linux-kernel - [PATCH 2/6] KVM: arm64: Add KVM_PGTABLE_WALK_PRE

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 for Android: free password hash cracker in your pocket

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

Message-Id: <20250911134648.58945-3-jackabt.amazon@gmail.com>
Date: Thu, 11 Sep 2025 14:46:44 +0100
From: Jack Thomson <jackabt.amazon@...il.com>
To: maz@...nel.org,
	oliver.upton@...ux.dev,
	pbonzini@...hat.com
Cc: joey.gouly@....com,
	suzuki.poulose@....com,
	yuzenghui@...wei.com,
	catalin.marinas@....com,
	will@...nel.org,
	shuah@...nel.org,
	linux-arm-kernel@...ts.infradead.org,
	kvmarm@...ts.linux.dev,
	linux-kernel@...r.kernel.org,
	linux-kselftest@...r.kernel.org,
	isaku.yamahata@...el.com,
	roypat@...zon.co.uk,
	kalyazin@...zon.co.uk,
	jackabt@...zon.com
Subject: [PATCH 2/6] KVM: arm64: Add KVM_PGTABLE_WALK_PRE_FAULT walk flag

From: Jack Thomson <jackabt@...zon.com>

Don't return -EAGAIN from stage2_map_walker_try_leaf during
KVM_PRE_FAULT_MEMORY.

During pre-faults, user_abort() is retried upon returning -EAGAIN,
meaning the ioctl would get stuck in an infinite loop if userspace
tries to pre-fault already existing mappings

Signed-off-by: Jack Thomson <jackabt@...zon.com>
---
 arch/arm64/include/asm/kvm_pgtable.h | 3 +++
 arch/arm64/kvm/hyp/pgtable.c         | 6 +++++-
 2 files changed, 8 insertions(+), 1 deletion(-)

diff --git a/arch/arm64/include/asm/kvm_pgtable.h b/arch/arm64/include/asm/kvm_pgtable.h
index 2888b5d03757..0789671d1c4f 100644
--- a/arch/arm64/include/asm/kvm_pgtable.h
+++ b/arch/arm64/include/asm/kvm_pgtable.h
@@ -296,6 +296,8 @@ typedef bool (*kvm_pgtable_force_pte_cb_t)(u64 addr, u64 end,
  * @KVM_PGTABLE_WALK_SKIP_CMO:		Visit and update table entries
  *					without Cache maintenance
  *					operations required.
+ * @KVM_PGTABLE_WALK_PRE_FAULT		Indicates the page-table walk was
+ *					invoked from a pre-fault request.
  */
 enum kvm_pgtable_walk_flags {
 	KVM_PGTABLE_WALK_LEAF			= BIT(0),
@@ -305,6 +307,7 @@ enum kvm_pgtable_walk_flags {
 	KVM_PGTABLE_WALK_HANDLE_FAULT		= BIT(4),
 	KVM_PGTABLE_WALK_SKIP_BBM_TLBI		= BIT(5),
 	KVM_PGTABLE_WALK_SKIP_CMO		= BIT(6),
+	KVM_PGTABLE_WALK_PRE_FAULT		= BIT(7),
 };
 
 struct kvm_pgtable_visit_ctx {
diff --git a/arch/arm64/kvm/hyp/pgtable.c b/arch/arm64/kvm/hyp/pgtable.c
index c351b4abd5db..140dccec2c5b 100644
--- a/arch/arm64/kvm/hyp/pgtable.c
+++ b/arch/arm64/kvm/hyp/pgtable.c
@@ -914,9 +914,13 @@ static int stage2_map_walker_try_leaf(const struct kvm_pgtable_visit_ctx *ctx,
 	 * same mapping or only change the access permissions. Instead,
 	 * the vCPU will exit one more time from guest if still needed
 	 * and then go through the path of relaxing permissions.
+	 *
+	 * When walking in the context of a pre-fault request, if the
+	 * mapping already exists we can return 0, as there's nothing
+	 * to do.
 	 */
 	if (!stage2_pte_needs_update(ctx->old, new))
-		return -EAGAIN;
+		return (ctx->flags & KVM_PGTABLE_WALK_PRE_FAULT) ? 0 : -EAGAIN;
 
 	/* If we're only changing software bits, then store them and go! */
 	if (!kvm_pgtable_walk_shared(ctx) &&
-- 
2.43.0