lines · Alexandru Elisei

Hi Ricardo,

Quoted 20 lines · Ricardo Koller

Extend vgic_v3_check_base() to verify that the redistributor regions don't go above the VM-specified IPA size (phys_size). This can happen when using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST attribute with:

base + size > phys_size AND base < phys_size

vgic_v3_check_base() is used to check the redist regions bases when setting them (with the vcpus added so far) and when attempting the first vcpu-run.

---
 arch/arm64/kvm/vgic/vgic-v3.c | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
index 66004f61cd83..5afd9f6f68f6 100644

arch/arm64/kvm/vgic/vgic-v3.c bool vgic_v3_check_base(struct kvm *kvm) · line 518

Quoted 7 lines
arch/arm64/kvm/vgic/vgic-v3.c +4 -0
@@ -512,6 +512,10 @@ bool vgic_v3_check_base(struct kvm *kvm)
 		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) <
 			rdreg->base)
 			return false;
+
+		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) >
+			kvm_phys_size(kvm))
+			return false;

Looks to me like this same check (and the overflow one before it) is done when adding a new Redistributor region in kvm_vgic_addr() -> vgic_v3_set_redist_base() -> vgic_v3_alloc_redist_region() -> vgic_check_ioaddr(). As far as I can tell, kvm_vgic_addr() handles both ways of setting the Redistributor address.

Quoted 8 lines
Quoted 3 lines

Without this patch, did you manage to set a base address such that base + size > kvm_phys_size()?

Yes, with the KVM_VGIC_V3_ADDR_TYPE_REDIST legacy API. The easiest way to get to this situation is with the selftest in patch 2. I then tried an extra experiment: map the first redistributor, run the first vcpu, and access the redist from inside the guest. KVM didn't complain in any of these steps.

Yes, Eric pointed out that I was mistaken and there is no check being done for base + size > kvm_phys_size().

Quoted 8 lines

What I was trying to say is that this check is better done when the user creates a Redistributor region, not when a VCPU is first run. We have everything we need to make the check when a region is created, why wait until the VCPU is run?

For example, vgic_v3_insert_redist_region() is called each time the adds a new Redistributor region (via either of the two APIs), and already has a check for the upper limit overflowing (identical to the check in vgic_v3_check_base()). I would add the check against the maximum IPA size there.

you seem to refer to an old kernel as vgic_v3_insert_redist_region was renamed into� vgic_v3_alloc_redist_region in
e5a35635464b kvm: arm64: vgic-v3: Introduce vgic_v3_free_redist_region()

Quoted 8 lines
Quoted 3 lines

I think in case you use the old rdist API you do not know yet the size of the redist region at this point (count=0), hence Ricardo's choice to do the check latter.

Just wanted to add one more detail. vgic_v3_check_base() is also called when creating the redistributor region (via vgic_v3_set_redist_base -> vgic_register_redist_iodev). This patch reuses that check for the old redist API to also check for "base + size > kvm_phys_size()" with a size calculated using the vcpus added so far.

@Eric: Indeed I was looking at an older kernel by mistake, thank you for pointing that out!

Thank you both for the explanations, the piece I was missing was the fact that KVM_VGIC_V3_ADDR_TYPE_REDIST specifies only the base address and the limit for the region is the number of VCPUs * (KVM_VGIC_V3_REDIST_SIZE = 128K), which makes it necessary to have the check when each VCPU is first run (as far as I can tell, VCPUs can be created at any time).

Quoted 6 lines
Quoted 4 lines
Quoted 4 lines

Also, because vgic_v3_insert_redist_region() already checks for overflow, I believe the overflow check in vgic_v3_check_base() is redundant.

It's redundant for the new redist API, but still needed for the old redist API.

Indeed.

Quoted 10 lines
Quoted 10 lines
Quoted 10 lines

As far as I can tell, vgic_v3_check_base() is there to make sure that the Distributor doesn't overlap with any of the Redistributors, and because the Redistributors and the Distributor can be created in any order, we defer the check until the first VCPU is run. I might be wrong about this, someone please correct me if I'm wrong.

Also, did you verify that KVM is also doing this check for GICv2? KVM does something similar and calls vgic_v2_check_base() when mapping the GIC resources, and I don't see a check for the maximum IPA size in that function either.

Quoted 6 lines
Quoted 3 lines

I think vgic_check_ioaddr() called in kvm_vgic_addr() does the job (it checks the base @)

It seems that GICv2 suffers from the same problem. The cpu interface base is checked but the end can extend above IPA size. Note that the cpu interface is 8KBs and vgic_check_ioaddr() is only checking that its base

... except that the doc for KVM_VGIC_V2_ADDR_TYPE_CPU says that the CPU interface region is 4K, while the check in vgic_v2_check_base() is done against KVM_VGIC_V2_CPU_SIZE, which is 8K. I suppose that the CPU interface region is 8K because ARM IHI 0048B.b strongly recommends that the virtual CPU interface control registers are in a separate 4KB region, and KVM wants to emulate a GICv2 as close to the real thing as possible?

Quoted 5 lines

is 4KB aligned and below IPA size. The distributor region is 4KB so vgic_check_ioaddr() is enough in that case.

What about the following?

Quoted 7 lines

I can work on the next version of this patch (v2 has the GICv2 issue) which adds vgic_check_range(), which is like vgic_check_ioaddr() but with a size arg. kvm_vgic_addr() can then call vgic_check_range() and do all the checks for GICv2 and GICv3. Note that for GICv2, there's no need to wait until first vcpu run to do the check. Also note that I will have to keep the change in vgic_v3_check_base() to check for the old v3 redist API at first vcpu run.

Sounds good.

Thanks,

Alex

Quoted 21 lines

Thanks,
Ricardo

Quoted 17 lines

Thanks

Eric

Quoted 14 lines

Thanks,

Alex

Quoted 10 lines

Thanks,
Ricardo

Quoted 7 lines

Thanks,

Alex

Quoted 3 lines
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
 	}
 
 	if (IS_VGIC_ADDR_UNDEF(d->vgic_dist_base))

_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

#13 lore raw

Re: [PATCH 1/2] KVM: arm64: vgic: check redist region is not above the VM IPA size

Hi Alexandru, Eric,

Quoted 111 lines · Alexandru Elisei

Hi Eric, Ricardo,

Quoted 68 lines · Ricardo Koller

Hi Alexandru and Eric,

Quoted 65 lines · Eric Auger

Hi Alexandru,

Quoted 50 lines · Alexandru Elisei

Hi Ricardo,

Quoted 36 lines · Ricardo Koller
Quoted 35 lines · Alexandru Elisei

Hi Ricardo,

Quoted 20 lines · Ricardo Koller

Extend vgic_v3_check_base() to verify that the redistributor regions don't go above the VM-specified IPA size (phys_size). This can happen when using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST attribute with:

base + size > phys_size AND base < phys_size

vgic_v3_check_base() is used to check the redist regions bases when setting them (with the vcpus added so far) and when attempting the first vcpu-run.

---
 arch/arm64/kvm/vgic/vgic-v3.c | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
index 66004f61cd83..5afd9f6f68f6 100644

arch/arm64/kvm/vgic/vgic-v3.c bool vgic_v3_check_base(struct kvm *kvm) · line 518

Quoted 7 lines
arch/arm64/kvm/vgic/vgic-v3.c +4 -0
@@ -512,6 +512,10 @@ bool vgic_v3_check_base(struct kvm *kvm)
 		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) <
 			rdreg->base)
 			return false;
+
+		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) >
+			kvm_phys_size(kvm))
+			return false;

Looks to me like this same check (and the overflow one before it) is done when adding a new Redistributor region in kvm_vgic_addr() -> vgic_v3_set_redist_base() -> vgic_v3_alloc_redist_region() -> vgic_check_ioaddr(). As far as I can tell, kvm_vgic_addr() handles both ways of setting the Redistributor address.

Quoted 8 lines
Quoted 3 lines

Without this patch, did you manage to set a base address such that base + size > kvm_phys_size()?

Yes, with the KVM_VGIC_V3_ADDR_TYPE_REDIST legacy API. The easiest way to get to this situation is with the selftest in patch 2. I then tried an extra experiment: map the first redistributor, run the first vcpu, and access the redist from inside the guest. KVM didn't complain in any of these steps.

Yes, Eric pointed out that I was mistaken and there is no check being done for base + size > kvm_phys_size().

Quoted 8 lines

What I was trying to say is that this check is better done when the user creates a Redistributor region, not when a VCPU is first run. We have everything we need to make the check when a region is created, why wait until the VCPU is run?

For example, vgic_v3_insert_redist_region() is called each time the adds a new Redistributor region (via either of the two APIs), and already has a check for the upper limit overflowing (identical to the check in vgic_v3_check_base()). I would add the check against the maximum IPA size there.

you seem to refer to an old kernel as vgic_v3_insert_redist_region was renamed into� vgic_v3_alloc_redist_region in
e5a35635464b kvm: arm64: vgic-v3: Introduce vgic_v3_free_redist_region()

Quoted 8 lines
Quoted 3 lines

I think in case you use the old rdist API you do not know yet the size of the redist region at this point (count=0), hence Ricardo's choice to do the check latter.

Just wanted to add one more detail. vgic_v3_check_base() is also called when creating the redistributor region (via vgic_v3_set_redist_base -> vgic_register_redist_iodev). This patch reuses that check for the old redist API to also check for "base + size > kvm_phys_size()" with a size calculated using the vcpus added so far.

@Eric: Indeed I was looking at an older kernel by mistake, thank you for pointing that out!

Thank you both for the explanations, the piece I was missing was the fact that KVM_VGIC_V3_ADDR_TYPE_REDIST specifies only the base address and the limit for the region is the number of VCPUs * (KVM_VGIC_V3_REDIST_SIZE = 128K), which makes it necessary to have the check when each VCPU is first run (as far as I can tell, VCPUs can be created at any time).

Quoted 6 lines
Quoted 4 lines
Quoted 4 lines

Also, because vgic_v3_insert_redist_region() already checks for overflow, I believe the overflow check in vgic_v3_check_base() is redundant.

It's redundant for the new redist API, but still needed for the old redist API.

Indeed.

Quoted 13 lines
Quoted 13 lines
Quoted 10 lines

As far as I can tell, vgic_v3_check_base() is there to make sure that the Distributor doesn't overlap with any of the Redistributors, and because the Redistributors and the Distributor can be created in any order, we defer the check until the first VCPU is run. I might be wrong about this, someone please correct me if I'm wrong.

Also, did you verify that KVM is also doing this check for GICv2? KVM does something similar and calls vgic_v2_check_base() when mapping the GIC resources, and I don't see a check for the maximum IPA size in that function either.

I think vgic_check_ioaddr() called in kvm_vgic_addr() does the job (it checks the base @)

Quoted 7 lines
Quoted 3 lines

It seems that GICv2 suffers from the same problem. The cpu interface base is checked but the end can extend above IPA size. Note that the cpu interface is 8KBs and vgic_check_ioaddr() is only checking that its base

... except that the doc for KVM_VGIC_V2_ADDR_TYPE_CPU says that the CPU interface region is 4K, while the check in vgic_v2_check_base() is done against KVM_VGIC_V2_CPU_SIZE, which is 8K.

The "GIC virtual CPU interface" alone is slightly more than 4K: GICV_DIR is at 0x1000. The documentation might need to be updated.

Quoted 4 lines

I suppose that the CPU interface region is 8K
because ARM IHI 0048B.b strongly recommends that the virtual CPU interface control registers are in a separate 4KB region, and KVM wants to emulate a GICv2 as close to the real thing as possible?

