diff mbox series

[v3,2/2] VMX: use a single, global APIC access page

Message ID 774a0bf2-d2a4-7dba-bf15-fec8b0ec8c5f@suse.com (mailing list archive)
State New
Headers show
Series VMX: apic access page handling adjustments | expand

Commit Message

Jan Beulich Feb. 22, 2021, 10:57 a.m. UTC
The address of this page is used by the CPU only to recognize when to
access the virtual APIC page instead. No accesses would ever go to this
page. It only needs to be present in the (CPU) page tables so that
address translation will produce its address as result for respective
accesses.

By making this page global, we also eliminate the need to refcount it,
or to assign it to any domain in the first place.

Signed-off-by: Jan Beulich <jbeulich@suse.com>
---
v3: Split p2m insertion change to a separate patch.
v2: Avoid insertion when !has_vlapic(). Split off change to
    p2m_get_iommu_flags().
---
I did further consider not allocating any real page at all, but just
using the address of some unpopulated space (which would require
announcing this page as reserved to Dom0, so it wouldn't put any PCI
MMIO BARs there). But I thought this would be too controversial, because
of the possible risks associated with this.

Comments

Tian, Kevin March 1, 2021, 2:34 a.m. UTC | #1
> From: Jan Beulich <jbeulich@suse.com>
> Sent: Monday, February 22, 2021 6:57 PM
> 
> The address of this page is used by the CPU only to recognize when to
> access the virtual APIC page instead. No accesses would ever go to this
> page. It only needs to be present in the (CPU) page tables so that
> address translation will produce its address as result for respective
> accesses.
> 
> By making this page global, we also eliminate the need to refcount it,
> or to assign it to any domain in the first place.
> 
> Signed-off-by: Jan Beulich <jbeulich@suse.com>

