diff mbox series

[v19,074/130] KVM: TDX: Create initial guest memory

Message ID bbac4998cfb34da496646491038b03f501964cbd.1708933498.git.isaku.yamahata@intel.com
State New
Headers
Series [v19,001/130] x86/virt/tdx: Rename _offset to _member for TD_SYSINFO_MAP() macro |

Commit Message

Isaku Yamahata Feb. 26, 2024, 8:26 a.m. UTC 
  From: Isaku Yamahata <isaku.yamahata@intel.com>

Because the guest memory is protected in TDX, the creation of the initial
guest memory requires a dedicated TDX module API, tdh_mem_page_add, instead
of directly copying the memory contents into the guest memory in the case
of the default VM type.  KVM MMU page fault handler callback,
set_private_spte, handles it.

Implement the hooks for KVM_MEMORY_MAPPING, pre_memory_mapping() and
post_memory_mapping() to handle it.

Signed-off-by: Isaku Yamahata <isaku.yamahata@intel.com>
---
v19:
- Switched to use KVM_MEMORY_MAPPING
- Dropped measurement extension
- updated commit message. private_page_add() => set_private_spte()
---
 arch/x86/kvm/vmx/main.c    | 23 ++++++++++
 arch/x86/kvm/vmx/tdx.c     | 93 ++++++++++++++++++++++++++++++++++++--
 arch/x86/kvm/vmx/tdx.h     |  4 ++
 arch/x86/kvm/vmx/x86_ops.h | 12 +++++
 4 files changed, 129 insertions(+), 3 deletions(-)
diff mbox series

Patch

diff --git a/arch/x86/kvm/vmx/main.c b/arch/x86/kvm/vmx/main.c
index c5672909fdae..7258a6304b4b 100644
--- a/arch/x86/kvm/vmx/main.c
+++ b/arch/x86/kvm/vmx/main.c
@@ -252,6 +252,27 @@  static int vt_gmem_max_level(struct kvm *kvm, kvm_pfn_t pfn, gfn_t gfn,
 	return 0;
 }
 
+static int vt_pre_memory_mapping(struct kvm_vcpu *vcpu,
+				 struct kvm_memory_mapping *mapping,
+				 u64 *error_code, u8 *max_level)
+{
+	if (is_td_vcpu(vcpu))
+		return tdx_pre_memory_mapping(vcpu, mapping, error_code, max_level);
+
+	if (mapping->source)
+		return -EINVAL;
+	return 0;
+}
+
+static void vt_post_memory_mapping(struct kvm_vcpu *vcpu,
+				   struct kvm_memory_mapping *mapping)
+{
+	if (!is_td_vcpu(vcpu))
+		return;
+
+	tdx_post_memory_mapping(vcpu, mapping);
+}
+
 #define VMX_REQUIRED_APICV_INHIBITS				\
 	(BIT(APICV_INHIBIT_REASON_DISABLE)|			\
 	 BIT(APICV_INHIBIT_REASON_ABSENT) |			\
@@ -415,6 +436,8 @@  struct kvm_x86_ops vt_x86_ops __initdata = {
 	.vcpu_mem_enc_ioctl = vt_vcpu_mem_enc_ioctl,
 
 	.gmem_max_level = vt_gmem_max_level,
+	.pre_memory_mapping = vt_pre_memory_mapping,
+	.post_memory_mapping = vt_post_memory_mapping,
 };
 
 struct kvm_x86_init_ops vt_init_ops __initdata = {
diff --git a/arch/x86/kvm/vmx/tdx.c b/arch/x86/kvm/vmx/tdx.c
index e65fff43cb1b..8cf6e5dab3e9 100644
--- a/arch/x86/kvm/vmx/tdx.c
+++ b/arch/x86/kvm/vmx/tdx.c
@@ -390,6 +390,7 @@  int tdx_vm_init(struct kvm *kvm)
 	 */
 	kvm->max_vcpus = min(kvm->max_vcpus, TDX_MAX_VCPUS);
 
+	mutex_init(&to_kvm_tdx(kvm)->source_lock);
 	return 0;
 }
 
@@ -541,6 +542,51 @@  static int tdx_mem_page_aug(struct kvm *kvm, gfn_t gfn,
 	return 0;
 }
 
+static int tdx_mem_page_add(struct kvm *kvm, gfn_t gfn,
+			    enum pg_level level, kvm_pfn_t pfn)
+{
+	struct kvm_tdx *kvm_tdx = to_kvm_tdx(kvm);
+	hpa_t hpa = pfn_to_hpa(pfn);
+	gpa_t gpa = gfn_to_gpa(gfn);
+	struct tdx_module_args out;
+	hpa_t source_pa;
+	u64 err;
+
+	lockdep_assert_held(&kvm_tdx->source_lock);
+
+	/*
+	 * KVM_MEMORY_MAPPING for TD supports only 4K page because
+	 * tdh_mem_page_add() supports only 4K page.
+	 */
+	if (KVM_BUG_ON(level != PG_LEVEL_4K, kvm))
+		return -EINVAL;
+
+	if (KVM_BUG_ON(!kvm_tdx->source_page, kvm)) {
+		tdx_unpin(kvm, pfn);
+		return -EINVAL;
+	}
+
+	source_pa = pfn_to_hpa(page_to_pfn(kvm_tdx->source_page));
+	do {
+		err = tdh_mem_page_add(kvm_tdx->tdr_pa, gpa, hpa, source_pa,
+				       &out);
+		/*
+		 * This path is executed during populating initial guest memory
+		 * image. i.e. before running any vcpu.  Race is rare.
+		 */
+	} while (unlikely(err == TDX_ERROR_SEPT_BUSY));
+	/*
+	 * Don't warn: This is for KVM_MEMORY_MAPPING. So tdh_mem_page_add() can
+	 * fail with parameters user provided.
+	 */
+	if (err) {
+		tdx_unpin(kvm, pfn);
+		return -EIO;
+	}
+
+	return 0;
+}
+
 static int tdx_sept_set_private_spte(struct kvm *kvm, gfn_t gfn,
 				     enum pg_level level, kvm_pfn_t pfn)
 {
@@ -563,9 +609,7 @@  static int tdx_sept_set_private_spte(struct kvm *kvm, gfn_t gfn,
 	if (likely(is_td_finalized(kvm_tdx)))
 		return tdx_mem_page_aug(kvm, gfn, level, pfn);
 
