x86: avoid EPT scanning errors when splitting superpages during live migration

Since Xen did not lock the p2m table for p2m table reading, when
splitting the large page during live migration, we should make sure
the path of EPT entries be modified are always there while other CPUs
may access the super entries at the same time.

Signed-off-by: Xin, Xiaohui <xiaohui.xin@intel.com>

diff --git a/xen/arch/x86/mm/hap/p2m-ept.c b/xen/arch/x86/mm/hap/p2m-ept.c
index bb1e8ee643a986b691bffae1ff3ec82b1cd365af..25311321879cb058906cf7224d8161dbbaa0adda 100644 (file)
--- a/xen/arch/x86/mm/hap/p2m-ept.c
+++ b/xen/arch/x86/mm/hap/p2m-ept.c
@@ -210,17 +210,18 @@ ept_set_entry(struct domain *d, unsigned long gfn, mfn_t mfn,
         ept_entry_t *split_ept_entry = NULL;
         unsigned long split_mfn = ept_entry->mfn;
         p2m_type_t split_p2mt = ept_entry->avail1;
+        ept_entry_t new_ept_entry;
 
         /* alloc new page for new ept middle level entry which is
          * before a leaf super entry
          */
 
-        if ( !ept_set_middle_entry(d, ept_entry) )
+        if ( !ept_set_middle_entry(d, &new_ept_entry) )
             goto out;
 
         /* split the super page before to 4k pages */
 
-        split_table = map_domain_page(ept_entry->mfn);
+        split_table = map_domain_page(new_ept_entry.mfn);
         offset = gfn & ((1 << EPT_TABLE_ORDER) - 1);
 
         for ( i = 0; i < 512; i++ )
@@ -257,6 +258,7 @@ ept_set_entry(struct domain *d, unsigned long gfn, mfn_t mfn,
         ept_p2m_type_to_flags(split_ept_entry, p2mt);
 
         unmap_domain_page(split_table);
+        *ept_entry = new_ept_entry;
     }
 
     /* Track the highest gfn for which we have ever had a valid mapping */