Reviewed-by: Kevin Tian <kevin.tian@intel.com>

> ---
> v3: Split p2m insertion change to a separate patch.
> v2: Avoid insertion when !has_vlapic(). Split off change to
>     p2m_get_iommu_flags().
> ---
> I did further consider not allocating any real page at all, but just
> using the address of some unpopulated space (which would require
> announcing this page as reserved to Dom0, so it wouldn't put any PCI
> MMIO BARs there). But I thought this would be too controversial, because
> of the possible risks associated with this.
> 
> --- a/xen/arch/x86/hvm/vmx/vmx.c
> +++ b/xen/arch/x86/hvm/vmx/vmx.c
> @@ -66,8 +66,7 @@ boolean_param("force-ept", opt_force_ept
>  static void vmx_ctxt_switch_from(struct vcpu *v);
>  static void vmx_ctxt_switch_to(struct vcpu *v);
> 
> -static int  vmx_alloc_vlapic_mapping(struct domain *d);
> -static void vmx_free_vlapic_mapping(struct domain *d);
> +static int alloc_vlapic_mapping(void);
>  static void vmx_install_vlapic_mapping(struct vcpu *v);
>  static void vmx_update_guest_cr(struct vcpu *v, unsigned int cr,
>                                  unsigned int flags);
> @@ -78,6 +77,8 @@ static int vmx_msr_read_intercept(unsign
>  static int vmx_msr_write_intercept(unsigned int msr, uint64_t msr_content);
>  static void vmx_invlpg(struct vcpu *v, unsigned long linear);
> 
> +static mfn_t __read_mostly apic_access_mfn;
> +
>  /* Values for domain's ->arch.hvm_domain.pi_ops.flags. */
>  #define PI_CSW_FROM (1u << 0)
>  #define PI_CSW_TO   (1u << 1)
> @@ -401,7 +402,6 @@ static int vmx_domain_initialise(struct
>          .to   = vmx_ctxt_switch_to,
>          .tail = vmx_do_resume,
>      };
> -    int rc;
> 
>      d->arch.ctxt_switch = &csw;
> 
> @@ -411,28 +411,14 @@ static int vmx_domain_initialise(struct
>       */
>      d->arch.hvm.vmx.exec_sp = is_hardware_domain(d) || opt_ept_exec_sp;
> 
> -    if ( !has_vlapic(d) )
> -        return 0;
> -
> -    if ( (rc = vmx_alloc_vlapic_mapping(d)) != 0 )
> -        return rc;
> -
>      return 0;
>  }
> 
> -static void vmx_domain_relinquish_resources(struct domain *d)
> -{
> -    if ( !has_vlapic(d) )
> -        return;
> -
> -    vmx_free_vlapic_mapping(d);
> -}
> -
>  static void domain_creation_finished(struct domain *d)
>  {
> -    if ( has_vlapic(d) && !mfn_eq(d->arch.hvm.vmx.apic_access_mfn,
> _mfn(0)) &&
> +    if ( has_vlapic(d) && !mfn_eq(apic_access_mfn, _mfn(0)) &&
>           set_mmio_p2m_entry(d, gaddr_to_gfn(APIC_DEFAULT_PHYS_BASE),
> -                            d->arch.hvm.vmx.apic_access_mfn, PAGE_ORDER_4K) )
> +                            apic_access_mfn, PAGE_ORDER_4K) )
>          domain_crash(d);
>  }
> 
> @@ -2415,7 +2401,6 @@ static struct hvm_function_table __initd
>      .cpu_up_prepare       = vmx_cpu_up_prepare,
>      .cpu_dead             = vmx_cpu_dead,
>      .domain_initialise    = vmx_domain_initialise,
> -    .domain_relinquish_resources = vmx_domain_relinquish_resources,
>      .domain_creation_finished = domain_creation_finished,
>      .vcpu_initialise      = vmx_vcpu_initialise,
>      .vcpu_destroy         = vmx_vcpu_destroy,
> @@ -2662,7 +2647,7 @@ const struct hvm_function_table * __init
>  {
>      set_in_cr4(X86_CR4_VMXE);
> 
> -    if ( vmx_vmcs_init() )
> +    if ( vmx_vmcs_init() || alloc_vlapic_mapping() )
>      {
>          printk("VMX: failed to initialise.\n");
>          return NULL;
> @@ -3217,7 +3202,7 @@ gp_fault:
>      return X86EMUL_EXCEPTION;
>  }
> 
> -static int vmx_alloc_vlapic_mapping(struct domain *d)
> +static int __init alloc_vlapic_mapping(void)
>  {
>      struct page_info *pg;
>      mfn_t mfn;
> @@ -3225,52 +3210,28 @@ static int vmx_alloc_vlapic_mapping(stru
>      if ( !cpu_has_vmx_virtualize_apic_accesses )
>          return 0;
> 
> -    pg = alloc_domheap_page(d, MEMF_no_refcount);
> +    pg = alloc_domheap_page(NULL, 0);
>      if ( !pg )
>          return -ENOMEM;
> 
> -    if ( !get_page_and_type(pg, d, PGT_writable_page) )
> -    {
> -        /*
> -         * The domain can't possibly know about this page yet, so failure
> -         * here is a clear indication of something fishy going on.
> -         */
> -        domain_crash(d);
> -        return -ENODATA;
> -    }
> -
>      mfn = page_to_mfn(pg);
>      clear_domain_page(mfn);
> -    d->arch.hvm.vmx.apic_access_mfn = mfn;
> +    apic_access_mfn = mfn;
> 
>      return 0;
>  }
> 
> -static void vmx_free_vlapic_mapping(struct domain *d)
> -{
> -    mfn_t mfn = d->arch.hvm.vmx.apic_access_mfn;
> -
> -    d->arch.hvm.vmx.apic_access_mfn = _mfn(0);
> -    if ( !mfn_eq(mfn, _mfn(0)) )
> -    {
> -        struct page_info *pg = mfn_to_page(mfn);
> -
> -        put_page_alloc_ref(pg);
> -        put_page_and_type(pg);
> -    }
> -}
> -
>  static void vmx_install_vlapic_mapping(struct vcpu *v)
>  {
>      paddr_t virt_page_ma, apic_page_ma;
> 
> -    if ( mfn_eq(v->domain->arch.hvm.vmx.apic_access_mfn, _mfn(0)) )
> +    if ( !has_vlapic(v->domain) || mfn_eq(apic_access_mfn, _mfn(0)) )
>          return;
> 
>      ASSERT(cpu_has_vmx_virtualize_apic_accesses);
> 
>      virt_page_ma = page_to_maddr(vcpu_vlapic(v)->regs_page);
> -    apic_page_ma = mfn_to_maddr(v->domain-
> >arch.hvm.vmx.apic_access_mfn);
> +    apic_page_ma = mfn_to_maddr(apic_access_mfn);
> 
>      vmx_vmcs_enter(v);
>      __vmwrite(VIRTUAL_APIC_PAGE_ADDR, virt_page_ma);
> --- a/xen/include/asm-x86/hvm/vmx/vmcs.h
> +++ b/xen/include/asm-x86/hvm/vmx/vmcs.h
> @@ -58,7 +58,6 @@ struct ept_data {
>  #define _VMX_DOMAIN_PML_ENABLED    0
>  #define VMX_DOMAIN_PML_ENABLED     (1ul <<
> _VMX_DOMAIN_PML_ENABLED)
>  struct vmx_domain {
> -    mfn_t apic_access_mfn;
>      /* VMX_DOMAIN_* */
>      unsigned int status;
>
Jan Beulich March 1, 2021, 8:18 a.m. UTC | #2
On 01.03.2021 03:34, Tian, Kevin wrote:
>> From: Jan Beulich <jbeulich@suse.com>
>> Sent: Monday, February 22, 2021 6:57 PM
>>
>> The address of this page is used by the CPU only to recognize when to
>> access the virtual APIC page instead. No accesses would ever go to this
>> page. It only needs to be present in the (CPU) page tables so that
>> address translation will produce its address as result for respective
>> accesses.
>>
>> By making this page global, we also eliminate the need to refcount it,
>> or to assign it to any domain in the first place.
>>
>> Signed-off-by: Jan Beulich <jbeulich@suse.com>
> 
> Reviewed-by: Kevin Tian <kevin.tian@intel.com>