Are the "virtual CPU interface control" registers the ones starting with GICH_? If yes, then I'm a bit confused, as those are not exposed to the guest (to my knowledge).

Quoted 14 lines
Quoted 6 lines

is 4KB aligned and below IPA size. The distributor region is 4KB so vgic_check_ioaddr() is enough in that case.

What about the following?

Quoted 7 lines

I can work on the next version of this patch (v2 has the GICv2 issue) which adds vgic_check_range(), which is like vgic_check_ioaddr() but with a size arg. kvm_vgic_addr() can then call vgic_check_range() and do all the checks for GICv2 and GICv3. Note that for GICv2, there's no need to wait until first vcpu run to do the check. Also note that I will have to keep the change in vgic_v3_check_base() to check for the old v3 redist API at first vcpu run.

Quoted 13 lines

Sounds good.

Thanks,

Alex

Quoted 7 lines

Thanks,
Ricardo

Quoted 3 lines

Thanks

Eric

Will do, thank you both.

Ricardo

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

Thanks,

Alex

Quoted 3 lines

Thanks,
Ricardo

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

Thanks,

Alex

Quoted 3 lines
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
 	}
 
 	if (IS_VGIC_ADDR_UNDEF(d->vgic_dist_base))

_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

#14 lore raw

Re: [PATCH 1/2] KVM: arm64: vgic: check redist region is not above the VM IPA size

Hi Ricardo,

(adding [email protected] to CC because the email this is a reply to got rejected because of html content)

Quoted 119 lines · Ricardo Koller

Hi Alexandru, Eric,

Quoted 107 lines · Alexandru Elisei

Hi Eric, Ricardo,

Quoted 68 lines · Ricardo Koller

Hi Alexandru and Eric,

Quoted 65 lines · Eric Auger

Hi Alexandru,

Quoted 50 lines · Alexandru Elisei

Hi Ricardo,

Quoted 36 lines · Ricardo Koller
Quoted 35 lines · Alexandru Elisei

Hi Ricardo,

Quoted 20 lines · Ricardo Koller

Extend vgic_v3_check_base() to verify that the redistributor regions don't go above the VM-specified IPA size (phys_size). This can happen when using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST attribute with:

base + size > phys_size AND base < phys_size

vgic_v3_check_base() is used to check the redist regions bases when setting them (with the vcpus added so far) and when attempting the first vcpu-run.

---
 arch/arm64/kvm/vgic/vgic-v3.c | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
index 66004f61cd83..5afd9f6f68f6 100644

arch/arm64/kvm/vgic/vgic-v3.c bool vgic_v3_check_base(struct kvm *kvm) · line 518

Quoted 7 lines
arch/arm64/kvm/vgic/vgic-v3.c +4 -0
@@ -512,6 +512,10 @@ bool vgic_v3_check_base(struct kvm *kvm)
 		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) <
 			rdreg->base)
 			return false;
+
+		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) >
+			kvm_phys_size(kvm))
+			return false;

Looks to me like this same check (and the overflow one before it) is done when adding a new Redistributor region in kvm_vgic_addr() -> vgic_v3_set_redist_base() -> vgic_v3_alloc_redist_region() -> vgic_check_ioaddr(). As far as I can tell, kvm_vgic_addr() handles both ways of setting the Redistributor address.

Quoted 8 lines
Quoted 3 lines

Without this patch, did you manage to set a base address such that base + size > kvm_phys_size()?

Yes, with the KVM_VGIC_V3_ADDR_TYPE_REDIST legacy API. The easiest way to get to this situation is with the selftest in patch 2. I then tried an extra experiment: map the first redistributor, run the first vcpu, and access the redist from inside the guest. KVM didn't complain in any of these steps.

Yes, Eric pointed out that I was mistaken and there is no check being done for base + size > kvm_phys_size().

Quoted 8 lines

What I was trying to say is that this check is better done when the user creates a Redistributor region, not when a VCPU is first run. We have everything we need to make the check when a region is created, why wait until the VCPU is run?

For example, vgic_v3_insert_redist_region() is called each time the adds a new Redistributor region (via either of the two APIs), and already has a check for the upper limit overflowing (identical to the check in vgic_v3_check_base()). I would add the check against the maximum IPA size there.

you seem to refer to an old kernel as vgic_v3_insert_redist_region was renamed into� vgic_v3_alloc_redist_region in
e5a35635464b kvm: arm64: vgic-v3: Introduce vgic_v3_free_redist_region()

Quoted 8 lines
Quoted 3 lines

I think in case you use the old rdist API you do not know yet the size of the redist region at this point (count=0), hence Ricardo's choice to do the check latter.

Just wanted to add one more detail. vgic_v3_check_base() is also called when creating the redistributor region (via vgic_v3_set_redist_base -> vgic_register_redist_iodev). This patch reuses that check for the old redist API to also check for "base + size > kvm_phys_size()" with a size calculated using the vcpus added so far.

@Eric: Indeed I was looking at an older kernel by mistake, thank you for pointing that out!

Thank you both for the explanations, the piece I was missing was the fact that KVM_VGIC_V3_ADDR_TYPE_REDIST specifies only the base address and the limit for the region is the number of VCPUs * (KVM_VGIC_V3_REDIST_SIZE = 128K), which makes it necessary to have the check when each VCPU is first run (as far as I can tell, VCPUs can be created at any time).

Quoted 5 lines
Quoted 3 lines
Quoted 3 lines

Also, because vgic_v3_insert_redist_region() already checks for overflow, I believe the overflow check in vgic_v3_check_base() is redundant.

It's redundant for the new redist API, but still needed for the old redist API.

Indeed.

Quoted 12 lines
Quoted 12 lines
Quoted 9 lines

As far as I can tell, vgic_v3_check_base() is there to make sure that the Distributor doesn't overlap with any of the Redistributors, and because the Redistributors and the Distributor can be created in any order, we defer the check until the first VCPU is run. I might be wrong about this, someone please correct me if I'm wrong.

