Skip to contentSkip to conversation

KVM: PPC: Improve KVM reference counting

PATCH Applied Alexey Kardashevskiy 3 participants 4 messages Feb 21, 2019 – Feb 22, 2019

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 KVM: PPC: Improve KVM reference counting

The anon fd's ops releases the KVM reference in the release hook. However we reference the KVM object after we create the fd so there is small window when the release function can be called and dereferenced the KVM object which potentially may free it.

It is not a problem at the moment as the file is created and KVM is referenced under the KVM lock and the release function obtains the same lock before dereferencing the KVM (although the lock is not held when calling kvm_put_kvm()) but it is a fragile against future changes.

This references the KVM object before creating a file.

---

The original bug is described here:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?idÏa393811
But in this case kvm_put_kvm() is called straight away with no locks before/after/around.

---
 arch/powerpc/kvm/book3s_64_vio.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
arch/powerpc/kvm/book3s_64_vio.c +4 -3
index 532ab797..d68c969 100644
@@ -338,14 +338,15 @@ long kvm_vm_ioctl_create_spapr_tce(struct kvm *kvm,
 		}
 	}
 
+	kvm_get_kvm(kvm);
 	if (!ret)
 		ret = anon_inode_getfd("kvm-spapr-tce", &kvm_spapr_tce_fops,
 				       stt, O_RDWR | O_CLOEXEC);
 
-	if (ret >= 0) {
+	if (ret >= 0)
 		list_add_rcu(&stt->list, &kvm->arch.spapr_tce_tables);
-		kvm_get_kvm(kvm);
-	}
+	else
+		kvm_put_kvm(kvm);
 
 	mutex_unlock(&kvm->lock);
 
Signature · 1 line
2.17.1
#1 lore raw
Quoted 4 lines · Alexey Kardashevskiy

The anon fd's ops releases the KVM reference in the release hook. However we reference the KVM object after we create the fd so there is small window when the release function can be called and dereferenced the KVM object which potentially may free it.

dereference

Quoted 31 lines

It is not a problem at the moment as the file is created and KVM is referenced under the KVM lock and the release function obtains the same lock before dereferencing the KVM (although the lock is not held when calling kvm_put_kvm()) but it is a fragile against future changes.

This references the KVM object before creating a file.

---

The original bug is described here:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?idÏa393811
But in this case kvm_put_kvm() is called straight away with no locks before/after/around.

---
 arch/powerpc/kvm/book3s_64_vio.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
arch/powerpc/kvm/book3s_64_vio.c +1 -0
index 532ab797..d68c969 100644
@@ -338,14 +338,15 @@ long kvm_vm_ioctl_create_spapr_tce(struct kvm *kvm,
 		}
 	}
 
+	kvm_get_kvm(kvm);
 	if (!ret)
 		ret = anon_inode_getfd("kvm-spapr-tce", &kvm_spapr_tce_fops,
 				       stt, O_RDWR | O_CLOEXEC);

arch/powerpc/kvm/book3s_64_vio.c long kvm_vm_ioctl_create_spapr_tce(struct kvm *kvm, · line 351

Quoted 10 lines
arch/powerpc/kvm/book3s_64_vio.c +3 -3
 
-	if (ret >= 0) {
+	if (ret >= 0)
 		list_add_rcu(&stt->list, &kvm->arch.spapr_tce_tables);
-		kvm_get_kvm(kvm);
-	}
+	else
+		kvm_put_kvm(kvm);
 
 	mutex_unlock(&kvm->lock);

This looks correct to me. But I feel like the logic could be cleaner, perhaps like this (patch below):

	mutex_lock(&kvm->lock);

