KVM: arm64: Protect the .hyp sections from the host

When KVM runs in nVHE protected mode, use the host stage 2 to unmap the
hypervisor sections by marking them as owned by the hypervisor itself.
The long-term goal is to ensure the EL2 code can remain robust
regardless of the host's state, so this starts by making sure the host
cannot e.g. write to the .hyp sections directly.

Acked-by: Will Deacon <will@kernel.org>
Signed-off-by: Quentin Perret <qperret@google.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210319100146.1149909-39-qperret@google.com

4 files changed, 90 insertions, 0 deletions

diff --git a/arch/arm64/kvm/arm.c b/arch/arm64/kvm/arm.c
index d237c378e6fb..368159021dee 100644
--- a/arch/arm64/kvm/arm.c
+++ b/arch/arm64/kvm/arm.c
@@ -1899,11 +1899,57 @@ void _kvm_host_prot_finalize(void *discard)
 	WARN_ON(kvm_call_hyp_nvhe(__pkvm_prot_finalize));
 }
 
+static inline int pkvm_mark_hyp(phys_addr_t start, phys_addr_t end)
+{
+	return kvm_call_hyp_nvhe(__pkvm_mark_hyp, start, end);
+}
+
+#define pkvm_mark_hyp_section(__section)		\
+	pkvm_mark_hyp(__pa_symbol(__section##_start),	\
+			__pa_symbol(__section##_end))
+
 static int finalize_hyp_mode(void)
 {
+	int cpu, ret;
+
 	if (!is_protected_kvm_enabled())
 		return 0;
 
+	ret = pkvm_mark_hyp_section(__hyp_idmap_text);
+	if (ret)
+		return ret;
+
+	ret = pkvm_mark_hyp_section(__hyp_text);
+	if (ret)
+		return ret;
+
+	ret = pkvm_mark_hyp_section(__hyp_rodata);
+	if (ret)
+		return ret;
+
+	ret = pkvm_mark_hyp_section(__hyp_bss);
+	if (ret)
+		return ret;
+
+	ret = pkvm_mark_hyp(hyp_mem_base, hyp_mem_base + hyp_mem_size);
+	if (ret)
+		return ret;
+
+	for_each_possible_cpu(cpu) {
+		phys_addr_t start = virt_to_phys((void *)kvm_arm_hyp_percpu_base[cpu]);
+		phys_addr_t end = start + (PAGE_SIZE << nvhe_percpu_order());
+
+		ret = pkvm_mark_hyp(start, end);
+		if (ret)
+			return ret;
+
+		start = virt_to_phys((void *)per_cpu(kvm_arm_hyp_stack_page, cpu));
+		end = start + PAGE_SIZE;
+		ret = pkvm_mark_hyp(start, end);
+		if (ret)
+			return ret;
+	}
+
 	/*
 	 * Flip the static key upfront as that may no longer be possible
 	 * once the host stage 2 is installed.
diff --git a/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h b/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h
index d293cb328cc4..42d81ec739fa 100644
--- a/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h
+++ b/arch/arm64/kvm/hyp/include/nvhe/mem_protect.h
@@ -21,6 +21,8 @@ struct host_kvm {
 extern struct host_kvm host_kvm;
 
 int __pkvm_prot_finalize(void);
+int __pkvm_mark_hyp(phys_addr_t start, phys_addr_t end);
+
 int kvm_host_prepare_stage2(void *mem_pgt_pool, void *dev_pgt_pool);
 void handle_host_mem_abort(struct kvm_cpu_context *host_ctxt);
 
diff --git a/arch/arm64/kvm/hyp/nvhe/hyp-main.c b/arch/arm64/kvm/hyp/nvhe/hyp-main.c
index 69163f2cbb63..b4eaa7ef13e0 100644
--- a/arch/arm64/kvm/hyp/nvhe/hyp-main.c
+++ b/arch/arm64/kvm/hyp/nvhe/hyp-main.c
@@ -156,6 +156,14 @@ static void handle___pkvm_prot_finalize(struct kvm_cpu_context *host_ctxt)
 {
 	cpu_reg(host_ctxt, 1) = __pkvm_prot_finalize();
 }
+
+static void handle___pkvm_mark_hyp(struct kvm_cpu_context *host_ctxt)
+{
+	DECLARE_REG(phys_addr_t, start, host_ctxt, 1);
+	DECLARE_REG(phys_addr_t, end, host_ctxt, 2);
+
+	cpu_reg(host_ctxt, 1) = __pkvm_mark_hyp(start, end);
+}
 typedef void (*hcall_t)(struct kvm_cpu_context *);
 
 #define HANDLE_FUNC(x)	[__KVM_HOST_SMCCC_FUNC_##x] = (hcall_t)handle_##x
@@ -180,6 +188,7 @@ static const hcall_t host_hcall[] = {
 	HANDLE_FUNC(__pkvm_create_mappings),
 	HANDLE_FUNC(__pkvm_create_private_mapping),
 	HANDLE_FUNC(__pkvm_prot_finalize),
+	HANDLE_FUNC(__pkvm_mark_hyp),
 };
 
 static void handle_host_hcall(struct kvm_cpu_context *host_ctxt)
diff --git a/arch/arm64/kvm/hyp/nvhe/mem_protect.c b/arch/arm64/kvm/hyp/nvhe/mem_protect.c
index 77b48c47344d..808e2471091b 100644
--- a/arch/arm64/kvm/hyp/nvhe/mem_protect.c
+++ b/arch/arm64/kvm/hyp/nvhe/mem_protect.c
@@ -27,6 +27,8 @@ struct host_kvm host_kvm;
 struct hyp_pool host_s2_mem;
 struct hyp_pool host_s2_dev;
 
+static const u8 pkvm_hyp_id = 1;
+
 static void *host_s2_zalloc_pages_exact(size_t size)
 {
 	return hyp_alloc_pages(&host_s2_mem, get_order(size));
@@ -182,6 +184,18 @@ static bool find_mem_range(phys_addr_t addr, struct kvm_mem_range *range)
 	return false;
 }
 
+static bool range_is_memory(u64 start, u64 end)
+{
+	struct kvm_mem_range r1, r2;
+
+	if (!find_mem_range(start, &r1) || !find_mem_range(end, &r2))
+		return false;
+	if (r1.start != r2.start)
+		return false;
+
+	return true;
+}
+
 static inline int __host_stage2_idmap(u64 start, u64 end,
 				      enum kvm_pgtable_prot prot,
 				      struct hyp_pool *pool)
@@ -229,6 +243,25 @@ unlock:
 	return ret;
 }
 
+int __pkvm_mark_hyp(phys_addr_t start, phys_addr_t end)
+{
+	int ret;
+
+	/*
+	 * host_stage2_unmap_dev_all() currently relies on MMIO mappings being
+	 * non-persistent, so don't allow changing page ownership in MMIO range.
+	 */
+	if (!range_is_memory(start, end))
+		return -EINVAL;
+
+	hyp_spin_lock(&host_kvm.lock);
+	ret = kvm_pgtable_stage2_set_owner(&host_kvm.pgt, start, end - start,
+					   &host_s2_mem, pkvm_hyp_id);
+	hyp_spin_unlock(&host_kvm.lock);
+
+	return ret != -EAGAIN ? ret : 0;
+}
+
 void handle_host_mem_abort(struct kvm_cpu_context *host_ctxt)
 {
 	struct kvm_vcpu_fault_info fault;