From: Roger Pau Monné <roger.pau@citrix.com>
Subject: x86/p2m: free the paging memory pool preemptively

The paging memory pool is currently freed in two different places:
from {shadow,hap}_teardown() via domain_relinquish_resources() and
from {shadow,hap}_final_teardown() via complete_domain_destroy().
While the former does handle preemption, the later doesn't.

Attempt to move as much p2m related freeing as possible to happen
before the call to {shadow,hap}_teardown(), so that most memory can be
freed in a preemptive way.  In order to avoid causing issues to
existing callers leave the root p2m page tables set and free them in
{hap,shadow}_final_teardown().  Also modify {hap,shadow}_free to free
the page immediately if the domain is dying, so that pages don't
accumulate in the pool when {shadow,hap}_final_teardown() get called.

Move altp2m_vcpu_disable_ve() to be done in hap_teardown(), as that's
the place where altp2m_active gets disabled now.

This is part of CVE-2022-33746 / XSA-410.

Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Tim Deegan <tim@xen.org>

--- a/xen/arch/x86/domain.c
+++ b/xen/arch/x86/domain.c
@@ -38,7 +38,6 @@
 #include <xen/livepatch.h>
 #include <public/sysctl.h>
 #include <public/hvm/hvm_vcpu.h>
-#include <asm/altp2m.h>
 #include <asm/regs.h>
 #include <asm/mc146818rtc.h>
 #include <asm/system.h>
@@ -2381,12 +2380,6 @@ int domain_relinquish_resources(struct d
             vpmu_destroy(v);
         }
 
-        if ( altp2m_active(d) )
-        {
-            for_each_vcpu ( d, v )
-                altp2m_vcpu_disable_ve(v);
-        }
-
         if ( is_pv_domain(d) )
         {
             for_each_vcpu ( d, v )
--- a/xen/arch/x86/mm/hap/hap.c
+++ b/xen/arch/x86/mm/hap/hap.c
@@ -28,6 +28,7 @@
 #include <xen/domain_page.h>
 #include <xen/guest_access.h>
 #include <xen/keyhandler.h>
+#include <asm/altp2m.h>
 #include <asm/event.h>
 #include <asm/page.h>
 #include <asm/current.h>
@@ -546,24 +547,8 @@ void hap_final_teardown(struct domain *d
     unsigned int i;
 
     if ( hvm_altp2m_supported() )
-    {
-        d->arch.altp2m_active = 0;
-
-        if ( d->arch.altp2m_eptp )
-        {
-            free_xenheap_page(d->arch.altp2m_eptp);
-            d->arch.altp2m_eptp = NULL;
-        }
-
-        if ( d->arch.altp2m_visible_eptp )
-        {
-            free_xenheap_page(d->arch.altp2m_visible_eptp);
-            d->arch.altp2m_visible_eptp = NULL;
-        }
-
         for ( i = 0; i < MAX_ALTP2M; i++ )
             p2m_teardown(d->arch.altp2m_p2m[i], true);
-    }
 
     /* Destroy nestedp2m's first */
     for (i = 0; i < MAX_NESTEDP2M; i++) {
@@ -578,6 +563,8 @@ void hap_final_teardown(struct domain *d
     paging_lock(d);
     hap_set_allocation(d, 0, NULL);
     ASSERT(d->arch.paging.hap.p2m_pages == 0);
+    ASSERT(d->arch.paging.hap.free_pages == 0);
+    ASSERT(d->arch.paging.hap.total_pages == 0);
     paging_unlock(d);
 }
 
@@ -603,6 +590,7 @@ void hap_vcpu_teardown(struct vcpu *v)
 void hap_teardown(struct domain *d, bool *preempted)
 {
     struct vcpu *v;
+    unsigned int i;
 
     ASSERT(d->is_dying);
     ASSERT(d != current->domain);
@@ -611,6 +599,28 @@ void hap_teardown(struct domain *d, bool
     for_each_vcpu ( d, v )
         hap_vcpu_teardown(v);
 
+    /* Leave the root pt in case we get further attempts to modify the p2m. */
+    if ( hvm_altp2m_supported() )
+    {
+        if ( altp2m_active(d) )
+            for_each_vcpu ( d, v )
+                altp2m_vcpu_disable_ve(v);
+
+        d->arch.altp2m_active = 0;
+
+        FREE_XENHEAP_PAGE(d->arch.altp2m_eptp);
+        FREE_XENHEAP_PAGE(d->arch.altp2m_visible_eptp);
+
+        for ( i = 0; i < MAX_ALTP2M; i++ )
+            p2m_teardown(d->arch.altp2m_p2m[i], false);
+    }
+
+    /* Destroy nestedp2m's after altp2m. */
+    for ( i = 0; i < MAX_NESTEDP2M; i++ )
+        p2m_teardown(d->arch.nested_p2m[i], false);
+
+    p2m_teardown(p2m_get_hostp2m(d), false);
+
     paging_lock(d); /* Keep various asserts happy */
 
     if ( d->arch.paging.hap.total_pages != 0 )
--- a/xen/arch/x86/mm/shadow/common.c
+++ b/xen/arch/x86/mm/shadow/common.c
@@ -2824,8 +2824,17 @@ void shadow_teardown(struct domain *d, b
     for_each_vcpu ( d, v )
         shadow_vcpu_teardown(v);
 
+    p2m_teardown(p2m_get_hostp2m(d), false);
+
     paging_lock(d);
 
+    /*
+     * Reclaim all shadow memory so that shadow_set_allocation() doesn't find
+     * in-use pages, as _shadow_prealloc() will no longer try to reclaim pages
+     * because the domain is dying.
+     */
+    shadow_blow_tables(d);
+
 #if (SHADOW_OPTIMIZATIONS & (SHOPT_VIRTUAL_TLB|SHOPT_OUT_OF_SYNC))
     /* Free the virtual-TLB array attached to each vcpu */
     for_each_vcpu(d, v)
@@ -2946,6 +2955,9 @@ void shadow_final_teardown(struct domain
                    d->arch.paging.shadow.total_pages,
                    d->arch.paging.shadow.free_pages,
                    d->arch.paging.shadow.p2m_pages);
+    ASSERT(!d->arch.paging.shadow.total_pages);
+    ASSERT(!d->arch.paging.shadow.free_pages);
+    ASSERT(!d->arch.paging.shadow.p2m_pages);
     paging_unlock(d);
 }