	/* Check this LIOBN hasn't been previously allocated */
	list_for_each_entry(siter, &kvm->arch.spapr_tce_tables, list) {
		if (siter->liobn = args->liobn) {
			ret = -EBUSY;
			goto fail_unlock;
		}
	}

	kvm_get_kvm(kvm);
	ret = anon_inode_getfd("kvm-spapr-tce", &kvm_spapr_tce_fops,
				stt, O_RDWR | O_CLOEXEC);

	if (ret < 0) {
		kvm_put_kvm(kvm);
		goto fail_unlock;
	}

	list_add_rcu(&stt->list, &kvm->arch.spapr_tce_tables);

	mutex_unlock(&kvm->lock);

	return ret;

fail_unlock:

	mutex_unlock(&kvm->lock);

fail:

cheers

arch/powerpc/kvm/book3s_64_vio.c +13 -11
index 532ab79734c7..5d74602db0d0 100644
@@ -296,7 +296,7 @@ long kvm_vm_ioctl_create_spapr_tce(struct kvm *kvm,
 	struct kvmppc_spapr_tce_table *stt = NULL;
 	struct kvmppc_spapr_tce_table *siter;
 	unsigned long npages, size = args->size;
-	int ret = -ENOMEM;
+	int ret;
 	int i;
 
 	if (!args->size || args->page_shift < 12 || args->page_shift > 34 ||
@@ -330,28 +330,30 @@ long kvm_vm_ioctl_create_spapr_tce(struct kvm *kvm,
 	mutex_lock(&kvm->lock);
 
 	/* Check this LIOBN hasn't been previously allocated */
-	ret = 0;
 	list_for_each_entry(siter, &kvm->arch.spapr_tce_tables, list) {
 		if (siter->liobn = args->liobn) {
 			ret = -EBUSY;
-			break;
+			goto fail_unlock;
 		}
 	}
 
-	if (!ret)
-		ret = anon_inode_getfd("kvm-spapr-tce", &kvm_spapr_tce_fops,
-				       stt, O_RDWR | O_CLOEXEC);
+	kvm_get_kvm(kvm);
+	ret = anon_inode_getfd("kvm-spapr-tce", &kvm_spapr_tce_fops,
+				stt, O_RDWR | O_CLOEXEC);
 
-	if (ret >= 0) {
-		list_add_rcu(&stt->list, &kvm->arch.spapr_tce_tables);
-		kvm_get_kvm(kvm);
+	if (ret < 0) {
+		kvm_put_kvm(kvm);
+		goto fail_unlock;
 	}
 
+	list_add_rcu(&stt->list, &kvm->arch.spapr_tce_tables);
+
 	mutex_unlock(&kvm->lock);
 
-	if (ret >= 0)
-		return ret;
+	return ret;
 
+ fail_unlock:
+	mutex_unlock(&kvm->lock);
  fail:
 	for (i = 0; i < npages; i++)
 		if (stt->pages[i])
#2 lore raw
Quoted 47 lines · Michael Ellerman
Quoted 4 lines · Alexey Kardashevskiy

The anon fd's ops releases the KVM reference in the release hook. However we reference the KVM object after we create the fd so there is small window when the release function can be called and dereferenced the KVM object which potentially may free it.

dereference

Quoted 40 lines

It is not a problem at the moment as the file is created and KVM is referenced under the KVM lock and the release function obtains the same lock before dereferencing the KVM (although the lock is not held when calling kvm_put_kvm()) but it is a fragile against future changes.

This references the KVM object before creating a file.

---

The original bug is described here:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?idÏa393811
But in this case kvm_put_kvm() is called straight away with no locks before/after/around.

---
 arch/powerpc/kvm/book3s_64_vio.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
arch/powerpc/kvm/book3s_64_vio.c +4 -3
index 532ab797..d68c969 100644
@@ -338,14 +338,15 @@ long kvm_vm_ioctl_create_spapr_tce(struct kvm *kvm,
 		}
 	}
 
+	kvm_get_kvm(kvm);
 	if (!ret)
 		ret = anon_inode_getfd("kvm-spapr-tce", &kvm_spapr_tce_fops,
 				       stt, O_RDWR | O_CLOEXEC);
 
-	if (ret >= 0) {
+	if (ret >= 0)
 		list_add_rcu(&stt->list, &kvm->arch.spapr_tce_tables);
-		kvm_get_kvm(kvm);
-	}
+	else
+		kvm_put_kvm(kvm);
 
Quoted 4 lines

arch/powerpc/kvm/book3s_64_vio.c long kvm_vm_ioctl_create_spapr_tce(struct kvm *kvm, · line 351

Quoted 1 line
arch/powerpc/kvm/book3s_64_vio.c +0 -0
 	mutex_unlock(&kvm->lock);

This looks correct to me. But I feel like the logic could be cleaner, perhaps like this (patch below):

And I feel that 1) your patch tries to hide what it actually does 2) having 2 unlocks for one lock is an invite for future bugs imho, I'd think the whole point of adding new gotos is exactly to have 1 unlock.

