[PATCH] KVM/arm64: vgic-its: Fix memory leak when vgic_its_set_abi() fails

[PATCH] KVM/arm64: vgic-its: Fix memory leak when vgic_its_set_abi() fails

[Jackie Liu]

From: Jackie Liu <liuyun01@kylinos.cn>

In vgic_its_create(), if vgic_its_set_abi() fails after allocating the
its structure and setting kvm state, the allocated 'its' is leaked
because the function returns without freeing it.

Fix by rolling back the kvm state flags and freeing the its structure
when vgic_its_set_abi() returns an error.

Fixes: 71afe470e20d ("KVM: arm64: vgic-its: Introduce migration ABI infrastructure")
Signed-off-by: Jackie Liu <liuyun01@kylinos.cn>
---
 arch/arm64/kvm/vgic/vgic-its.c | 18 +++++++++++-------
 1 file changed, 11 insertions(+), 7 deletions(-)

diff --git a/arch/arm64/kvm/vgic/vgic-its.c b/arch/arm64/kvm/vgic/vgic-its.c
index 1d7e5d560af4..83718eab4e06 100644
--- a/arch/arm64/kvm/vgic/vgic-its.c
+++ b/arch/arm64/kvm/vgic/vgic-its.c
@@ -1878,8 +1878,6 @@ static int vgic_its_create(struct kvm_device *dev, u32 type)
 	INIT_LIST_HEAD(&its->collection_list);
 	xa_init(&its->translation_cache);
 
-	dev->kvm->arch.vgic.msis_require_devid = true;
-	dev->kvm->arch.vgic.has_its = true;
 	its->enabled = false;
 	its->dev = dev;
 
@@ -1887,15 +1885,21 @@ static int vgic_its_create(struct kvm_device *dev, u32 type)
 		((u64)GITS_BASER_TYPE_DEVICE << GITS_BASER_TYPE_SHIFT);
 	its->baser_coll_table = INITIAL_BASER_VALUE |
 		((u64)GITS_BASER_TYPE_COLLECTION << GITS_BASER_TYPE_SHIFT);
-	dev->kvm->arch.vgic.propbaser = INITIAL_PROPBASER_VALUE;
-
-	dev->private = its;
 
 	ret = vgic_its_set_abi(its, NR_ITS_ABIS - 1);
+	if (ret) {
+		mutex_unlock(&dev->kvm->arch.config_lock);
+		kfree(its);
+		return ret;
+	}
 
-	mutex_unlock(&dev->kvm->arch.config_lock);
+	dev->kvm->arch.vgic.msis_require_devid = true;
+	dev->kvm->arch.vgic.has_its = true;
+	dev->kvm->arch.vgic.propbaser = INITIAL_PROPBASER_VALUE;
+	dev->private = its;
 
-	return ret;
+	mutex_unlock(&dev->kvm->arch.config_lock);
+	return 0;
 }
 
 static void vgic_its_destroy(struct kvm_device *kvm_dev)
-- 
2.54.0

Re: [PATCH] KVM/arm64: vgic-its: Fix memory leak when vgic_its_set_abi() fails

[Oliver Upton]

On Thu, Jun 04, 2026 at 11:14:26AM +0800, Jackie Liu wrote:
> From: Jackie Liu <liuyun01@kylinos.cn>
> 
> In vgic_its_create(), if vgic_its_set_abi() fails after allocating the
> its structure and setting kvm state, the allocated 'its' is leaked
> because the function returns without freeing it.
> 
> Fix by rolling back the kvm state flags and freeing the its structure
> when vgic_its_set_abi() returns an error.
> 
> Fixes: 71afe470e20d ("KVM: arm64: vgic-its: Introduce migration ABI infrastructure")
> Signed-off-by: Jackie Liu <liuyun01@kylinos.cn>

vgic_its_set_abi() always succeeds so there's no actual memory leak
here. I'd prefer getting rid of the return value instead.

Thanks,
Oliver

Re: [PATCH] KVM/arm64: vgic-its: Fix memory leak when vgic_its_set_abi() fails

[Jackie Liu]

2026年6月4日 14:41, "Oliver Upton" <oupton@kernel.org mailto:oupton@kernel.org?to=%22Oliver%20Upton%22%20%3Coupton%40kernel.org%3E > 写到:


> 
> On Thu, Jun 04, 2026 at 11:14:26AM +0800, Jackie Liu wrote:
> 
> > 
> > From: Jackie Liu <liuyun01@kylinos.cn>
> >  
> >  In vgic_its_create(), if vgic_its_set_abi() fails after allocating the
> >  its structure and setting kvm state, the allocated 'its' is leaked
> >  because the function returns without freeing it.
> >  
> >  Fix by rolling back the kvm state flags and freeing the its structure
> >  when vgic_its_set_abi() returns an error.
> >  
> >  Fixes: 71afe470e20d ("KVM: arm64: vgic-its: Introduce migration ABI infrastructure")
> >  Signed-off-by: Jackie Liu <liuyun01@kylinos.cn>
> > 
> vgic_its_set_abi() always succeeds so there's no actual memory leak
> here. I'd prefer getting rid of the return value instead.

That was my original thought as well. However, I kept the defensive cleanup logic in place
to make future extensions easier. This way, if additional resource allocation paths are 
introduced later, we're less likely to miss the corresponding cleanup.

That said, looking at the current code, the return value no longer serves any real 
purpose, so removing it would also be a reasonable and cleaner approach.

I'll send a V2 with this change.

Thanks.
Jackie


> 
> Thanks,
> Oliver
>