-	/* TODO: tdh_mem_page_add() comes here for the initial memory. */
-
-	return 0;
+	return tdx_mem_page_add(kvm, gfn, level, pfn);
 }
 
 static int tdx_sept_drop_private_spte(struct kvm *kvm, gfn_t gfn,
@@ -1469,6 +1513,49 @@  int tdx_gmem_max_level(struct kvm *kvm, kvm_pfn_t pfn, gfn_t gfn,
 	return 0;
 }
 
+#define TDX_SEPT_PFERR	(PFERR_WRITE_MASK | PFERR_GUEST_ENC_MASK)
+
+int tdx_pre_memory_mapping(struct kvm_vcpu *vcpu,
+			   struct kvm_memory_mapping *mapping,
+			   u64 *error_code, u8 *max_level)
+{
+	struct kvm_tdx *kvm_tdx = to_kvm_tdx(vcpu->kvm);
+	struct page *page;
+	int r = 0;
+
+	/* memory contents is needed for encryption. */
+	if (!mapping->source)
+		return -EINVAL;
+
+	/* Once TD is finalized, the initial guest memory is fixed. */
+	if (is_td_finalized(to_kvm_tdx(vcpu->kvm)))
+		return -EINVAL;
+
+	/* TDX supports only 4K to pre-populate. */
+	*max_level = PG_LEVEL_4K;
+	*error_code = TDX_SEPT_PFERR;
+
+	r = get_user_pages_fast(mapping->source, 1, 0, &page);
+	if (r < 0)
+		return r;
+	if (r != 1)
+		return -ENOMEM;
+
+	mutex_lock(&kvm_tdx->source_lock);
+	kvm_tdx->source_page = page;
+	return 0;
+}
+
+void tdx_post_memory_mapping(struct kvm_vcpu *vcpu,
+			     struct kvm_memory_mapping *mapping)
+{
+	struct kvm_tdx *kvm_tdx = to_kvm_tdx(vcpu->kvm);
+
+	put_page(kvm_tdx->source_page);
+	kvm_tdx->source_page = NULL;
+	mutex_unlock(&kvm_tdx->source_lock);
+}
+
 #define TDX_MD_MAP(_fid, _ptr)			\
 	{ .fid = MD_FIELD_ID_##_fid,		\
 	  .ptr = (_ptr), }
diff --git a/arch/x86/kvm/vmx/tdx.h b/arch/x86/kvm/vmx/tdx.h
index 75596b9dcf3f..d822e790e3e5 100644
--- a/arch/x86/kvm/vmx/tdx.h
+++ b/arch/x86/kvm/vmx/tdx.h
@@ -21,6 +21,10 @@  struct kvm_tdx {
 	atomic_t tdh_mem_track;
 
 	u64 tsc_offset;
+
+	/* For KVM_MEMORY_MAPPING */
+	struct mutex source_lock;
+	struct page *source_page;
 };
 
 struct vcpu_tdx {
diff --git a/arch/x86/kvm/vmx/x86_ops.h b/arch/x86/kvm/vmx/x86_ops.h
index 5335d35bc655..191f2964ec8e 100644
--- a/arch/x86/kvm/vmx/x86_ops.h
+++ b/arch/x86/kvm/vmx/x86_ops.h
@@ -160,6 +160,11 @@  int tdx_sept_flush_remote_tlbs(struct kvm *kvm);
 void tdx_load_mmu_pgd(struct kvm_vcpu *vcpu, hpa_t root_hpa, int root_level);
 int tdx_gmem_max_level(struct kvm *kvm, kvm_pfn_t pfn, gfn_t gfn,
 		       bool is_private, u8 *max_level);
+int tdx_pre_memory_mapping(struct kvm_vcpu *vcpu,
+			   struct kvm_memory_mapping *mapping,
+			   u64 *error_code, u8 *max_level);
+void tdx_post_memory_mapping(struct kvm_vcpu *vcpu,
+			     struct kvm_memory_mapping *mapping);
 #else
 static inline int tdx_hardware_setup(struct kvm_x86_ops *x86_ops) { return -EOPNOTSUPP; }
 static inline void tdx_hardware_unsetup(void) {}
@@ -192,6 +197,13 @@  static inline int tdx_gmem_max_level(struct kvm *kvm, kvm_pfn_t pfn, gfn_t gfn,
 {
 	return -EOPNOTSUPP;
 }
+int tdx_pre_memory_mapping(struct kvm_vcpu *vcpu,
+			   struct kvm_memory_mapping *mapping,
+			   u64 *error_code, u8 *max_level)
+{
+	return -EOPNOTSUPP;
+}
+void tdx_post_memory_mapping(struct kvm_vcpu *vcpu, struct kvm_memory_mapping *mapping) {}
 #endif
 
 #endif /* __KVM_X86_VMX_X86_OPS_H */