Nested Virtualization: p2m phystable

Signed-off-by Christoph Egger <[email protected]>

# HG changeset patch
# User cegger
# Date 1271330296 -7200
Move phys_table from struct domain to struct p2m_domain. This prepares p2m code to deal with multiple p2m tables per-domain. Multiple p2m tables are needed to use hap with nested virtualization.

diff -r 96efed7dedaf -r 96cbc9586732 xen/arch/x86/hvm/svm/svm.c

xen/arch/x86/hvm/svm/svm.c +2 -1

@@ -212,6 +212,7 @@ static int svm_vmcb_restore(struct vcpu 
     unsigned long mfn = 0;
     p2m_type_t p2mt;
     struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
+    struct p2m_domain *p2m = p2m_get_hostp2m(v->domain);
 
     if ( c->pending_valid &&
          ((c->pending_type == 1) || (c->pending_type > 6) ||

@@ -257,7 +258,7 @@ static int svm_vmcb_restore(struct vcpu 
     {
         vmcb->np_enable = 1;
         vmcb->g_pat = MSR_IA32_CR_PAT_RESET; /* guest PAT */
-        vmcb->h_cr3 = pagetable_get_paddr(v->domain->arch.phys_table);
+        vmcb->h_cr3 = pagetable_get_paddr(p2m_get_pagetable(p2m));
     }
 
     if ( c->pending_valid ) 

diff -r 96efed7dedaf -r 96cbc9586732 xen/arch/x86/hvm/svm/vmcb.c

xen/arch/x86/hvm/svm/vmcb.c +2 -2

@@ -26,7 +26,7 @@
 #include <asm/cpufeature.h>
 #include <asm/processor.h>
 #include <asm/msr.h>
-#include <asm/paging.h>
+#include <asm/p2m.h>
 #include <asm/hvm/hvm.h>
 #include <asm/hvm/io.h>
 #include <asm/hvm/support.h>

@@ -232,7 +232,7 @@ static int construct_vmcb(struct vcpu *v
     {
         vmcb->np_enable = 1; /* enable nested paging */
         vmcb->g_pat = MSR_IA32_CR_PAT_RESET; /* guest PAT */
-        vmcb->h_cr3 = pagetable_get_paddr(v->domain->arch.phys_table);
+        vmcb->h_cr3 = pagetable_get_paddr(p2m_get_pagetable(p2m_get_hostp2m(v->domain)));
 
         /* No point in intercepting CR3 reads/writes. */
         vmcb->cr_intercepts &= ~(CR_INTERCEPT_CR3_READ|CR_INTERCEPT_CR3_WRITE);

diff -r 96efed7dedaf -r 96cbc9586732 xen/arch/x86/hvm/vmx/vmx.c

xen/arch/x86/hvm/vmx/vmx.c +1 -1

@@ -81,7 +81,7 @@ static int vmx_domain_initialise(struct 
     d->arch.hvm_domain.vmx.ept_control.etmt = EPT_DEFAULT_MT;
     d->arch.hvm_domain.vmx.ept_control.gaw  = EPT_DEFAULT_GAW;
     d->arch.hvm_domain.vmx.ept_control.asr  =
-        pagetable_get_pfn(d->arch.phys_table);
+        pagetable_get_pfn(p2m_get_pagetable(p2m_get_hostp2m(d)));
 
 
     if ( (rc = vmx_alloc_vlapic_mapping(d)) != 0 )

diff -r 96efed7dedaf -r 96cbc9586732 xen/arch/x86/mm/hap/hap.c

xen/arch/x86/mm/hap/hap.c +4 -3

@@ -410,7 +410,7 @@ static void hap_install_xen_entries_in_l
 
     /* Install the domain-specific P2M table */
     l4e[l4_table_offset(RO_MPT_VIRT_START)] =
-        l4e_from_pfn(mfn_x(pagetable_get_mfn(d->arch.phys_table)),
+        l4e_from_pfn(mfn_x(pagetable_get_mfn(p2m_get_pagetable(p2m_get_hostp2m(d)))),
                      __PAGE_HYPERVISOR);
 
     hap_unmap_domain_page(l4e);

@@ -421,6 +421,7 @@ static void hap_install_xen_entries_in_l
 static void hap_install_xen_entries_in_l2h(struct vcpu *v, mfn_t l2hmfn)
 {
     struct domain *d = v->domain;
+    struct p2m_domain *hostp2m = p2m_get_hostp2m(d);
     l2_pgentry_t *l2e;
     l3_pgentry_t *p2m;
     int i;

@@ -446,8 +447,8 @@ static void hap_install_xen_entries_in_l
             l2e_empty();
 
