Re: [PATCH v4 10/10] KVM: arm64: np-guest CMOs with PMD_SIZE fixmap

With the introduction of stage-2 huge mappings in the pKVM hypervisor,
guest pages CMO is needed for PMD_SIZE size. Fixmap only supports
PAGE_SIZE and iterating over the huge-page is time consuming (mostly due
to TLBI on hyp_fixmap_unmap) which is a problem for EL2 latency.

Introduce a shared PMD_SIZE fixmap (hyp_fixblock_map/hyp_fixblock_unmap)
to improve guest page CMOs when stage-2 huge mappings are installed.

On a Pixel6, the iterative solution resulted in a latency of ~700us,
while the PMD_SIZE fixmap reduces it to ~100us.

Because of the horrendous private range allocation that would be
necessary, this is disabled for 64KiB pages systems.

Suggested-by: Quentin Perret <redacted>
Signed-off-by: Vincent Donnefort <redacted>
Signed-off-by: Quentin Perret <redacted>

diff --git a/arch/arm64/include/asm/kvm_pgtable.h b/arch/arm64/include/asm/kvm_pgtable.h
index 1b43bcd2a679..2888b5d03757 100644
--- a/arch/arm64/include/asm/kvm_pgtable.h
+++ b/arch/arm64/include/asm/kvm_pgtable.h

@@ -59,6 +59,11 @@ typedef u64 kvm_pte_t;
 
 #define KVM_PHYS_INVALID		(-1ULL)
 
+#define KVM_PTE_TYPE			BIT(1)
+#define KVM_PTE_TYPE_BLOCK		0
+#define KVM_PTE_TYPE_PAGE		1
+#define KVM_PTE_TYPE_TABLE		1
+
 #define KVM_PTE_LEAF_ATTR_LO		GENMASK(11, 2)
 
 #define KVM_PTE_LEAF_ATTR_LO_S1_ATTRIDX	GENMASK(4, 2)

diff --git a/arch/arm64/kvm/hyp/include/nvhe/mm.h b/arch/arm64/kvm/hyp/include/nvhe/mm.h
index 230e4f2527de..b0c72bc2d5ba 100644
--- a/arch/arm64/kvm/hyp/include/nvhe/mm.h
+++ b/arch/arm64/kvm/hyp/include/nvhe/mm.h

@@ -13,9 +13,11 @@
 extern struct kvm_pgtable pkvm_pgtable;
 extern hyp_spinlock_t pkvm_pgd_lock;
 
-int hyp_create_pcpu_fixmap(void);
+int hyp_create_fixmap(void);
 void *hyp_fixmap_map(phys_addr_t phys);
 void hyp_fixmap_unmap(void);
+void *hyp_fixblock_map(phys_addr_t phys);
+void hyp_fixblock_unmap(void);
 
 int hyp_create_idmap(u32 hyp_va_bits);
 int hyp_map_vectors(void);

diff --git a/arch/arm64/kvm/hyp/nvhe/mem_protect.c b/arch/arm64/kvm/hyp/nvhe/mem_protect.c
index 97e0fea9db4e..9f3ffa4e0690 100644
--- a/arch/arm64/kvm/hyp/nvhe/mem_protect.c
+++ b/arch/arm64/kvm/hyp/nvhe/mem_protect.c

@@ -220,16 +220,52 @@ static void guest_s2_put_page(void *addr)
 	hyp_put_page(&current_vm->pool, addr);
 }
 
+static void *__fixmap_guest_page(void *va, size_t *size)
+{
+	if (IS_ALIGNED(*size, PMD_SIZE)) {
+		void *addr = hyp_fixblock_map(__hyp_pa(va));
+
+		if (addr)
+			return addr;
+
+		*size = PAGE_SIZE;
+	}
+
+	if (IS_ALIGNED(*size, PAGE_SIZE))
+		return hyp_fixmap_map(__hyp_pa(va));
+
+	WARN_ON(1);
+
+	return NULL;
+}
+
+static void __fixunmap_guest_page(size_t size)
+{
+	switch (size) {
+	case PAGE_SIZE:
+		hyp_fixmap_unmap();
+		break;
+	case PMD_SIZE:
+		hyp_fixblock_unmap();
+		break;
+	default:
+		WARN_ON(1);
+	}

