linux-kernel - Re: [PATCH] x86/kvm: Avoid freeing stack-allocated node in kvm_async_pf_queue

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <87cy4vlmv8.fsf@redhat.com>
Date: Wed, 03 Dec 2025 14:24:11 +0100
From: Vitaly Kuznetsov <vkuznets@...hat.com>
To: Ryosuke Yasuoka <ryasuoka@...hat.com>
Cc: x86@...nel.org, kvm@...r.kernel.org, linux-kernel@...r.kernel.org,
 pbonzini@...hat.com, tglx@...utronix.de, mingo@...hat.com, bp@...en8.de,
 dave.hansen@...ux.intel.com, hpa@...or.com
Subject: Re: [PATCH] x86/kvm: Avoid freeing stack-allocated node in
 kvm_async_pf_queue_task

Ryosuke Yasuoka <ryasuoka@...hat.com> writes:

> kvm_async_pf_queue_task() can incorrectly remove a node allocated on the
> stack of kvm_async_pf_task_wait_schedule(). This occurs when a task
> request a PF while another task's PF request with the same token is
> still pending. 

The important missing part here is what the 'token' is. exc_page_fault()
sets it to (u32)read_cr2() so indeed I see possibilities that two
different tasks will generate the same token.

> Currently, kvm_async_pf_queue_task() assumes that any
> entry in the list is a dummy entry and tries to kfree(). To fix this,
> add a dummy flag to the node structure and the function should check
> this flag and kfree() only if it is a dummy entry.
>
> Signed-off-by: Ryosuke Yasuoka <ryasuoka@...hat.com>
> ---
>  arch/x86/kernel/kvm.c | 13 +++++++++++--
>  1 file changed, 11 insertions(+), 2 deletions(-)
>
> diff --git a/arch/x86/kernel/kvm.c b/arch/x86/kernel/kvm.c
> index b67d7c59dca0..2c92ec528379 100644
> --- a/arch/x86/kernel/kvm.c
> +++ b/arch/x86/kernel/kvm.c
> @@ -88,6 +88,7 @@ struct kvm_task_sleep_node {
>  	struct swait_queue_head wq;
>  	u32 token;
>  	int cpu;
> +	bool dummy;
>  };
>  
>  static struct kvm_task_sleep_head {
> @@ -119,10 +120,17 @@ static bool kvm_async_pf_queue_task(u32 token, struct kvm_task_sleep_node *n)
>  	raw_spin_lock(&b->lock);
>  	e = _find_apf_task(b, token);
>  	if (e) {
> +		struct kvm_task_sleep_node *dummy = NULL;
> +
>  		/* dummy entry exist -> wake up was delivered ahead of PF */
> -		hlist_del(&e->link);
> +		/* Otherwise it should not be freed here. */

I think we can merge these two comments together and write something
better, e.g.
	/* 
         * The entry can either be a 'dummy' entry (which is put on the
         * list when wake-up happens ahead of APF handling completion)
         * or a token from another task which should not be touched.
         */

> +		if (e->dummy) {
> +			hlist_del(&e->link);
> +			dummy = e;
> +		}
> +
>  		raw_spin_unlock(&b->lock);
> -		kfree(e);
> +		kfree(dummy);
>  		return false;
>  	}


I think you also need to do 

diff --git a/arch/x86/kernel/kvm.c b/arch/x86/kernel/kvm.c
index b67d7c59dca0..0a84a3100e72 100644
--- a/arch/x86/kernel/kvm.c
+++ b/arch/x86/kernel/kvm.c
@@ -128,6 +128,7 @@ static bool kvm_async_pf_queue_task(u32 token, struct kvm_task_sleep_node *n)
 
        n->token = token;
        n->cpu = smp_processor_id();
+       n->dummy = false;
        init_swait_queue_head(&n->wq);
        hlist_add_head(&n->link, &b->list);
        raw_spin_unlock(&b->lock);

now as kvm_async_pf_task_wait_schedule() allocates struct
kvm_task_sleep_node() on the stack and thus 'n->dummy' can be anything.

>  
> @@ -230,6 +238,7 @@ static void kvm_async_pf_task_wake(u32 token)
>  		}
>  		dummy->token = token;
>  		dummy->cpu = smp_processor_id();
> +		dummy->dummy = true;
>  		init_swait_queue_head(&dummy->wq);
>  		hlist_add_head(&dummy->link, &b->list);
>  		dummy = NULL;
>
> base-commit: 2eba5e05d9bcf4cdea995ed51b0f07ba0275794a

Sorry for delayed reply!

-- 
Vitaly