Thanks, but for the record: As said on an unrelated thread already,
I also need to make an adjustment to shadow mode code here. Since
that doesn't affect VMX code itself, I'll retain the R-b on v4
(which I intend to submit only once we have 4.15 branched).

Jan
diff mbox series

Patch

--- a/xen/arch/x86/hvm/vmx/vmx.c
+++ b/xen/arch/x86/hvm/vmx/vmx.c
@@ -66,8 +66,7 @@  boolean_param("force-ept", opt_force_ept
 static void vmx_ctxt_switch_from(struct vcpu *v);
 static void vmx_ctxt_switch_to(struct vcpu *v);
 
-static int  vmx_alloc_vlapic_mapping(struct domain *d);
-static void vmx_free_vlapic_mapping(struct domain *d);
+static int alloc_vlapic_mapping(void);
 static void vmx_install_vlapic_mapping(struct vcpu *v);
 static void vmx_update_guest_cr(struct vcpu *v, unsigned int cr,
                                 unsigned int flags);
@@ -78,6 +77,8 @@  static int vmx_msr_read_intercept(unsign
 static int vmx_msr_write_intercept(unsigned int msr, uint64_t msr_content);
 static void vmx_invlpg(struct vcpu *v, unsigned long linear);
 
+static mfn_t __read_mostly apic_access_mfn;
+
 /* Values for domain's ->arch.hvm_domain.pi_ops.flags. */
 #define PI_CSW_FROM (1u << 0)
 #define PI_CSW_TO   (1u << 1)
@@ -401,7 +402,6 @@  static int vmx_domain_initialise(struct
         .to   = vmx_ctxt_switch_to,
         .tail = vmx_do_resume,
     };
-    int rc;
 
     d->arch.ctxt_switch = &csw;
 
@@ -411,28 +411,14 @@  static int vmx_domain_initialise(struct
      */
     d->arch.hvm.vmx.exec_sp = is_hardware_domain(d) || opt_ept_exec_sp;
 
-    if ( !has_vlapic(d) )
-        return 0;
-
-    if ( (rc = vmx_alloc_vlapic_mapping(d)) != 0 )
-        return rc;
-
     return 0;
 }
 
-static void vmx_domain_relinquish_resources(struct domain *d)
-{
-    if ( !has_vlapic(d) )
-        return;
-
-    vmx_free_vlapic_mapping(d);
-}
-
 static void domain_creation_finished(struct domain *d)
 {
-    if ( has_vlapic(d) && !mfn_eq(d->arch.hvm.vmx.apic_access_mfn, _mfn(0)) &&
+    if ( has_vlapic(d) && !mfn_eq(apic_access_mfn, _mfn(0)) &&
          set_mmio_p2m_entry(d, gaddr_to_gfn(APIC_DEFAULT_PHYS_BASE),
-                            d->arch.hvm.vmx.apic_access_mfn, PAGE_ORDER_4K) )
+                            apic_access_mfn, PAGE_ORDER_4K) )
         domain_crash(d);
 }
 
