Received-SPF: pass (google.com: domain of linux-kernel+bounces-86065-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
From: David Stevens <stevensd@chromium.org>
To: Sean Christopherson <seanjc@google.com>,
	Paolo Bonzini <pbonzini@redhat.com>
Cc: Yu Zhang <yu.c.zhang@linux.intel.com>,
	Isaku Yamahata <isaku.yamahata@gmail.com>,
	Zhi Wang <zhi.wang.linux@gmail.com>,
	Maxim Levitsky <mlevitsk@redhat.com>,
	kvmarm@lists.linux.dev,
	linux-kernel@vger.kernel.org,
	kvm@vger.kernel.org
Subject: [PATCH v11 1/8] KVM: Assert that a page's refcount is elevated when marking accessed/dirty
Date: Thu, 29 Feb 2024 11:57:52 +0900
Message-ID: <20240229025759.1187910-2-stevensd@google.com>
In-Reply-To: <20240229025759.1187910-1-stevensd@google.com>
References: <20240229025759.1187910-1-stevensd@google.com>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit

From: Sean Christopherson <seanjc@google.com>

Assert that a page's refcount is elevated, i.e. that _something_ holds a
reference to the page, when KVM marks a page as accessed and/or dirty.
KVM typically doesn't hold a reference to pages that are mapped into the
guest, e.g. to allow page migration, compaction, swap, etc., and instead
relies on mmu_notifiers to react to changes in the primary MMU.

Incorrect handling of mmu_notifier events (or similar mechanisms) can
result in KVM keeping a mapping beyond the lifetime of the backing page,
i.e. can (and often does) result in use-after-free.  Yelling if KVM marks
a freed page as accessed/dirty doesn't prevent badness as KVM usually
only does A/D updates when unmapping memory from the guest, i.e. the
assertion fires well after an underlying bug has occurred, but yelling
does help detect, triage, and debug use-after-free bugs.

Note, the assertion must use page_count(), NOT page_ref_count()!  For
hugepages, the returned struct page may be a tailpage and thus not have
its own refcount.

Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 virt/kvm/kvm_main.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 10bfc88a69f7..c5e4bf7c48f9 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -3204,6 +3204,19 @@ EXPORT_SYMBOL_GPL(kvm_vcpu_unmap);
 
 static bool kvm_is_ad_tracked_page(struct page *page)
 {
+	/*
+	 * Assert that KVM isn't attempting to mark a freed page as Accessed or
+	 * Dirty, i.e. that KVM's MMU doesn't have a use-after-free bug.  KVM
+	 * (typically) doesn't pin pages that are mapped in KVM's MMU, and
+	 * instead relies on mmu_notifiers to know when a mapping needs to be
+	 * zapped/invalidated.  Unmapping from KVM's MMU must happen _before_
+	 * KVM returns from its mmu_notifier, i.e. the page should have an
+	 * elevated refcount at this point even though KVM doesn't hold a
+	 * reference of its own.
+	 */
+	if (WARN_ON_ONCE(!page_count(page)))
+		return false;
+
 	/*
 	 * Per page-flags.h, pages tagged PG_reserved "should in general not be
 	 * touched (e.g. set dirty) except by its owner".
-- 
2.44.0.rc1.240.g4c46232300-goog