Skip to contentSkip to conversation

KVM: x86: Handle SRCU initialization failure during page track init

PATCH In review [email protected] 4 participants 4 messages Sep 3, 2021 – Sep 21, 2021

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

PATCH v2 1/1 KVM: x86: Handle SRCU initialization failure during page track init

From: Haimin Zhang <[email protected]>

Check the return of init_srcu_struct(), which can fail due to OOM, when initializing the page track mechanism. Lack of checking leads to a NULL pointer deref found by a modified syzkaller.

---
 arch/x86/include/asm/kvm_page_track.h | 2 +-
 arch/x86/kvm/mmu/page_track.c         | 4 ++--
 arch/x86/kvm/x86.c                    | 6 +++++-
 3 files changed, 8 insertions(+), 4 deletions(-)
arch/x86/include/asm/kvm_page_track.h +1 -1
index 8766adb..9cd9230 100644
@@ -46,7 +46,7 @@ struct kvm_page_track_notifier_node {
 			    struct kvm_page_track_notifier_node *node);
 };
 
-void kvm_page_track_init(struct kvm *kvm);
+int kvm_page_track_init(struct kvm *kvm);
 void kvm_page_track_cleanup(struct kvm *kvm);
 
 void kvm_page_track_free_memslot(struct kvm_memory_slot *slot);
arch/x86/kvm/mmu/page_track.c +2 -2
index a9e2e02..859800f 100644
@@ -170,13 +170,13 @@ void kvm_page_track_cleanup(struct kvm *kvm)
 	cleanup_srcu_struct(&head->track_srcu);
 }
 
-void kvm_page_track_init(struct kvm *kvm)
+int kvm_page_track_init(struct kvm *kvm)
 {
 	struct kvm_page_track_notifier_head *head;
 
 	head = &kvm->arch.track_notifier_head;
-	init_srcu_struct(&head->track_srcu);
 	INIT_HLIST_HEAD(&head->track_notifier_list);
+	return init_srcu_struct(&head->track_srcu);
 }
 
 /*
arch/x86/kvm/x86.c +5 -1
index 86539c1..9a122af 100644
@@ -11143,6 +11143,8 @@ void kvm_arch_free_vm(struct kvm *kvm)
 
 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
 {
+	int ret;
+
 	if (type)
 		return -EINVAL;
 
@@ -11178,7 +11180,9 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
 
 	kvm_apicv_init(kvm);
 	kvm_hv_init_vm(kvm);
-	kvm_page_track_init(kvm);
+	ret = kvm_page_track_init(kvm);
+	if (ret)
+		return ret;
 	kvm_mmu_init_vm(kvm);
 	kvm_xen_init_vm(kvm);
 
Signature · 1 line
1.8.3.1
Acked-by on v2 1/1
#1 lore raw
Quoted 10 lines · On Fri, 2021-09-03 at 10:37 +0800, [email protected] wrote:

From: Haimin Zhang <[email protected]>

Check the return of init_srcu_struct(), which can fail due to OOM, when initializing the page track mechanism. Lack of checking leads to a NULL pointer deref found by a modified syzkaller.

I'd drop reported-by. It's not a person (I guess) and the desc is self-contained already.

Anyway,

/Jarkko

#2 lore raw
In reply to [email protected]
Quoted 9 lines · On Fri, Sep 03, 2021, [email protected] wrote:

From: Haimin Zhang <[email protected]>

Check the return of init_srcu_struct(), which can fail due to OOM, when initializing the page track mechanism. Lack of checking leads to a NULL pointer deref found by a modified syzkaller.

Neither myself nor Vitaly provided an SOB, nor is one needed. Review feedback can be attributed/noted in the part that git ignores (below the three dashes).

Quoted 3 lines

---

Notes about version changes, e.g. to document/attribute review feedback, go here.

v2:
- Blah blah blah [Vitaly, Sean]

Quoted 5 lines
 arch/x86/include/asm/kvm_page_track.h | 2 +-
 arch/x86/kvm/mmu/page_track.c         | 4 ++--
 arch/x86/kvm/x86.c                    | 6 +++++-
 3 files changed, 8 insertions(+), 4 deletions(-)

...

Quoted 14 lines
arch/x86/kvm/x86.c +2 -0
index 86539c1..9a122af 100644
@@ -11143,6 +11143,8 @@ void kvm_arch_free_vm(struct kvm *kvm)
 
 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
 {
+	int ret;
+
 	if (type)
 		return -EINVAL;
 

arch/x86/kvm/x86.c int kvm_arch_init_vm(struct kvm *kvm, unsigned long type) · line 11185

Quoted 7 lines
arch/x86/kvm/x86.c +3 -1
@@ -11178,7 +11180,9 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
 
 	kvm_apicv_init(kvm);
 	kvm_hv_init_vm(kvm);
-	kvm_page_track_init(kvm);
+	ret = kvm_page_track_init(kvm);
+	if (ret)
+		return ret;

Before moving forward with a fix, I'd like to get Paolo's input on dropping track_srcu in favor of kvm->srcu and avoiding this altogheter. Note, Paolo is on vacation at the moment, so this won't get attention for a week or more.

[*] https://lkml.kernel.org/r/[email protected]

Quoted 6 lines
arch/x86/kvm/x86.c +0 -0
 	kvm_mmu_init_vm(kvm);
 	kvm_xen_init_vm(kvm);
 
Signature · 1 line
1.8.3.1
maintainer
#3 lore raw
Quoted 3 lines · Sean Christopherson

Before moving forward with a fix, I'd like to get Paolo's input on dropping track_srcu in favor of kvm->srcu and avoiding this altogheter. Note, Paolo is on vacation at the moment, so this won't get attention for a week or more.

The reason for track_srcu's existence is to avoid complications in kvm_arch_flush_shadow_memslot, which is called from the _write_ side of kvm->srcu but is on the _read_ side of track_srcu.

I think this should be fixed easily by taking slots_lock in kvm_page_track_register_notifier and kvm_page_track_unregister_notifier, however it's a bit more complicated from the point of view of the lock hierarchy and possible deadlocks.

So I'm open to patches that drop track_srcu, but for now I applied this patch.

Paolo