gnttab: correct maptrack table accesses

In order to observe a consistent (limit,pointer-table) pair, the reader
needs to either hold the maptrack lock (in line with documentation) or
both sides need to order their accesses suitably (the writer side
barrier was removed by commit dff515dfea ["gnttab: use per-VCPU
maptrack free lists"], and a read side barrier has never been there).

Make the writer publish a new table page before limit (for bounds
checks to work), and new list head last (for racing maptrack_entry()
invocations to work). At the same time add read barriers to lockless
readers.

Additionally get_maptrack_handle() must not assume ->maptrack_head to
not change behind its back: Another handle may be put (updating only
->maptrack_tail) and then got or stolen (updating ->maptrack_head).

This is part of XSA-218.

Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
Gbp-Pq: Name gnttab-correct-maptrack-table-accesses

diff --git a/xen/common/grant_table.c b/xen/common/grant_table.c
index ba10e76b5b09866b7a90d68a29fae48dd18a6f6b..627947a45f74658ea260e5eb6a2a1fd110748c35 100644 (file)
--- a/xen/common/grant_table.c
+++ b/xen/common/grant_table.c
@@ -395,7 +395,7 @@ get_maptrack_handle(
     struct grant_table *lgt)
 {
     struct vcpu          *curr = current;
-    int                   i;
+    unsigned int          i, head;
     grant_handle_t        handle;
     struct grant_mapping *new_mt;
 
@@ -451,17 +451,20 @@ get_maptrack_handle(
         new_mt[i].ref = handle + i + 1;
         new_mt[i].vcpu = curr->vcpu_id;
     }
-    new_mt[i - 1].ref = curr->maptrack_head;
 
     /* Set tail directly if this is the first page for this VCPU. */
     if ( curr->maptrack_tail == MAPTRACK_TAIL )
         curr->maptrack_tail = handle + MAPTRACK_PER_PAGE - 1;
 
-    write_atomic(&curr->maptrack_head, handle + 1);
-
     lgt->maptrack[nr_maptrack_frames(lgt)] = new_mt;
+    smp_wmb();
     lgt->maptrack_limit += MAPTRACK_PER_PAGE;
 
+    do {
+        new_mt[i - 1].ref = read_atomic(&curr->maptrack_head);
+        head = cmpxchg(&curr->maptrack_head, new_mt[i - 1].ref, handle + 1);
+    } while ( head != new_mt[i - 1].ref );
+
     spin_unlock(&lgt->maptrack_lock);
 
     return handle;
@@ -727,6 +730,7 @@ static unsigned int mapkind(
     for ( handle = 0; !(kind & MAPKIND_WRITE) &&
                       handle < lgt->maptrack_limit; handle++ )
     {
+        smp_rmb();
         map = &maptrack_entry(lgt, handle);
         if ( !(map->flags & (GNTMAP_device_map|GNTMAP_host_map)) ||
              map->domid != rd->domain_id )
@@ -1094,6 +1098,7 @@ __gnttab_unmap_common(
         return;
     }
 
+    smp_rmb();
     map = &maptrack_entry(lgt, op->handle);
 
     grant_read_lock(lgt);