Skip to contentSkip to conversation

KVM: x86/mmu: Check write tracking in all address spaces

PATCH Applied Artem Dinaburg 2 participants 4 messages Aug 26 – Aug 28

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

cover KVM: x86/mmu: Check write tracking in all address spaces

Please queue upstream commit 0f38453cdb2e17566ccb7c0f3dabd5bd21caca26 ("KVM: x86/mmu: Check write tracking in all address spaces") for the stable trees. It is in mainline as of v7.2-rc6.

The fix is absent by code state from 7.1.10, 6.18.46, 6.12.105, 6.6.153 and 6.1.184.

7.1.y, 6.18.y, 6.12.y: please cherry-pick the upstream commit directly. No patch is included for these trees. Commit 0f38453cdb2e applies and compiles unmodified on all three, so there is nothing to adapt:

git cherry-pick -x 0f38453cdb2e17566ccb7c0f3dabd5bd21caca26

6.6.y and 6.1.y need adapted backports; they are patches 1/2 and 2/2 of this series. On 6.6 the upstream commit applies but does not build, because KVM_MAX_NR_ADDRESS_SPACES and kvm_arch_nr_memslot_as_ids() do not exist there. On 6.1 the upstream commit conflicts: the function is still kvm_slot_page_track_is_active() and the tracking array is mode-indexed gfn_track[].

5.15.y and 5.10.y are not covered by this submission. Those trees were not assessed.

Jinu Kim (2):
[6.6.y] KVM: x86/mmu: Check write tracking in all address spaces [6.1.y] KVM: x86/mmu: Check write tracking in all address spaces

Each patch touches only arch/x86/kvm/mmu/page_track.c:
1/2 (6.6.y): 26 insertions(+), 6 deletions(-) 2/2 (6.1.y): 22 insertions(+), 6 deletions(-)

Signature · 1 line
2.43.0
#1 lore raw

PATCH 1/2 KVM: x86/mmu: Check write tracking in all address spaces

From: Jinu Kim <[email protected]>

commit 0f38453cdb2e17566ccb7c0f3dabd5bd21caca26 upstream.

kvm_gfn_is_write_tracked() checks only the supplied memslot, but page tracking is per-address-space and shadow pages are shared across all address spaces. With SMM, a GFN can therefore be write-tracked in one address space and appear untracked through the other.

Check the supplied slot first, then the slot for the other address space. This ensures all callers honor write tracking regardless of the active address space. In particular, it prevents mmu_try_to_unsync_pages() from marking an upper-level shadow page unsync and eventually triggering the BUG in pte_list_remove().

Fixes 699023e23965 ("KVM: x86: add SMM to the MMU role, support SMRAM address space")
Assisted-by Codex:GPT-5

Message-ID: <[email protected]>

[invert direction of the conditional. - Paolo]

[ artem: adapt to 6.6, which has neither KVM_MAX_NR_ADDRESS_SPACES nor
kvm_arch_nr_memslot_as_ids(); use KVM_ADDRESS_SPACE_NUM instead. That macro is CONFIG_KVM_SMM-conditional in 6.6, so the guard on KVM_ADDRESS_SPACE_NUM > 1 around the peer-slot lookup is retained ]

---
Target tree: linux-6.6.y (stable).

Verified: applies to v6.6.153; arch/x86/kvm/mmu/page_track.o builds clean with CONFIG_KVM_SMM=y and with CONFIG_KVM_SMM=n.

 arch/x86/kvm/mmu/page_track.c | 32 ++++++++++++++++++++++++++------
 1 file changed, 26 insertions(+), 6 deletions(-)
arch/x86/kvm/mmu/page_track.c +26 -6
index c87da11f3a04..2615074a4bff 100644
@@ -117,13 +117,23 @@ void __kvm_write_track_remove_gfn(struct kvm *kvm,
 	kvm_mmu_gfn_allow_lpage(slot, gfn);
 }
 
-/*
- * check if the corresponding access on the specified guest page is tracked.
- */
+static bool __kvm_gfn_is_write_tracked(const struct kvm_memory_slot *slot,
+				       gfn_t gfn)
+{
+	int index;
+
+	if (!slot)
+		return false;
+
+	index = gfn_to_index(gfn, slot->base_gfn, PG_LEVEL_4K);
+	return !!READ_ONCE(slot->arch.gfn_write_track[index]);
+}
+
+/* check if write access is tracked on the specified guest page.  */
 bool kvm_gfn_is_write_tracked(struct kvm *kvm,
 			      const struct kvm_memory_slot *slot, gfn_t gfn)
 {
-	int index;
+	const struct kvm_memory_slot *other_slot;
 
 	if (!slot)
 		return false;
@@ -131,8 +141,18 @@ bool kvm_gfn_is_write_tracked(struct kvm *kvm,
 	if (!kvm_page_track_write_tracking_enabled(kvm))
 		return false;
 
