[Xen-devel] [PATCH 5/6] x86/HVM: prefill cache with PDPTEs when possible

["Jan Beulich" <JBeulich@xxxxxxxx>]

Since strictly speaking it is incorrect for guest_walk_tables() to read
L3 entries during PAE page walks, try to overcome this where possible by
pre-loading the values from hardware into the cache. Sadly the
information is available in the EPT case only. On the positive side for
NPT the spec spells out that L3 entries are actually read on walks, so
us reading them is consistent with hardware behavior in that case.

Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>

--- a/xen/arch/x86/hvm/emulate.c
+++ b/xen/arch/x86/hvm/emulate.c
@@ -2294,6 +2294,23 @@ static int _hvm_emulate_one(struct hvm_e
 
     vio->mmio_retry = 0;
 
+    if ( !curr->arch.hvm_vcpu.data_cache->num_ents &&
+         curr->arch.paging.mode->guest_levels == 3 )
+    {
+        unsigned int i;
+
+        for ( i = 0; i < 4; ++i )
+        {
+            uint64_t pdpte;
+
+            if ( hvm_read_pdpte(curr, i, &pdpte) )
+                hvmemul_write_cache(curr->arch.hvm_vcpu.data_cache,
+                                    (curr->arch.hvm_vcpu.guest_cr[3] &
+                                     (PADDR_MASK & ~0x1f)) + i * sizeof(pdpte),
+                                    3, &pdpte, sizeof(pdpte));
+        }
+    }
+
     rc = x86_emulate(&hvmemul_ctxt->ctxt, ops);
     if ( rc == X86EMUL_OKAY && vio->mmio_retry )
         rc = X86EMUL_RETRY;
--- a/xen/arch/x86/hvm/vmx/vmx.c
+++ b/xen/arch/x86/hvm/vmx/vmx.c
@@ -1361,6 +1361,25 @@ static void vmx_set_interrupt_shadow(str
     __vmwrite(GUEST_INTERRUPTIBILITY_INFO, intr_shadow);
 }
 
+static bool read_pdpte(struct vcpu *v, unsigned int idx, uint64_t *pdpte)
+{
+    if ( !paging_mode_hap(v->domain) || !hvm_pae_enabled(v) ||
+         (v->arch.hvm_vcpu.guest_efer & EFER_LMA) )
+        return false;
+
+    if ( idx >= 4 )
+    {
+        ASSERT_UNREACHABLE();
+        return false;
+    }
+
+    vmx_vmcs_enter(v);
+    __vmread(GUEST_PDPTE(idx), pdpte);
+    vmx_vmcs_exit(v);
+
+    return true;
+}
+
 static bool vmx_load_pdptrs(struct vcpu *v)
 {
     unsigned long cr3 = v->arch.hvm_vcpu.guest_cr[3];
@@ -2464,6 +2483,8 @@ const struct hvm_function_table * __init
         if ( cpu_has_vmx_ept_1gb )
             vmx_function_table.hap_capabilities |= HVM_HAP_SUPERPAGE_1GB;
 
+        vmx_function_table.read_pdpte = read_pdpte;
+
         setup_ept_dump();
     }
 
--- a/xen/include/asm-x86/hvm/hvm.h
+++ b/xen/include/asm-x86/hvm/hvm.h
@@ -146,6 +146,8 @@ struct hvm_function_table {
 
     void (*fpu_leave)(struct vcpu *v);
 
+    bool (*read_pdpte)(struct vcpu *v, unsigned int index, uint64_t *pdpte);
+
     int  (*get_guest_pat)(struct vcpu *v, u64 *);
     int  (*set_guest_pat)(struct vcpu *v, u64);
 
@@ -383,6 +385,12 @@ static inline unsigned long hvm_get_shad
     return hvm_funcs.get_shadow_gs_base(v);
 }
 
+static inline bool hvm_read_pdpte(struct vcpu *v, unsigned int index, uint64_t 
*pdpte)
+{
+    return hvm_funcs.read_pdpte &&
+           hvm_funcs.read_pdpte(v, index, pdpte);
+}
+
 static inline bool hvm_get_guest_bndcfgs(struct vcpu *v, u64 *val)
 {
     return hvm_funcs.get_guest_bndcfgs &&





_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel