[PATCH 1/4] nested vmx: Use a list to store the launched vvmcs for L1 VMM

[Dongxiao Xu <dongxiao.xu@intel.com>]

Originally we use a virtual VMCS field to store the launch state of
a certain vmcs. However if we introduce VMCS shadowing feature, this
virtual VMCS should also be able to load into real hardware,
and VMREAD/VMWRITE operate invalid fields.

The new approach is to store the launch state into a list for L1 VMM.

Signed-off-by: Dongxiao Xu <dongxiao.xu@intel.com>
---
 xen/arch/x86/hvm/vmx/vvmx.c        |   88 ++++++++++++++++++++++++++++++++----
 xen/include/asm-x86/hvm/vmx/vmcs.h |    2 -
 xen/include/asm-x86/hvm/vmx/vvmx.h |    6 +++
 3 files changed, 85 insertions(+), 11 deletions(-)

diff --git a/xen/arch/x86/hvm/vmx/vvmx.c b/xen/arch/x86/hvm/vmx/vvmx.c
index d4e9b02..1c7b1d4 100644
--- a/xen/arch/x86/hvm/vmx/vvmx.c
+++ b/xen/arch/x86/hvm/vmx/vvmx.c
@@ -51,6 +51,7 @@ int nvmx_vcpu_initialise(struct vcpu *v)
     nvmx->iobitmap[0] = NULL;
     nvmx->iobitmap[1] = NULL;
     nvmx->msrbitmap = NULL;
+    INIT_LIST_HEAD(&nvmx->launched_list);
     return 0;
 out:
     return -ENOMEM;
@@ -58,7 +59,9 @@ out:
  
 void nvmx_vcpu_destroy(struct vcpu *v)
 {
+    struct nestedvmx *nvmx = &vcpu_2_nvmx(v);
     struct nestedvcpu *nvcpu = &vcpu_nestedhvm(v);
+    struct vvmcs_list *item, *n;
 
     /* 
      * When destroying the vcpu, it may be running on behalf of L2 guest.
@@ -74,6 +77,11 @@ void nvmx_vcpu_destroy(struct vcpu *v)
         free_xenheap_page(nvcpu->nv_n2vmcx);
         nvcpu->nv_n2vmcx = NULL;
     }
+
+    list_for_each_entry_safe(item, n, &nvmx->launched_list, node) {
+        list_del(&item->node);
+        xfree(item);
+    }
 }
  
 void nvmx_domain_relinquish_resources(struct domain *d)
@@ -1198,6 +1206,59 @@ int nvmx_handle_vmxoff(struct cpu_user_regs *regs)
     return X86EMUL_OKAY;
 }
 
+static int vvmcs_launched(struct list_head *launched_list, paddr_t vvmcs_pa)
+{
+    struct vvmcs_list *vvmcs = NULL;
+    struct list_head *pos;
+    int launched = 0;
+
+    list_for_each(pos, launched_list)
+    {
+        vvmcs = list_entry(pos, struct vvmcs_list, node);
+        if ( vvmcs_pa == vvmcs->vvmcs_pa )
+        {
+            launched = 1;
+            break;
+        }
+    }
+
+    return launched;
+}
+
+static int set_vvmcs_launched(struct list_head *launched_list, paddr_t vvmcs_pa)
+{
+    struct vvmcs_list *vvmcs;
+
+    if ( vvmcs_launched(launched_list, vvmcs_pa) )
+        return 0;
+
+    vvmcs = xzalloc(struct vvmcs_list);
+    if ( !vvmcs )
+        return -ENOMEM;
+
+    vvmcs->vvmcs_pa = vvmcs_pa;
+    list_add(&vvmcs->node, launched_list);
+
+    return 0;
+}
+
+static void clear_vvmcs_launched(struct list_head *launched_list, paddr_t vvmcs_pa)
+{
+    struct vvmcs_list *vvmcs;
+    struct list_head *pos;
+
+    list_for_each(pos, launched_list)
+    {
+        vvmcs = list_entry(pos, struct vvmcs_list, node);
+        if ( vvmcs_pa == vvmcs->vvmcs_pa )
+        {
+            list_del(&vvmcs->node);
+            xfree(vvmcs);
+            break;
+        }
+    }
+}
+
 int nvmx_vmresume(struct vcpu *v, struct cpu_user_regs *regs)
 {
     struct nestedvmx *nvmx = &vcpu_2_nvmx(v);
@@ -1221,8 +1282,10 @@ int nvmx_vmresume(struct vcpu *v, struct cpu_user_regs *regs)
 
 int nvmx_handle_vmresume(struct cpu_user_regs *regs)
 {
-    int launched;
     struct vcpu *v = current;
+    struct nestedvcpu *nvcpu = &vcpu_nestedhvm(v);
+    struct nestedvmx *nvmx = &vcpu_2_nvmx(v);
+    int launched;
 
     if ( vcpu_nestedhvm(v).nv_vvmcxaddr == VMCX_EADDR )
     {
@@ -1230,8 +1293,8 @@ int nvmx_handle_vmresume(struct cpu_user_regs *regs)
         return X86EMUL_OKAY;        
     }
 