Also, did you verify that KVM is also doing this check for GICv2? KVM does something similar and calls vgic_v2_check_base() when mapping the GIC resources, and I don't see a check for the maximum IPA size in that function either.

I think vgic_check_ioaddr() called in kvm_vgic_addr() does the job (it checks the base @)

Quoted 6 lines
Quoted 3 lines

It seems that GICv2 suffers from the same problem. The cpu interface base is checked but the end can extend above IPA size. Note that the cpu interface is 8KBs and vgic_check_ioaddr() is only checking that its base

... except that the doc for KVM_VGIC_V2_ADDR_TYPE_CPU says that the CPU interface region is 4K, while the check in vgic_v2_check_base() is done against KVM_VGIC_V2_CPU_SIZE, which is 8K.

The "GIC virtual CPU interface" alone is slightly more than 4K: GICV_DIR is at 0x1000. The documentation might need to be updated.

Quoted 7 lines
Quoted 4 lines

I suppose that the CPU interface region is 8K
because ARM IHI 0048B.b strongly recommends that the virtual CPU interface control registers are in a separate 4KB region, and KVM wants to emulate a GICv2 as close to the real thing as possible?

Are the "virtual CPU interface control" registers the ones starting with GICH_? If yes, then I'm a bit confused, as those are not exposed to the guest (to my knowledge).

Yes, those are the ones, and I also did find that they are not exposed to the guest.

Comparing the KVM documentation with what KVM actually does, I assumed that the 8KB was a forward looking decision, in case nested virtualization will support GICv2, which means that the GICH_* registers would also have to be exposed. Making the CPU interface for a CPU 8KB from the start would avoid changes or additions to the API if that happens.

However, after further digging through the spec, I found that the virtual CPU interface is specified to be 8KB (Table 5-10 of ARM IHI 0048B.b). I think that's the reason KVM treats it as 8KB.

Thanks,

Alex

Quoted 22 lines
Quoted 22 lines
Quoted 6 lines

is 4KB aligned and below IPA size. The distributor region is 4KB so vgic_check_ioaddr() is enough in that case.

What about the following?

Quoted 7 lines

I can work on the next version of this patch (v2 has the GICv2 issue) which adds vgic_check_range(), which is like vgic_check_ioaddr() but with a size arg. kvm_vgic_addr() can then call vgic_check_range() and do all the checks for GICv2 and GICv3. Note that for GICv2, there's no need to wait until first vcpu run to do the check. Also note that I will have to keep the change in vgic_v3_check_base() to check for the old v3 redist API at first vcpu run.

Sounds good.

Thanks,

Alex

Quoted 3 lines

Thanks,
Ricardo

Quoted 21 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

Thanks

Eric

Will do, thank you both.

Ricardo

Quoted 14 lines
Quoted 14 lines
Quoted 14 lines
Quoted 14 lines

Thanks,

Alex

Quoted 10 lines

Thanks,
Ricardo

Quoted 7 lines

Thanks,

Alex

Quoted 3 lines
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
 	}
 
 	if (IS_VGIC_ADDR_UNDEF(d->vgic_dist_base))

_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

#15 lore raw

Re: [PATCH 1/2] KVM: arm64: vgic: check redist region is not above the VM IPA size

Quoted 141 lines · Alexandru Elisei

Hi Ricardo,

(adding [email protected] to CC because the email this is a reply to got rejected because of html content)

Quoted 119 lines · Ricardo Koller

Hi Alexandru, Eric,

Quoted 107 lines · Alexandru Elisei

Hi Eric, Ricardo,

Quoted 68 lines · Ricardo Koller

Hi Alexandru and Eric,

Quoted 65 lines · Eric Auger

Hi Alexandru,

Quoted 50 lines · Alexandru Elisei

Hi Ricardo,

Quoted 36 lines · Ricardo Koller
Quoted 35 lines · Alexandru Elisei

Hi Ricardo,

Quoted 20 lines · Ricardo Koller

Extend vgic_v3_check_base() to verify that the redistributor regions don't go above the VM-specified IPA size (phys_size). This can happen when using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST attribute with:

base + size > phys_size AND base < phys_size

vgic_v3_check_base() is used to check the redist regions bases when setting them (with the vcpus added so far) and when attempting the first vcpu-run.

---
 arch/arm64/kvm/vgic/vgic-v3.c | 4 ++++
 1 file changed, 4 insertions(+)
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
index 66004f61cd83..5afd9f6f68f6 100644

arch/arm64/kvm/vgic/vgic-v3.c bool vgic_v3_check_base(struct kvm *kvm) · line 518

Quoted 7 lines
arch/arm64/kvm/vgic/vgic-v3.c +4 -0
@@ -512,6 +512,10 @@ bool vgic_v3_check_base(struct kvm *kvm)
 		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) <
 			rdreg->base)
 			return false;
+
+		if (rdreg->base + vgic_v3_rd_region_size(kvm, rdreg) >
+			kvm_phys_size(kvm))
+			return false;

Looks to me like this same check (and the overflow one before it) is done when adding a new Redistributor region in kvm_vgic_addr() -> vgic_v3_set_redist_base() -> vgic_v3_alloc_redist_region() -> vgic_check_ioaddr(). As far as I can tell, kvm_vgic_addr() handles both ways of setting the Redistributor address.

Quoted 8 lines
Quoted 3 lines

Without this patch, did you manage to set a base address such that base + size > kvm_phys_size()?

Yes, with the KVM_VGIC_V3_ADDR_TYPE_REDIST legacy API. The easiest way to get to this situation is with the selftest in patch 2. I then tried an extra experiment: map the first redistributor, run the first vcpu, and access the redist from inside the guest. KVM didn't complain in any of these steps.

Yes, Eric pointed out that I was mistaken and there is no check being done for base + size > kvm_phys_size().

