Skip to contentSkip to conversation

Use new wrappers to copy userspace arrays

PATCH Applied Philipp Stanner 4 participants 8 messages Nov 2, 2023 – Dec 1, 2023

Subsystems (5)
x86 KVM s390 KVM x86 +1 more subsystems
lore
Share & tools
lore
#0 lore raw

cover Use new wrappers to copy userspace arrays

Hi,

Linus recently merged [1] the wrapper functions memdup_array_user() and vmemdup_array_user() in include/linux/string.h for Kernel v6.7

I am currently adding them to all places where (v)memdup_user() had been used to copy arrays.

The wrapper is different to the wrapped functions only in that it might return -EOVERFLOW. So this new error code might get pushed up to userspace. I hope this is fine.

I felt that it might be a good idea to land those three patches here with a single series, since they all touch something KVM-related.

Kind regards,
P.

[1] https://lore.kernel.org/all/[email protected]/

Philipp Stanner (3):
arch/x86/kvm: copy user-array with overflow-check arch/s390/kvm: copy userspace-array safely virt/kvm: copy userspace-array safely

 arch/s390/kvm/guestdbg.c | 4 ++--
 arch/x86/kvm/cpuid.c     | 4 ++--
 virt/kvm/kvm_main.c      | 5 ++---
 3 files changed, 6 insertions(+), 7 deletions(-)
Signature · 1 line
2.41.0
#1 lore raw

PATCH 1/3 arch/x86/kvm: copy user-array with overflow-check

cpuid.c utilizes vmemdup_user() and array_size() to copy two userspace arrays. This, currently, does not check for an overflow.

Use the new wrapper vmemdup_array_user() to copy the arrays more safely.

---
 arch/x86/kvm/cpuid.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
arch/x86/kvm/cpuid.c +2 -2
index 773132c3bf5a..4a15b2a20f84 100644
@@ -467,7 +467,7 @@ int kvm_vcpu_ioctl_set_cpuid(struct kvm_vcpu *vcpu,
 		return -E2BIG;
 
 	if (cpuid->nent) {
-		e = vmemdup_user(entries, array_size(sizeof(*e), cpuid->nent));
+		e = vmemdup_array_user(entries, cpuid->nent, sizeof(*e));
 		if (IS_ERR(e))
 			return PTR_ERR(e);
 
@@ -511,7 +511,7 @@ int kvm_vcpu_ioctl_set_cpuid2(struct kvm_vcpu *vcpu,
 		return -E2BIG;
 
 	if (cpuid->nent) {
-		e2 = vmemdup_user(entries, array_size(sizeof(*e2), cpuid->nent));
+		e2 = vmemdup_array_user(entries, cpuid->nent, sizeof(*e2));
 		if (IS_ERR(e2))
 			return PTR_ERR(e2);
 	}
Signature · 1 line
2.41.0
#2 lore raw

PATCH 2/3 arch/s390/kvm: copy userspace-array safely

In reply to Philipp Stanner

guestdbg.c utilizes memdup_user() to copy a userspace array. This, currently, does not check for an overflow.

Use the new wrapper memdup_array_user() to copy the array more safely.

---
 arch/s390/kvm/guestdbg.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
arch/s390/kvm/guestdbg.c +2 -2
index 3765c4223bf9..80879fc73c90 100644
@@ -213,8 +213,8 @@ int kvm_s390_import_bp_data(struct kvm_vcpu *vcpu,
 	else if (dbg->arch.nr_hw_bp > MAX_BP_COUNT)
 		return -EINVAL;
 
-	bp_data = memdup_user(dbg->arch.hw_bp,
-			      sizeof(*bp_data) * dbg->arch.nr_hw_bp);
+	bp_data = memdup_array_user(dbg->arch.hw_bp, dbg->arch.nr_hw_bp,
+				    sizeof(*bp_data));
 	if (IS_ERR(bp_data))
 		return PTR_ERR(bp_data);
 
Signature · 1 line
2.41.0
maintainer Acked-by on v1 2/3
#3 lore raw

Re: [PATCH 2/3] arch/s390/kvm: copy userspace-array safely

Quoted 7 lines · On Thu, 2 Nov 2023 19:15:25 +0100 Philipp Stanner <[email protected]> wrote:

guestdbg.c utilizes memdup_user() to copy a userspace array. This, currently, does not check for an overflow.

Use the new wrapper memdup_array_user() to copy the array more safely.