     /* Install the domain-specific p2m table */
-    ASSERT(pagetable_get_pfn(d->arch.phys_table) != 0);
-    p2m = hap_map_domain_page(pagetable_get_mfn(d->arch.phys_table));
+    ASSERT(pagetable_get_pfn(p2m_get_pagetable(hostp2m)) != 0);
+    p2m = hap_map_domain_page(pagetable_get_mfn(p2m_get_pagetable(hostp2m)));
     for ( i = 0; i < MACHPHYS_MBYTES>>1; i++ )
     {
         l2e[l2_table_offset(RO_MPT_VIRT_START) + i] =

diff -r 96efed7dedaf -r 96cbc9586732 xen/arch/x86/mm/hap/p2m-ept.c

xen/arch/x86/mm/hap/p2m-ept.c +7 -6

@@ -240,12 +240,13 @@ ept_set_entry(struct domain *d, unsigned
     int direct_mmio = (p2mt == p2m_mmio_direct);
     uint8_t ipat = 0;
     int need_modify_vtd_table = 1;
+    struct p2m_domain *p2m = p2m_get_hostp2m(d);
 
     if (  order != 0 )
         if ( (gfn & ((1UL << order) - 1)) )
             return 1;
 
-    table = map_domain_page(mfn_x(pagetable_get_mfn(d->arch.phys_table)));
+    table = map_domain_page(mfn_x(pagetable_get_mfn(p2m_get_pagetable(p2m))));
 
     ASSERT(table != NULL);
 

@@ -368,7 +369,7 @@ static mfn_t ept_get_entry(struct domain
                            p2m_query_t q)
 {
     ept_entry_t *table =
-        map_domain_page(mfn_x(pagetable_get_mfn(d->arch.phys_table)));
+        map_domain_page(mfn_x(pagetable_get_mfn(p2m_get_pagetable(p2m_get_hostp2m(d)))));
     unsigned long gfn_remainder = gfn;
     ept_entry_t *ept_entry;
     u32 index;

@@ -462,7 +463,7 @@ out:
 static ept_entry_t ept_get_entry_content(struct domain *d, unsigned long gfn, int *level)
 {
     ept_entry_t *table =
-        map_domain_page(mfn_x(pagetable_get_mfn(d->arch.phys_table)));
+        map_domain_page(mfn_x(pagetable_get_mfn(p2m_get_pagetable(p2m_get_hostp2m(d)))));
     unsigned long gfn_remainder = gfn;
     ept_entry_t *ept_entry;
     ept_entry_t content = { .epte = 0 };

@@ -497,7 +498,7 @@ static ept_entry_t ept_get_entry_content
 void ept_walk_table(struct domain *d, unsigned long gfn)
 {
     ept_entry_t *table =
-        map_domain_page(mfn_x(pagetable_get_mfn(d->arch.phys_table)));
+        map_domain_page(mfn_x(pagetable_get_mfn(p2m_get_pagetable(p2m_get_hostp2m(d)))));
     unsigned long gfn_remainder = gfn;
 
     int i;

@@ -631,12 +632,12 @@ static void ept_change_entry_type_global
     int i2;
     int i1;
 
-    if ( pagetable_get_pfn(d->arch.phys_table) == 0 )
+    if ( pagetable_get_pfn(p2m_get_pagetable(p2m_get_hostp2m(d))) == 0 )
         return;
 
     BUG_ON(EPT_DEFAULT_GAW != 3);
 
-    l4e = map_domain_page(mfn_x(pagetable_get_mfn(d->arch.phys_table)));
+    l4e = map_domain_page(mfn_x(pagetable_get_mfn(p2m_get_pagetable(p2m_get_hostp2m(d)))));
     for (i4 = 0; i4 < EPT_PAGETABLE_ENTRIES; i4++ )
     {
         if ( !l4e[i4].epte )

diff -r 96efed7dedaf -r 96cbc9586732 xen/arch/x86/mm/p2m.c

xen/arch/x86/mm/p2m.c +15 -14

@@ -1253,7 +1253,7 @@ p2m_set_entry(struct domain *d, unsigned
               unsigned int page_order, p2m_type_t p2mt)
 {
     // XXX -- this might be able to be faster iff current->domain == d
-    mfn_t table_mfn = pagetable_get_mfn(d->arch.phys_table);
+    mfn_t table_mfn = pagetable_get_mfn(p2m_get_pagetable(p2m_get_hostp2m(d)));
     void *table =map_domain_page(mfn_x(table_mfn));
     unsigned long i, gfn_remainder = gfn;
     l1_pgentry_t *p2m_entry;