Quoted 8 lines

What I was trying to say is that this check is better done when the user creates a Redistributor region, not when a VCPU is first run. We have everything we need to make the check when a region is created, why wait until the VCPU is run?

For example, vgic_v3_insert_redist_region() is called each time the adds a new Redistributor region (via either of the two APIs), and already has a check for the upper limit overflowing (identical to the check in vgic_v3_check_base()). I would add the check against the maximum IPA size there.

you seem to refer to an old kernel as vgic_v3_insert_redist_region was renamed into� vgic_v3_alloc_redist_region in
e5a35635464b kvm: arm64: vgic-v3: Introduce vgic_v3_free_redist_region()

Quoted 8 lines
Quoted 3 lines

I think in case you use the old rdist API you do not know yet the size of the redist region at this point (count=0), hence Ricardo's choice to do the check latter.

Just wanted to add one more detail. vgic_v3_check_base() is also called when creating the redistributor region (via vgic_v3_set_redist_base -> vgic_register_redist_iodev). This patch reuses that check for the old redist API to also check for "base + size > kvm_phys_size()" with a size calculated using the vcpus added so far.

@Eric: Indeed I was looking at an older kernel by mistake, thank you for pointing that out!

Thank you both for the explanations, the piece I was missing was the fact that KVM_VGIC_V3_ADDR_TYPE_REDIST specifies only the base address and the limit for the region is the number of VCPUs * (KVM_VGIC_V3_REDIST_SIZE = 128K), which makes it necessary to have the check when each VCPU is first run (as far as I can tell, VCPUs can be created at any time).

Quoted 5 lines
Quoted 3 lines
Quoted 3 lines

Also, because vgic_v3_insert_redist_region() already checks for overflow, I believe the overflow check in vgic_v3_check_base() is redundant.

It's redundant for the new redist API, but still needed for the old redist API.

Indeed.

Quoted 12 lines
Quoted 12 lines
Quoted 9 lines

As far as I can tell, vgic_v3_check_base() is there to make sure that the Distributor doesn't overlap with any of the Redistributors, and because the Redistributors and the Distributor can be created in any order, we defer the check until the first VCPU is run. I might be wrong about this, someone please correct me if I'm wrong.

Also, did you verify that KVM is also doing this check for GICv2? KVM does something similar and calls vgic_v2_check_base() when mapping the GIC resources, and I don't see a check for the maximum IPA size in that function either.

I think vgic_check_ioaddr() called in kvm_vgic_addr() does the job (it checks the base @)

Quoted 6 lines
Quoted 3 lines

It seems that GICv2 suffers from the same problem. The cpu interface base is checked but the end can extend above IPA size. Note that the cpu interface is 8KBs and vgic_check_ioaddr() is only checking that its base

... except that the doc for KVM_VGIC_V2_ADDR_TYPE_CPU says that the CPU interface region is 4K, while the check in vgic_v2_check_base() is done against KVM_VGIC_V2_CPU_SIZE, which is 8K.

The "GIC virtual CPU interface" alone is slightly more than 4K: GICV_DIR is at 0x1000. The documentation might need to be updated.

Quoted 7 lines
Quoted 4 lines

I suppose that the CPU interface region is 8K
because ARM IHI 0048B.b strongly recommends that the virtual CPU interface control registers are in a separate 4KB region, and KVM wants to emulate a GICv2 as close to the real thing as possible?

Are the "virtual CPU interface control" registers the ones starting with GICH_? If yes, then I'm a bit confused, as those are not exposed to the guest (to my knowledge).

Yes, those are the ones, and I also did find that they are not exposed to the guest.

Comparing the KVM documentation with what KVM actually does, I assumed that the 8KB was a forward looking decision, in case nested virtualization will support GICv2, which means that the GICH_* registers would also have to be exposed. Making the CPU interface for a CPU 8KB from the start would avoid changes or additions to the API if that happens.

Quoted 3 lines

However, after further digging through the spec, I found that the virtual CPU interface is specified to be 8KB (Table 5-10 of ARM IHI 0048B.b). I think that's the reason KVM treats it as 8KB.

Thanks for checking Alexandru. Will send a commit to update the documentation as well then (v3).

Regards,
Ricardo

Quoted 48 lines

Thanks,

Alex

Quoted 43 lines
Quoted 13 lines
Quoted 6 lines

is 4KB aligned and below IPA size. The distributor region is 4KB so vgic_check_ioaddr() is enough in that case.

What about the following?

Quoted 7 lines

I can work on the next version of this patch (v2 has the GICv2 issue) which adds vgic_check_range(), which is like vgic_check_ioaddr() but with a size arg. kvm_vgic_addr() can then call vgic_check_range() and do all the checks for GICv2 and GICv3. Note that for GICv2, there's no need to wait until first vcpu run to do the check. Also note that I will have to keep the change in vgic_v3_check_base() to check for the old v3 redist API at first vcpu run.

Quoted 12 lines

Sounds good.

Thanks,

Alex

Quoted 6 lines

Thanks,
Ricardo

Quoted 3 lines

Thanks

Eric

Will do, thank you both.

Ricardo

Quoted 14 lines
Quoted 14 lines
Quoted 14 lines
Quoted 14 lines

Thanks,

Alex

Quoted 10 lines

Thanks,
Ricardo

Quoted 7 lines

Thanks,

Alex

Quoted 3 lines
arch/arm64/kvm/vgic/vgic-v3.c +0 -0
 	}
 
 	if (IS_VGIC_ADDR_UNDEF(d->vgic_dist_base))

_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

#16 lore raw

PATCH 2/2 KVM: arm64: selftests: test for vgic redist above the VM IPA size

In reply to Ricardo Koller

This test attempts (and fails) to set a redistributor region using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST that's partially above the VM-specified IPA size.