Quoted 88 lines
	mutex_lock(&kvm->lock);

	/* Check this LIOBN hasn't been previously allocated */
	list_for_each_entry(siter, &kvm->arch.spapr_tce_tables, list) {
		if (siter->liobn = args->liobn) {
			ret = -EBUSY;
			goto fail_unlock;
		}
	}

	kvm_get_kvm(kvm);
	ret = anon_inode_getfd("kvm-spapr-tce", &kvm_spapr_tce_fops,
				stt, O_RDWR | O_CLOEXEC);

	if (ret < 0) {
		kvm_put_kvm(kvm);
		goto fail_unlock;
	}

	list_add_rcu(&stt->list, &kvm->arch.spapr_tce_tables);

	mutex_unlock(&kvm->lock);

	return ret;

fail_unlock:

	mutex_unlock(&kvm->lock);

fail:

cheers

arch/powerpc/kvm/book3s_64_vio.c +13 -11
index 532ab79734c7..5d74602db0d0 100644
@@ -296,7 +296,7 @@ long kvm_vm_ioctl_create_spapr_tce(struct kvm *kvm,
 	struct kvmppc_spapr_tce_table *stt = NULL;
 	struct kvmppc_spapr_tce_table *siter;
 	unsigned long npages, size = args->size;
-	int ret = -ENOMEM;
+	int ret;
 	int i;
 
 	if (!args->size || args->page_shift < 12 || args->page_shift > 34 ||
@@ -330,28 +330,30 @@ long kvm_vm_ioctl_create_spapr_tce(struct kvm *kvm,
 	mutex_lock(&kvm->lock);
 
 	/* Check this LIOBN hasn't been previously allocated */
-	ret = 0;
 	list_for_each_entry(siter, &kvm->arch.spapr_tce_tables, list) {
 		if (siter->liobn = args->liobn) {
 			ret = -EBUSY;
-			break;
+			goto fail_unlock;
 		}
 	}
 
-	if (!ret)
-		ret = anon_inode_getfd("kvm-spapr-tce", &kvm_spapr_tce_fops,
-				       stt, O_RDWR | O_CLOEXEC);
+	kvm_get_kvm(kvm);
+	ret = anon_inode_getfd("kvm-spapr-tce", &kvm_spapr_tce_fops,
+				stt, O_RDWR | O_CLOEXEC);
 
-	if (ret >= 0) {
-		list_add_rcu(&stt->list, &kvm->arch.spapr_tce_tables);
-		kvm_get_kvm(kvm);
+	if (ret < 0) {
+		kvm_put_kvm(kvm);
+		goto fail_unlock;
 	}
 
+	list_add_rcu(&stt->list, &kvm->arch.spapr_tce_tables);
+
 	mutex_unlock(&kvm->lock);
 
-	if (ret >= 0)
-		return ret;
+	return ret;
 
+ fail_unlock:
+	mutex_unlock(&kvm->lock);
  fail:
 	for (i = 0; i < npages; i++)
 		if (stt->pages[i])
Signature · 1 line
Alexey
#3 lore raw
In reply to Alexey Kardashevskiy
Quoted 5 lines · Alexey Kardashevskiy

The anon fd's ops releases the KVM reference in the release hook. However we reference the KVM object after we create the fd so there is small window when the release function can be called and dereferenced the KVM object which potentially may free it.

Quoted 8 lines

It is not a problem at the moment as the file is created and KVM is referenced under the KVM lock and the release function obtains the same lock before dereferencing the KVM (although the lock is not held when calling kvm_put_kvm()) but it is a fragile against future changes.

This references the KVM object before creating a file.

Thanks, applied to my kvm-ppc-next tree.

Paul.