 	}
 }
 

@@ -238,11 +274,14 @@ static void invalidate_icache_guest_page(void *va, size_t size)
 	WARN_ON(!PAGE_ALIGNED(size));
 
 	while (size) {
-		__invalidate_icache_guest_page(hyp_fixmap_map(__hyp_pa(va)),
-					       PAGE_SIZE);
-		hyp_fixmap_unmap();
-		va += PAGE_SIZE;
-		size -= PAGE_SIZE;
+		size_t fixmap_size = size == PMD_SIZE ? size : PAGE_SIZE;
+		void *addr = __fixmap_guest_page(va, &fixmap_size);
+
+		__invalidate_icache_guest_page(addr, fixmap_size);
+		__fixunmap_guest_page(fixmap_size);
+
+		size -= fixmap_size;
+		va += fixmap_size;
 	}
 }
 

diff --git a/arch/arm64/kvm/hyp/nvhe/mm.c b/arch/arm64/kvm/hyp/nvhe/mm.c
index f41c7440b34b..e3b1bece8504 100644
--- a/arch/arm64/kvm/hyp/nvhe/mm.c
+++ b/arch/arm64/kvm/hyp/nvhe/mm.c

@@ -229,9 +229,8 @@ int hyp_map_vectors(void)
 	return 0;
 }
 
-void *hyp_fixmap_map(phys_addr_t phys)
+static void *fixmap_map_slot(struct hyp_fixmap_slot *slot, phys_addr_t phys)
 {
-	struct hyp_fixmap_slot *slot = this_cpu_ptr(&fixmap_slots);
 	kvm_pte_t pte, *ptep = slot->ptep;
 
 	pte = *ptep;

@@ -243,10 +242,21 @@ void *hyp_fixmap_map(phys_addr_t phys)
 	return (void *)slot->addr;
 }
 
+void *hyp_fixmap_map(phys_addr_t phys)
+{
+	return fixmap_map_slot(this_cpu_ptr(&fixmap_slots), phys);
+}
+
 static void fixmap_clear_slot(struct hyp_fixmap_slot *slot)
 {
 	kvm_pte_t *ptep = slot->ptep;
 	u64 addr = slot->addr;
+	u32 level;
+
+	if (FIELD_GET(KVM_PTE_TYPE, *ptep) == KVM_PTE_TYPE_PAGE)
+		level = KVM_PGTABLE_LAST_LEVEL;
+	else
+		level = KVM_PGTABLE_LAST_LEVEL - 1; /* create_fixblock() guarantees PMD level */

 	WRITE_ONCE(*ptep, *ptep & ~KVM_PTE_VALID);
 

@@ -260,7 +270,7 @@ static void fixmap_clear_slot(struct hyp_fixmap_slot *slot)
 	 * https://lore.kernel.org/kvm/20221017115209.2099-1-will@kernel.org/T/#mf10dfbaf1eaef9274c581b81c53758918c1d0f03
 	 */
 	dsb(ishst);
-	__tlbi_level(vale2is, __TLBI_VADDR(addr, 0), KVM_PGTABLE_LAST_LEVEL);
+	__tlbi_level(vale2is, __TLBI_VADDR(addr, 0), level);
 	dsb(ish);
 	isb();
 }

@@ -273,9 +283,9 @@ void hyp_fixmap_unmap(void)
 static int __create_fixmap_slot_cb(const struct kvm_pgtable_visit_ctx *ctx,
 				   enum kvm_pgtable_walk_flags visit)
 {
-	struct hyp_fixmap_slot *slot = per_cpu_ptr(&fixmap_slots, (u64)ctx->arg);
+	struct hyp_fixmap_slot *slot = (struct hyp_fixmap_slot *)ctx->arg;
 
-	if (!kvm_pte_valid(ctx->old) || ctx->level != KVM_PGTABLE_LAST_LEVEL)
+	if (!kvm_pte_valid(ctx->old) || (ctx->end - ctx->start) != kvm_granule_size(ctx->level))
 		return -EINVAL;
 
 	slot->addr = ctx->addr;

@@ -296,13 +306,73 @@ static int create_fixmap_slot(u64 addr, u64 cpu)
 	struct kvm_pgtable_walker walker = {
 		.cb	= __create_fixmap_slot_cb,
 		.flags	= KVM_PGTABLE_WALK_LEAF,
-		.arg = (void *)cpu,
+		.arg = (void *)per_cpu_ptr(&fixmap_slots, cpu),