@@ -1408,7 +1408,7 @@ p2m_gfn_to_mfn(struct domain *d, unsigne
      * XXX we will return p2m_invalid for unmapped gfns */
     *t = p2m_mmio_dm;
 
-    mfn = pagetable_get_mfn(d->arch.phys_table);
+    mfn = pagetable_get_mfn(p2m_get_pagetable(p2m_get_hostp2m(d)));
 
     if ( gfn > d->arch.p2m->max_mapped_pfn )
         /* This pfn is higher than the highest the p2m map currently holds */

@@ -1622,11 +1622,11 @@ int p2m_alloc_table(struct domain *d,
     struct page_info *page, *p2m_top;
     unsigned int page_count = 0;
     unsigned long gfn = -1UL;
-    struct p2m_domain *p2m = d->arch.p2m;
+    struct p2m_domain *p2m = p2m_get_hostp2m(d);
 
     p2m_lock(p2m);
 
-    if ( pagetable_get_pfn(d->arch.phys_table) != 0 )
+    if ( pagetable_get_pfn(p2m_get_pagetable(p2m)) != 0 )
     {
         P2M_ERROR("p2m already allocated for this domain\n");
         p2m_unlock(p2m);

@@ -1655,7 +1655,7 @@ int p2m_alloc_table(struct domain *d,
 #endif
         | 1 | PGT_validated;
 
-    d->arch.phys_table = pagetable_from_mfn(page_to_mfn(p2m_top));
+    p2m->phys_table = pagetable_from_mfn(page_to_mfn(p2m_top));
 
     P2M_PRINTK("populating p2m table\n");
 

@@ -1699,7 +1699,7 @@ void p2m_teardown(struct domain *d)
  * We know we don't have any extra mappings to these pages */
 {
     struct page_info *pg;
-    struct p2m_domain *p2m = d->arch.p2m;
+    struct p2m_domain *p2m = p2m_get_hostp2m(d);
     unsigned long gfn;
     p2m_type_t t;
     mfn_t mfn;

@@ -1711,7 +1711,7 @@ void p2m_teardown(struct domain *d)
         if(mfn_valid(mfn) && (t == p2m_ram_shared))
             BUG_ON(mem_sharing_unshare_page(d, gfn, MEM_SHARING_DESTROY_GFN));
     }
-    d->arch.phys_table = pagetable_null();
+    p2m->phys_table = pagetable_null();
 
     while ( (pg = page_list_remove_head(&p2m->pages)) )
         p2m->free_page(d, pg);

@@ -1822,7 +1822,7 @@ static void audit_p2m(struct domain *d)
     spin_unlock(&d->page_alloc_lock);
 
     /* Audit part two: walk the domain's p2m table, checking the entries. */
-    if ( pagetable_get_pfn(d->arch.phys_table) != 0 )
+    if ( pagetable_get_pfn(p2m_get_pagetable(p2m_get_hostp2m(d)) != 0 )
     {
         l2_pgentry_t *l2e;
         l1_pgentry_t *l1e;

@@ -1832,11 +1832,11 @@ static void audit_p2m(struct domain *d)
         l4_pgentry_t *l4e;
         l3_pgentry_t *l3e;
         int i3, i4;
-        l4e = map_domain_page(mfn_x(pagetable_get_mfn(d->arch.phys_table)));
+        l4e = map_domain_page(mfn_x(pagetable_get_mfn(p2m_get_pagetable(p2m_get_hostp2m(d)))));
 #else /* CONFIG_PAGING_LEVELS == 3 */
         l3_pgentry_t *l3e;
         int i3;
-        l3e = map_domain_page(mfn_x(pagetable_get_mfn(d->arch.phys_table)));
+        l3e = map_domain_page(mfn_x(pagetable_get_mfn(p2m_get_pagetable(p2m_get_hostp2m(d)))));
 #endif
 
         gfn = 0;

@@ -2248,22 +2248,23 @@ void p2m_change_type_global(struct domai
     l4_pgentry_t *l4e;
     unsigned long i4;
 #endif /* CONFIG_PAGING_LEVELS == 4 */
+    struct p2m_domain *p2m = p2m_get_hostp2m(d);
 
     BUG_ON(p2m_is_grant(ot) || p2m_is_grant(nt));
 
     if ( !paging_mode_translate(d) )
         return;
 
-    if ( pagetable_get_pfn(d->arch.phys_table) == 0 )
+    if ( pagetable_get_pfn(p2m_get_pagetable(p2m)) == 0 )
         return;
 
     ASSERT(p2m_locked_by_me(d->arch.p2m));
 