Quoted 19 lines
---
 arch/s390/kvm/guestdbg.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
arch/s390/kvm/guestdbg.c +2 -2
index 3765c4223bf9..80879fc73c90 100644
@@ -213,8 +213,8 @@ int kvm_s390_import_bp_data(struct kvm_vcpu *vcpu,
 	else if (dbg->arch.nr_hw_bp > MAX_BP_COUNT)
 		return -EINVAL;
 
-	bp_data = memdup_user(dbg->arch.hw_bp,
-			      sizeof(*bp_data) * dbg->arch.nr_hw_bp);
+	bp_data = memdup_array_user(dbg->arch.hw_bp, dbg->arch.nr_hw_bp,
+				    sizeof(*bp_data));
 	if (IS_ERR(bp_data))
 		return PTR_ERR(bp_data);
 
#4 lore raw

PATCH 3/3 virt/kvm: copy userspace-array safely

In reply to Philipp Stanner

kvm_main.c utilizes vmemdup_user() and array_size() to copy a userspace array. Currently, this does not check for an overflow.

Use the new wrapper vmemdup_array_user() to copy the array more safely.

---
 virt/kvm/kvm_main.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)
virt/kvm/kvm_main.c +2 -3
index 486800a7024b..2a2f409c2a7d 100644
@@ -4932,9 +4932,8 @@ static long kvm_vm_ioctl(struct file *filp,
 			goto out;
 		if (routing.nr) {
 			urouting = argp;
-			entries = vmemdup_user(urouting->entries,
-					       array_size(sizeof(*entries),
-							  routing.nr));
+			entries = vmemdup_array_user(urouting->entries,
+						     routing.nr, sizeof(*entries));
 			if (IS_ERR(entries)) {
 				r = PTR_ERR(entries);
 				goto out;
Signature · 1 line
2.41.0
maintainer Applied
#5 lore raw
In reply to Philipp Stanner
Quoted 11 lines · Philipp Stanner

Linus recently merged [1] the wrapper functions memdup_array_user() and vmemdup_array_user() in include/linux/string.h for Kernel v6.7

I am currently adding them to all places where (v)memdup_user() had been used to copy arrays.

The wrapper is different to the wrapped functions only in that it might return -EOVERFLOW. So this new error code might get pushed up to userspace. I hope this is fine.

[...]

Applied to kvm-x86 generic. Claudio (or anyone else from s390), holler if you want to take the s390 patch through the s390 tree.

I massaged the shortlogs to use KVM's standard scopes, and to make it clear that these are hardening patches, i.e. that there is no unsafe/buggy behavior that is being fixed. I also added a note at the end of each changelog to call out that KVM pre-checks the sizes before copying, again to make it clear that using the safer helper isn't expected to actually change KVM's behavior.

[1/3] KVM: x86: Harden copying of userspace-array against overflow

      https://github.com/kvm-x86/linux/commit/573cc0e5cf14
[2/3] KVM: s390: Harden copying of userspace-array against overflow
      https://github.com/kvm-x86/linux/commit/8b81a8d7c6b7
[3/3] KVM: Harden copying of userspace-array against overflow
      https://github.com/kvm-x86/linux/commit/bc2cad56094c

--
https://github.com/kvm-x86/linux/tree/next

maintainer Acked-by
#6 lore raw

Am 01.12.23 um 02:52 schrieb Sean Christopherson:

Quoted 7 lines · Philipp Stanner
Quoted 6 lines

Linus recently merged [1] the wrapper functions memdup_array_user() and vmemdup_array_user() in include/linux/string.h for Kernel v6.7

I am currently adding them to all places where (v)memdup_user() had been used to copy arrays.

Quoted 8 lines
Quoted 5 lines

The wrapper is different to the wrapped functions only in that it might return -EOVERFLOW. So this new error code might get pushed up to userspace. I hope this is fine.

[...]

Applied to kvm-x86 generic. Claudio (or anyone else from s390), holler if you want to take the s390 patch through the s390 tree.

I think this is fine via your tree.

Feel free to add

to patch 2 if the commit id is not yet final.

#7 lore raw
Quoted 16 lines · Christian Borntraeger

Am 01.12.23 um 02:52 schrieb Sean Christopherson:

Quoted 13 lines · Philipp Stanner
Quoted 11 lines

Linus recently merged [1] the wrapper functions memdup_array_user() and vmemdup_array_user() in include/linux/string.h for Kernel v6.7

I am currently adding them to all places where (v)memdup_user() had been used to copy arrays.

The wrapper is different to the wrapped functions only in that it might return -EOVERFLOW. So this new error code might get pushed up to userspace. I hope this is fine.

[...]

Quoted 8 lines
Quoted 2 lines

Applied to kvm-x86 generic. Claudio (or anyone else from s390), holler if you want to take the s390 patch through the s390 tree.

I think this is fine via your tree.

Feel free to add

to patch 2 if the commit id is not yet final.

Done, thanks much! New hashes if anyone cares:

[1/3] KVM: x86: Harden copying of userspace-array against overflow

      https://github.com/kvm-x86/linux/commit/573cc0e5cf14
[2/3] KVM: s390: Harden copying of userspace-array against overflow
      https://github.com/kvm-x86/linux/commit/8c4976772d9b
[3/3] KVM: Harden copying of userspace-array against overflow
      https://github.com/kvm-x86/linux/commit/1f829359c8c3