@@ -2415,7 +2401,6 @@  static struct hvm_function_table __initd
     .cpu_up_prepare       = vmx_cpu_up_prepare,
     .cpu_dead             = vmx_cpu_dead,
     .domain_initialise    = vmx_domain_initialise,
-    .domain_relinquish_resources = vmx_domain_relinquish_resources,
     .domain_creation_finished = domain_creation_finished,
     .vcpu_initialise      = vmx_vcpu_initialise,
     .vcpu_destroy         = vmx_vcpu_destroy,
@@ -2662,7 +2647,7 @@  const struct hvm_function_table * __init
 {
     set_in_cr4(X86_CR4_VMXE);
 
-    if ( vmx_vmcs_init() )
+    if ( vmx_vmcs_init() || alloc_vlapic_mapping() )
     {
         printk("VMX: failed to initialise.\n");
         return NULL;
@@ -3217,7 +3202,7 @@  gp_fault:
     return X86EMUL_EXCEPTION;
 }
 
-static int vmx_alloc_vlapic_mapping(struct domain *d)
+static int __init alloc_vlapic_mapping(void)
 {
     struct page_info *pg;
     mfn_t mfn;
@@ -3225,52 +3210,28 @@  static int vmx_alloc_vlapic_mapping(stru
     if ( !cpu_has_vmx_virtualize_apic_accesses )
         return 0;
 
-    pg = alloc_domheap_page(d, MEMF_no_refcount);
+    pg = alloc_domheap_page(NULL, 0);
     if ( !pg )
         return -ENOMEM;
 
-    if ( !get_page_and_type(pg, d, PGT_writable_page) )
-    {
-        /*
-         * The domain can't possibly know about this page yet, so failure
-         * here is a clear indication of something fishy going on.
-         */
-        domain_crash(d);
-        return -ENODATA;
-    }
-
     mfn = page_to_mfn(pg);
     clear_domain_page(mfn);
-    d->arch.hvm.vmx.apic_access_mfn = mfn;
+    apic_access_mfn = mfn;
 
     return 0;
 }
 
-static void vmx_free_vlapic_mapping(struct domain *d)
-{
-    mfn_t mfn = d->arch.hvm.vmx.apic_access_mfn;
-
-    d->arch.hvm.vmx.apic_access_mfn = _mfn(0);
-    if ( !mfn_eq(mfn, _mfn(0)) )
-    {
-        struct page_info *pg = mfn_to_page(mfn);
-
-        put_page_alloc_ref(pg);
-        put_page_and_type(pg);
-    }
-}
-
 static void vmx_install_vlapic_mapping(struct vcpu *v)
 {
     paddr_t virt_page_ma, apic_page_ma;
 
-    if ( mfn_eq(v->domain->arch.hvm.vmx.apic_access_mfn, _mfn(0)) )
+    if ( !has_vlapic(v->domain) || mfn_eq(apic_access_mfn, _mfn(0)) )
         return;
 
     ASSERT(cpu_has_vmx_virtualize_apic_accesses);
 
     virt_page_ma = page_to_maddr(vcpu_vlapic(v)->regs_page);
-    apic_page_ma = mfn_to_maddr(v->domain->arch.hvm.vmx.apic_access_mfn);
+    apic_page_ma = mfn_to_maddr(apic_access_mfn);
 
     vmx_vmcs_enter(v);
     __vmwrite(VIRTUAL_APIC_PAGE_ADDR, virt_page_ma);
--- a/xen/include/asm-x86/hvm/vmx/vmcs.h
+++ b/xen/include/asm-x86/hvm/vmx/vmcs.h
@@ -58,7 +58,6 @@  struct ept_data {
 #define _VMX_DOMAIN_PML_ENABLED    0
 #define VMX_DOMAIN_PML_ENABLED     (1ul << _VMX_DOMAIN_PML_ENABLED)
 struct vmx_domain {
-    mfn_t apic_access_mfn;
     /* VMX_DOMAIN_* */
     unsigned int status;