-	index = gfn_to_index(gfn, slot->base_gfn, PG_LEVEL_4K);
-	return !!READ_ONCE(slot->arch.gfn_write_track[index]);
+	BUILD_BUG_ON(KVM_ADDRESS_SPACE_NUM > 2);
+
+	if (__kvm_gfn_is_write_tracked(slot, gfn))
+		return true;
+
+	if (KVM_ADDRESS_SPACE_NUM > 1) {
+		other_slot = __gfn_to_memslot(__kvm_memslots(kvm, slot->as_id ^ 1), gfn);
+		if (__kvm_gfn_is_write_tracked(other_slot, gfn))
+			return true;
+	}
+
+	return false;
 }
 
 #ifdef CONFIG_KVM_EXTERNAL_WRITE_TRACKING
Signature · 1 line
2.43.0
#2 lore raw

PATCH 2/2 KVM: x86/mmu: Check write tracking in all address spaces

In reply to Artem Dinaburg

From: Jinu Kim <[email protected]>

commit 0f38453cdb2e17566ccb7c0f3dabd5bd21caca26 upstream.

kvm_gfn_is_write_tracked() checks only the supplied memslot, but page tracking is per-address-space and shadow pages are shared across all address spaces. With SMM, a GFN can therefore be write-tracked in one address space and appear untracked through the other.

Check the supplied slot first, then the slot for the other address space. This ensures all callers honor write tracking regardless of the active address space. In particular, it prevents mmu_try_to_unsync_pages() from marking an upper-level shadow page unsync and eventually triggering the BUG in pte_list_remove().

Fixes 699023e23965 ("KVM: x86: add SMM to the MMU role, support SMRAM address space")
Assisted-by Codex:GPT-5

Message-ID: <[email protected]>

[invert direction of the conditional. - Paolo]

[ artem: adapt to 6.1's kvm_slot_page_track_is_active() and mode-indexed
gfn_track[]; KVM_ADDRESS_SPACE_NUM is unconditionally 2 on x86 in 6.1, so the peer-slot lookup needs no guard ]

---
Target tree: linux-6.1.y (stable).

Verified: applies to v6.1.184; arch/x86/kvm/ builds clean with x86_64 defconfig plus CONFIG_KVM=m, CONFIG_KVM_INTEL=m and CONFIG_KVM_AMD=m, gcc 13.3.0.

 arch/x86/kvm/mmu/page_track.c | 28 ++++++++++++++++++++++------
 1 file changed, 22 insertions(+), 6 deletions(-)
arch/x86/kvm/mmu/page_track.c +22 -6
index 2e09d1b6249f..51d5a912c553 100644
@@ -171,14 +171,25 @@ void kvm_slot_page_track_remove_page(struct kvm *kvm,
 }
 EXPORT_SYMBOL_GPL(kvm_slot_page_track_remove_page);
 
-/*
- * check if the corresponding access on the specified guest page is tracked.
- */
+static bool __kvm_slot_page_track_is_active(const struct kvm_memory_slot *slot,
+					    gfn_t gfn,
+					    enum kvm_page_track_mode mode)
+{
+	int index;
+
+	if (!slot)
+		return false;
+
+	index = gfn_to_index(gfn, slot->base_gfn, PG_LEVEL_4K);
+	return !!READ_ONCE(slot->arch.gfn_track[mode][index]);
+}
+
+/* check if write access is tracked on the specified guest page. */
 bool kvm_slot_page_track_is_active(struct kvm *kvm,
 				   const struct kvm_memory_slot *slot,
 				   gfn_t gfn, enum kvm_page_track_mode mode)
 {
-	int index;
+	const struct kvm_memory_slot *other_slot;
 
 	if (WARN_ON(!page_track_mode_is_valid(mode)))
 		return false;
@@ -190,8 +201,13 @@ bool kvm_slot_page_track_is_active(struct kvm *kvm,
 	    !kvm_page_track_write_tracking_enabled(kvm))
 		return false;
 
-	index = gfn_to_index(gfn, slot->base_gfn, PG_LEVEL_4K);
-	return !!READ_ONCE(slot->arch.gfn_track[mode][index]);
+	BUILD_BUG_ON(KVM_ADDRESS_SPACE_NUM > 2);
+
+	if (__kvm_slot_page_track_is_active(slot, gfn, mode))
+		return true;
+
+	other_slot = __gfn_to_memslot(__kvm_memslots(kvm, slot->as_id ^ 1), gfn);
+	return __kvm_slot_page_track_is_active(other_slot, gfn, mode);
 }
 
 void kvm_page_track_cleanup(struct kvm *kvm)
Signature · 1 line
2.43.0
maintainer Applied
#3 lore raw
In reply to Artem Dinaburg
Quoted 5 lines

7.1.y, 6.18.y, 6.12.y: please cherry-pick the upstream commit directly. No patch is included for these trees.

6.6.y and 6.1.y need adapted backports; they are patches 1/2 and 2/2 of this series.

Queued for 7.1, 6.18, 6.12, 6.6 and 6.1, thanks.

Signature · 2 lines
Thanks,
Sasha