---
 .../testing/selftests/kvm/aarch64/vgic_init.c | 44 +++++++++++++++++++
 1 file changed, 44 insertions(+)
tools/testing/selftests/kvm/aarch64/vgic_init.c +44 -0
index 623f31a14326..6dd7b5e91421 100644
@@ -285,6 +285,49 @@ static void test_vcpus_then_vgic(void)
 	vm_gic_destroy(&v);
 }
 
+static void test_redist_above_vm_pa_bits(enum vm_guest_mode mode)
+{
+	struct vm_gic v;
+	int ret, i;
+	uint32_t vcpuids[] = { 1, 2, 3, 4, };
+	int pa_bits = vm_guest_mode_params[mode].pa_bits;
+	uint64_t addr, psize = 1ULL << pa_bits;
+
+	/* Add vcpu 1 */
+	v.vm = vm_create_with_vcpus(mode, 1, DEFAULT_GUEST_PHY_PAGES,
+				    0, 0, guest_code, vcpuids);
+	v.gic_fd = kvm_create_device(v.vm, KVM_DEV_TYPE_ARM_VGIC_V3, false);
+
+	/* Set space for half a redist, we have 1 vcpu, so this fails. */
+	addr = psize - 0x10000;
+	ret = _kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+				 KVM_VGIC_V3_ADDR_TYPE_REDIST, &addr, true);
+	TEST_ASSERT(ret && errno == EINVAL, "not enough space for one redist");
+
+	/* Set space for 3 redists, we have 1 vcpu, so this succeeds. */
+	addr = psize - (3 * 2 * 0x10000);
+	kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+				 KVM_VGIC_V3_ADDR_TYPE_REDIST, &addr, true);
+
+	addr = 0x00000;
+	kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+			KVM_VGIC_V3_ADDR_TYPE_DIST, &addr, true);
+
+	/* Add three vcpus (2, 3, 4). */
+	for (i = 1; i < 4; ++i)
+		vm_vcpu_add_default(v.vm, vcpuids[i], guest_code);
+
+	kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_CTRL,
+			  KVM_DEV_ARM_VGIC_CTRL_INIT, NULL, true);
+
+	/* Attempt to run a vcpu without enough redist space. */
+	ret = run_vcpu(v.vm, vcpuids[3]);
+	TEST_ASSERT(ret && errno == EINVAL,
+			"redist base+size above IPA detected on 1st vcpu run");
+
+	vm_gic_destroy(&v);
+}
+
 static void test_new_redist_regions(void)
 {
 	void *dummy = NULL;
@@ -542,6 +585,7 @@ int main(int ac, char **av)
 	test_kvm_device();
 	test_vcpus_then_vgic();
 	test_vgic_then_vcpus();
+	test_redist_above_vm_pa_bits(VM_MODE_DEFAULT);
 	test_new_redist_regions();
 	test_typer_accesses();
 	test_last_bit_redist_regions();
Signature · 1 line
2.33.0.153.gba50c8fa24-goog
#17 lore raw

Re: [PATCH 2/2] KVM: arm64: selftests: test for vgic redist above the VM IPA size

Hi Ricardo,

Quoted 30 lines · Ricardo Koller

This test attempts (and fails) to set a redistributor region using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST that's partially above the VM-specified IPA size.

---
 .../testing/selftests/kvm/aarch64/vgic_init.c | 44 +++++++++++++++++++
 1 file changed, 44 insertions(+)
tools/testing/selftests/kvm/aarch64/vgic_init.c +13 -0
index 623f31a14326..6dd7b5e91421 100644
@@ -285,6 +285,49 @@ static void test_vcpus_then_vgic(void)
 	vm_gic_destroy(&v);
 }
 
+static void test_redist_above_vm_pa_bits(enum vm_guest_mode mode)
+{
+	struct vm_gic v;
+	int ret, i;
+	uint32_t vcpuids[] = { 1, 2, 3, 4, };
+	int pa_bits = vm_guest_mode_params[mode].pa_bits;
+	uint64_t addr, psize = 1ULL << pa_bits;
+
+	/* Add vcpu 1 */
+	v.vm = vm_create_with_vcpus(mode, 1, DEFAULT_GUEST_PHY_PAGES,
+				    0, 0, guest_code, vcpuids);
+	v.gic_fd = kvm_create_device(v.vm, KVM_DEV_TYPE_ARM_VGIC_V3, false);
+

tools/testing/selftests/kvm/aarch64/vgic_init.c static void test_vcpus_then_vgic(void) · line 310

Quoted 10 lines
tools/testing/selftests/kvm/aarch64/vgic_init.c +10 -0
+	/* Set space for half a redist, we have 1 vcpu, so this fails. */
+	addr = psize - 0x10000;
+	ret = _kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+				 KVM_VGIC_V3_ADDR_TYPE_REDIST, &addr, true);
+	TEST_ASSERT(ret && errno == EINVAL, "not enough space for one redist");
+
+	/* Set space for 3 redists, we have 1 vcpu, so this succeeds. */
+	addr = psize - (3 * 2 * 0x10000);
+	kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+				 KVM_VGIC_V3_ADDR_TYPE_REDIST, &addr, true);

I think you need to test both the old API (KVM_VGIC_V3_ADDR_TYPE_REDIST) and the new one (KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION).

Can't you integrate those new checks in existing tests, subtest_redist_regions() and subtest_dist_rdist() which already tests base addr beyond IPA limit (but not range end unfortunately). look for E2BIG.

Thanks

Eric

Quoted 24 lines
tools/testing/selftests/kvm/aarch64/vgic_init.c +20 -0
+
+	addr = 0x00000;
+	kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+			KVM_VGIC_V3_ADDR_TYPE_DIST, &addr, true);
+
+	/* Add three vcpus (2, 3, 4). */
+	for (i = 1; i < 4; ++i)
+		vm_vcpu_add_default(v.vm, vcpuids[i], guest_code);
+
+	kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_CTRL,
+			  KVM_DEV_ARM_VGIC_CTRL_INIT, NULL, true);
+
+	/* Attempt to run a vcpu without enough redist space. */
+	ret = run_vcpu(v.vm, vcpuids[3]);
+	TEST_ASSERT(ret && errno == EINVAL,
+			"redist base+size above IPA detected on 1st vcpu run");
+
+	vm_gic_destroy(&v);
+}
+
 static void test_new_redist_regions(void)
 {
 	void *dummy = NULL;

tools/testing/selftests/kvm/aarch64/vgic_init.c int main(int ac, char **av) · line 591

Quoted 7 lines
tools/testing/selftests/kvm/aarch64/vgic_init.c +1 -0
@@ -542,6 +585,7 @@ int main(int ac, char **av)
 	test_kvm_device();
 	test_vcpus_then_vgic();
 	test_vgic_then_vcpus();
+	test_redist_above_vm_pa_bits(VM_MODE_DEFAULT);
 	test_new_redist_regions();
 	test_typer_accesses();
 	test_last_bit_redist_regions();

_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

#18 lore raw

Re: [PATCH 2/2] KVM: arm64: selftests: test for vgic redist above the VM IPA size

Quoted 47 lines · Eric Auger

Hi Ricardo,

Quoted 30 lines · Ricardo Koller

This test attempts (and fails) to set a redistributor region using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST that's partially above the VM-specified IPA size.

---
 .../testing/selftests/kvm/aarch64/vgic_init.c | 44 +++++++++++++++++++
 1 file changed, 44 insertions(+)
tools/testing/selftests/kvm/aarch64/vgic_init.c +13 -0
index 623f31a14326..6dd7b5e91421 100644
@@ -285,6 +285,49 @@ static void test_vcpus_then_vgic(void)
 	vm_gic_destroy(&v);
 }
 
+static void test_redist_above_vm_pa_bits(enum vm_guest_mode mode)
+{
+	struct vm_gic v;
+	int ret, i;
+	uint32_t vcpuids[] = { 1, 2, 3, 4, };
+	int pa_bits = vm_guest_mode_params[mode].pa_bits;
+	uint64_t addr, psize = 1ULL << pa_bits;
+
+	/* Add vcpu 1 */
+	v.vm = vm_create_with_vcpus(mode, 1, DEFAULT_GUEST_PHY_PAGES,
+				    0, 0, guest_code, vcpuids);
+	v.gic_fd = kvm_create_device(v.vm, KVM_DEV_TYPE_ARM_VGIC_V3, false);
+

tools/testing/selftests/kvm/aarch64/vgic_init.c static void test_vcpus_then_vgic(void) · line 310

Quoted 10 lines
tools/testing/selftests/kvm/aarch64/vgic_init.c +10 -0
+	/* Set space for half a redist, we have 1 vcpu, so this fails. */
+	addr = psize - 0x10000;
+	ret = _kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+				 KVM_VGIC_V3_ADDR_TYPE_REDIST, &addr, true);
+	TEST_ASSERT(ret && errno == EINVAL, "not enough space for one redist");
+
+	/* Set space for 3 redists, we have 1 vcpu, so this succeeds. */
+	addr = psize - (3 * 2 * 0x10000);
+	kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+				 KVM_VGIC_V3_ADDR_TYPE_REDIST, &addr, true);

I think you need to test both the old API (KVM_VGIC_V3_ADDR_TYPE_REDIST) and the new one (KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION).

Quoted 5 lines

Can't you integrate those new checks in existing tests, subtest_redist_regions() and subtest_dist_rdist() which already tests base addr beyond IPA limit (but not range end unfortunately). look for E2BIG.

Had some issues adapting subtest_dist_rdist() as the IPA range check for ADDR_TYPE_REDIST is done at 1st vcpu run. subtest_dist_rdist() is already used to set overlapping dist/redist regions, which is then checked to generate EINVAL on 1st vcpu run. If subtest_dist_rdist() is also used to set the redist region above phys_size, then there won't be a way of checking that the vcpu run fails because of both the overlap and IPA issue. It was simpler and cleaner to just add a new function for the ADDR_TYPE_REDIST IPA range test. Will adapt subtest_redist_regions() as the check for ADDR_TYPE_REDIST_REGION can be done when setting the regions.

Related Question:

Both the KVM_VGIC_V3_ADDR_TYPE_REDIST and KVM_RUN currently return EINVAL with my proposed change (not E2BIG). I will change KVM_VGIC_V3_ADDR_TYPE_REDIST to fail with E2BIG, but will leave KVM_RUN failing with EINVAL. Would you say that's the correct behavior?

Thanks,
Ricardo

Quoted 26 lines

Thanks

Eric

Quoted 23 lines
tools/testing/selftests/kvm/aarch64/vgic_init.c +20 -0
+
+	addr = 0x00000;
+	kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+			KVM_VGIC_V3_ADDR_TYPE_DIST, &addr, true);
+
+	/* Add three vcpus (2, 3, 4). */
+	for (i = 1; i < 4; ++i)
+		vm_vcpu_add_default(v.vm, vcpuids[i], guest_code);
+
+	kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_CTRL,
+			  KVM_DEV_ARM_VGIC_CTRL_INIT, NULL, true);
+
+	/* Attempt to run a vcpu without enough redist space. */
+	ret = run_vcpu(v.vm, vcpuids[3]);
+	TEST_ASSERT(ret && errno == EINVAL,
+			"redist base+size above IPA detected on 1st vcpu run");
+
+	vm_gic_destroy(&v);
+}
+
 static void test_new_redist_regions(void)
 {
 	void *dummy = NULL;
Quoted 9 lines

tools/testing/selftests/kvm/aarch64/vgic_init.c int main(int ac, char **av) · line 591

Quoted 8 lines
tools/testing/selftests/kvm/aarch64/vgic_init.c +1 -0
@@ -542,6 +585,7 @@ int main(int ac, char **av)
 	test_kvm_device();
 	test_vcpus_then_vgic();
 	test_vgic_then_vcpus();
+	test_redist_above_vm_pa_bits(VM_MODE_DEFAULT);
 	test_new_redist_regions();
 	test_typer_accesses();
 	test_last_bit_redist_regions();

_______________________________________________
kvmarm mailing list
[email protected]
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm

#19 lore raw

Re: [PATCH 2/2] KVM: arm64: selftests: test for vgic redist above the VM IPA size

Hi Ricardo,

Quoted 52 lines · Ricardo Koller
Quoted 46 lines · Eric Auger

Hi Ricardo,

Quoted 30 lines · Ricardo Koller

This test attempts (and fails) to set a redistributor region using the legacy KVM_VGIC_V3_ADDR_TYPE_REDIST that's partially above the VM-specified IPA size.