 #if CONFIG_PAGING_LEVELS == 4
-    l4e = map_domain_page(mfn_x(pagetable_get_mfn(d->arch.phys_table)));
+    l4e = map_domain_page(mfn_x(pagetable_get_mfn(p2m_get_pagetable(p2m))));
 #else /* CONFIG_PAGING_LEVELS == 3 */
-    l3mfn = _mfn(mfn_x(pagetable_get_mfn(d->arch.phys_table)));
-    l3e = map_domain_page(mfn_x(pagetable_get_mfn(d->arch.phys_table)));
+    l3mfn = _mfn(mfn_x(pagetable_get_mfn(p2m_get_pagetable(p2m))));
+    l3e = map_domain_page(mfn_x(pagetable_get_mfn(p2m_get_pagetable(p2m))));
 #endif
 
 #if CONFIG_PAGING_LEVELS >= 4

diff -r 96efed7dedaf -r 96cbc9586732 xen/arch/x86/mm/shadow/common.c

xen/arch/x86/mm/shadow/common.c +1 -1

@@ -3177,7 +3177,7 @@ int shadow_enable(struct domain *d, u32 
  out_locked:
     shadow_unlock(d);
  out_unlocked:
-    if ( rv != 0 && !pagetable_is_null(d->arch.phys_table) )
+    if ( rv != 0 && !pagetable_is_null(p2m_get_pagetable(p2m_get_hostp2m(d))) )
         p2m_teardown(d);
     if ( rv != 0 && pg != NULL )
         shadow_free_p2m_page(d, pg);

diff -r 96efed7dedaf -r 96cbc9586732 xen/arch/x86/mm/shadow/multi.c

xen/arch/x86/mm/shadow/multi.c +3 -3

@@ -1473,7 +1473,7 @@ void sh_install_xen_entries_in_l4(struct
     {
         /* install domain-specific P2M table */
         sl4e[shadow_l4_table_offset(RO_MPT_VIRT_START)] =
-            shadow_l4e_from_mfn(pagetable_get_mfn(d->arch.phys_table),
+            shadow_l4e_from_mfn(pagetable_get_mfn(p2m_get_pagetable(p2m_get_hostp2m(d))),
                                 __PAGE_HYPERVISOR);
     }
 

@@ -1532,8 +1532,8 @@ static void sh_install_xen_entries_in_l2
     {
         /* Install the domain-specific p2m table */
         l3_pgentry_t *p2m;
-        ASSERT(pagetable_get_pfn(d->arch.phys_table) != 0);
-        p2m = sh_map_domain_page(pagetable_get_mfn(d->arch.phys_table));
+        ASSERT(pagetable_get_pfn(p2m_get_pagetable(p2m_get_hostp2m(d))) != 0);
+        p2m = sh_map_domain_page(pagetable_get_mfn(p2m_get_pagetable(p2m_get_hostp2m(d))));
         for ( i = 0; i < MACHPHYS_MBYTES>>1; i++ )
         {
             sl2e[shadow_l2_table_offset(RO_MPT_VIRT_START) + i] =

diff -r 96efed7dedaf -r 96cbc9586732 xen/include/asm-x86/domain.h

xen/include/asm-x86/domain.h +0 -3

@@ -272,9 +272,6 @@ struct arch_domain
     struct paging_domain paging;
     struct p2m_domain *p2m;
 
-    /* Shadow translated domain: P2M mapping */
-    pagetable_t phys_table;
-
     /* NB. protected by d->event_lock and by irq_desc[irq].lock */
     int *irq_pirq;
     int *pirq_irq;

diff -r 96efed7dedaf -r 96cbc9586732 xen/include/asm-x86/p2m.h

xen/include/asm-x86/p2m.h +9 -0

@@ -29,6 +29,7 @@
 #include <xen/config.h>
 #include <xen/paging.h>
 #include <asm/mem_sharing.h>
+#include <asm/page.h>    /* for pagetable_t */
 
 /*
  * The phys_to_machine_mapping maps guest physical frame numbers 

@@ -166,6 +167,9 @@ struct p2m_domain {
     int                locker;   /* processor which holds the lock */
     const char        *locker_function; /* Func that took it */
 
+    /* Shadow translated domain: p2m mapping */
+    pagetable_t        phys_table;
+
     /* Pages used to construct the p2m */
     struct page_list_head pages;
 

@@ -212,6 +216,11 @@ struct p2m_domain {
     } pod;
 };
 
+/* get host p2m table */
+#define p2m_get_hostp2m(d)      ((d)->arch.p2m)
+
+#define p2m_get_pagetable(p2m)  ((p2m)->phys_table)
+
 /*
  * The P2M lock.  This protects all updates to the p2m table.
  * Updates are expected to be safe against concurrent reads,