-    launched = __get_vvmcs(vcpu_nestedhvm(v).nv_vvmcx,
-                           NVMX_LAUNCH_STATE);
+    launched = vvmcs_launched(&nvmx->launched_list,
+                              virt_to_maddr(nvcpu->nv_vvmcx));
     if ( !launched ) {
        vmreturn (regs, VMFAIL_VALID);
        return X86EMUL_OKAY;
@@ -1244,6 +1307,8 @@ int nvmx_handle_vmlaunch(struct cpu_user_regs *regs)
     int launched;
     int rc;
     struct vcpu *v = current;
+    struct nestedvcpu *nvcpu = &vcpu_nestedhvm(v);
+    struct nestedvmx *nvmx = &vcpu_2_nvmx(v);
 
     if ( vcpu_nestedhvm(v).nv_vvmcxaddr == VMCX_EADDR )
     {
@@ -1251,8 +1316,8 @@ int nvmx_handle_vmlaunch(struct cpu_user_regs *regs)
         return X86EMUL_OKAY;
     }
 
-    launched = __get_vvmcs(vcpu_nestedhvm(v).nv_vvmcx,
-                           NVMX_LAUNCH_STATE);
+    launched = vvmcs_launched(&nvmx->launched_list,
+                              virt_to_maddr(nvcpu->nv_vvmcx));
     if ( launched ) {
        vmreturn (regs, VMFAIL_VALID);
        return X86EMUL_OKAY;
@@ -1260,8 +1325,12 @@ int nvmx_handle_vmlaunch(struct cpu_user_regs *regs)
     else {
         rc = nvmx_vmresume(v,regs);
         if ( rc == X86EMUL_OKAY )
-            __set_vvmcs(vcpu_nestedhvm(v).nv_vvmcx,
-                        NVMX_LAUNCH_STATE, 1);
+        {
+            if ( set_vvmcs_launched(&nvmx->launched_list,
+                                    virt_to_maddr(nvcpu->nv_vvmcx)) < 0 )
+                return X86EMUL_UNHANDLEABLE;
+        }
+          
     }
     return rc;
 }
@@ -1328,6 +1397,7 @@ int nvmx_handle_vmclear(struct cpu_user_regs *regs)
     struct vcpu *v = current;
     struct vmx_inst_decoded decode;
     struct nestedvcpu *nvcpu = &vcpu_nestedhvm(v);
+    struct nestedvmx *nvmx = &vcpu_2_nvmx(v);
     unsigned long gpa = 0;
     void *vvmcs;
     int rc;
@@ -1344,7 +1414,7 @@ int nvmx_handle_vmclear(struct cpu_user_regs *regs)
     
     if ( gpa == nvcpu->nv_vvmcxaddr ) 
     {
-        __set_vvmcs(nvcpu->nv_vvmcx, NVMX_LAUNCH_STATE, 0);
+        clear_vvmcs_launched(&nvmx->launched_list, virt_to_maddr(nvcpu->nv_vvmcx));
         nvmx_purge_vvmcs(v);
     }
     else 
@@ -1352,7 +1422,7 @@ int nvmx_handle_vmclear(struct cpu_user_regs *regs)
         /* Even if this VMCS isn't the current one, we must clear it. */
         vvmcs = hvm_map_guest_frame_rw(gpa >> PAGE_SHIFT);
         if ( vvmcs ) 
-            __set_vvmcs(vvmcs, NVMX_LAUNCH_STATE, 0);
+            clear_vvmcs_launched(&nvmx->launched_list, virt_to_maddr(vvmcs));
         hvm_unmap_guest_frame(vvmcs);
     }
 
diff --git a/xen/include/asm-x86/hvm/vmx/vmcs.h b/xen/include/asm-x86/hvm/vmx/vmcs.h
index 51df81e..9ff741f 100644
--- a/xen/include/asm-x86/hvm/vmx/vmcs.h
+++ b/xen/include/asm-x86/hvm/vmx/vmcs.h
@@ -421,8 +421,6 @@ enum vmcs_field {
     HOST_SYSENTER_EIP               = 0x00006c12,
     HOST_RSP                        = 0x00006c14,
     HOST_RIP                        = 0x00006c16,
-    /* A virtual VMCS field used for nestedvmx only */
-    NVMX_LAUNCH_STATE               = 0x00006c20,
 };
 
 #define VMCS_VPID_WIDTH 16
diff --git a/xen/include/asm-x86/hvm/vmx/vvmx.h b/xen/include/asm-x86/hvm/vmx/vvmx.h
index 9e1dc77..1c5313d 100644
--- a/xen/include/asm-x86/hvm/vmx/vvmx.h
+++ b/xen/include/asm-x86/hvm/vmx/vvmx.h
@@ -23,6 +23,11 @@
 #ifndef __ASM_X86_HVM_VVMX_H__
 #define __ASM_X86_HVM_VVMX_H__
 
+struct vvmcs_list {
+    paddr_t vvmcs_pa;
+    struct list_head node;
+};
+
 struct nestedvmx {
     paddr_t    vmxon_region_pa;
     void       *iobitmap[2];		/* map (va) of L1 guest I/O bitmap */
@@ -38,6 +43,7 @@ struct nestedvmx {
         uint32_t exit_qual;
     } ept;
     uint32_t guest_vpid;
+    struct list_head launched_list;
 };
 
 #define vcpu_2_nvmx(v)	(vcpu_nestedhvm(v).u.nvmx)
-- 
1.7.1