---
 .../testing/selftests/kvm/aarch64/vgic_init.c | 44 +++++++++++++++++++
 1 file changed, 44 insertions(+)
tools/testing/selftests/kvm/aarch64/vgic_init.c +13 -0
index 623f31a14326..6dd7b5e91421 100644
@@ -285,6 +285,49 @@ static void test_vcpus_then_vgic(void)
 	vm_gic_destroy(&v);
 }
 
+static void test_redist_above_vm_pa_bits(enum vm_guest_mode mode)
+{
+	struct vm_gic v;
+	int ret, i;
+	uint32_t vcpuids[] = { 1, 2, 3, 4, };
+	int pa_bits = vm_guest_mode_params[mode].pa_bits;
+	uint64_t addr, psize = 1ULL << pa_bits;
+
+	/* Add vcpu 1 */
+	v.vm = vm_create_with_vcpus(mode, 1, DEFAULT_GUEST_PHY_PAGES,
+				    0, 0, guest_code, vcpuids);
+	v.gic_fd = kvm_create_device(v.vm, KVM_DEV_TYPE_ARM_VGIC_V3, false);
+

tools/testing/selftests/kvm/aarch64/vgic_init.c static void test_vcpus_then_vgic(void) · line 310

Quoted 10 lines
tools/testing/selftests/kvm/aarch64/vgic_init.c +10 -0
+	/* Set space for half a redist, we have 1 vcpu, so this fails. */
+	addr = psize - 0x10000;
+	ret = _kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+				 KVM_VGIC_V3_ADDR_TYPE_REDIST, &addr, true);
+	TEST_ASSERT(ret && errno == EINVAL, "not enough space for one redist");
+
+	/* Set space for 3 redists, we have 1 vcpu, so this succeeds. */
+	addr = psize - (3 * 2 * 0x10000);
+	kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+				 KVM_VGIC_V3_ADDR_TYPE_REDIST, &addr, true);

I think you need to test both the old API (KVM_VGIC_V3_ADDR_TYPE_REDIST) and the new one (KVM_VGIC_V3_ADDR_TYPE_REDIST_REGION).

Quoted 5 lines

Can't you integrate those new checks in existing tests, subtest_redist_regions() and subtest_dist_rdist() which already tests base addr beyond IPA limit (but not range end unfortunately). look for E2BIG.

Quoted 8 lines

Had some issues adapting subtest_dist_rdist() as the IPA range check for ADDR_TYPE_REDIST is done at 1st vcpu run. subtest_dist_rdist() is already used to set overlapping dist/redist regions, which is then checked to generate EINVAL on 1st vcpu run. If subtest_dist_rdist() is also used to set the redist region above phys_size, then there won't be a way of checking that the vcpu run fails because of both the overlap and IPA issue. It was simpler and cleaner to just add a new function for the ADDR_TYPE_REDIST IPA range test. Will adapt

OK I see, then effectively adding a new test was more straightforward.

Quoted 2 lines

subtest_redist_regions() as the check for ADDR_TYPE_REDIST_REGION can be done when setting the regions.

OK

Quoted 7 lines

Related Question:

Both the KVM_VGIC_V3_ADDR_TYPE_REDIST and KVM_RUN currently return EINVAL with my proposed change (not E2BIG). I will change KVM_VGIC_V3_ADDR_TYPE_REDIST to fail with E2BIG, but will leave KVM_RUN failing with EINVAL. Would you say that's the correct behavior?

This looks OK to me, as long as the KVM uapi doc documents is aligned.

Thanks

Eric

Quoted 38 lines

Thanks,
Ricardo

Quoted 34 lines

Thanks

Eric

Quoted 31 lines
tools/testing/selftests/kvm/aarch64/vgic_init.c +21 -0
+
+	addr = 0x00000;
+	kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_ADDR,
+			KVM_VGIC_V3_ADDR_TYPE_DIST, &addr, true);
+
+	/* Add three vcpus (2, 3, 4). */
+	for (i = 1; i < 4; ++i)
+		vm_vcpu_add_default(v.vm, vcpuids[i], guest_code);
+
+	kvm_device_access(v.gic_fd, KVM_DEV_ARM_VGIC_GRP_CTRL,
+			  KVM_DEV_ARM_VGIC_CTRL_INIT, NULL, true);
+
+	/* Attempt to run a vcpu without enough redist space. */
+	ret = run_vcpu(v.vm, vcpuids[3]);
+	TEST_ASSERT(ret && errno == EINVAL,
+			"redist base+size above IPA detected on 1st vcpu run");
+
+	vm_gic_destroy(&v);
+}
+
 static void test_new_redist_regions(void)
 {
 	void *dummy = NULL;
@@ -542,6 +585,7 @@ int main(int ac, char **av)
 	test_kvm_device();
 	test_vcpus_then_vgic();
 	test_vgic_then_vcpus();
+	test_redist_above_vm_pa_bits(VM_MODE_DEFAULT);
 	test_new_redist_regions();
 	test_typer_accesses();
 	test_last_bit_redist_regions();