[Qemu-devel] [PATCH v2 for-2.0 26/47] qcow2: Don't rely on free_cluster_index in alloc_refcount_block() (CVE-2014-0147)

[Stefan Hajnoczi <stefanha@redhat.com>]

From: Kevin Wolf <kwolf@redhat.com>

free_cluster_index is only correct if update_refcount() was called from
an allocation function, and even there it's brittle because it's used to
protect unfinished allocations which still have a refcount of 0 - if it
moves in the wrong place, the unfinished allocation can be corrupted.

So not using it any more seems to be a good idea. Instead, use the
first requested cluster to do the calculations. Return -EAGAIN if
unfinished allocations could become invalid and let the caller restart
its search for some free clusters.

The context of creating a snapsnot is one situation where
update_refcount() is called outside of a cluster allocation. For this
case, the change fixes a buffer overflow if a cluster is referenced in
an L2 table that cannot be represented by an existing refcount block.
(new_table[refcount_table_index] was out of bounds)

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
---
v2:
 * Fill new refcount block with zeroes when creating image.

   In v1 a dangling refcount table entry was created.  When a qcow2 image is
   created on a block device containing previous data (non-zero), the
   dangling refcount table entry would be read!

   Failure scenario:
   $ qemu-img create -f qcow2 /dev/loop0 30G
   $ qemu-img create -f qcow2 /dev/loop0 30G
   Huh, first cluster in empty image is already in use?

 block/qcow2-refcount.c     | 72 ++++++++++++++++++++++++----------------------
 block/qcow2.c              | 11 +++----
 tests/qemu-iotests/044.out |  2 +-
 tests/qemu-iotests/080     | 11 +++++++
 tests/qemu-iotests/080.out |  7 +++++
 5 files changed, 62 insertions(+), 41 deletions(-)

diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index e3c7ecd..220b322 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -194,10 +194,11 @@ static int alloc_refcount_block(BlockDriverState *bs,
      *   they can describe them themselves.
      *
      * - We need to consider that at this point we are inside update_refcounts
-     *   and doing the initial refcount increase. This means that some clusters
-     *   have already been allocated by the caller, but their refcount isn't
-     *   accurate yet. free_cluster_index tells us where this allocation ends
-     *   as long as we don't overwrite it by freeing clusters.
+     *   and potentially doing an initial refcount increase. This means that
+     *   some clusters have already been allocated by the caller, but their
+     *   refcount isn't accurate yet. If we allocate clusters for metadata, we
+     *   need to return -EAGAIN to signal the caller that it needs to restart
+     *   the search for free clusters.
      *
      * - alloc_clusters_noref and qcow2_free_clusters may load a different
      *   refcount block into the cache
@@ -282,7 +283,10 @@ static int alloc_refcount_block(BlockDriverState *bs,
         }
 
         s->refcount_table[refcount_table_index] = new_block;
-        return 0;
+
+        /* The new refcount block may be where the caller intended to put its
+         * data, so let it restart the search. */
+        return -EAGAIN;
     }
 
     ret = qcow2_cache_put(bs, s->refcount_block_cache, (void**) refcount_block);
@@ -305,8 +309,7 @@ static int alloc_refcount_block(BlockDriverState *bs,
 
     /* Calculate the number of refcount blocks needed so far */
     uint64_t refcount_block_clusters = 1 << (s->cluster_bits - REFCOUNT_SHIFT);
-    uint64_t blocks_used = (s->free_cluster_index +
-        refcount_block_clusters - 1) / refcount_block_clusters;
+    uint64_t blocks_used = DIV_ROUND_UP(cluster_index, refcount_block_clusters);
 
     /* And now we need at least one block more for the new metadata */
     uint64_t table_size = next_refcount_table_size(s, blocks_used + 1);
@@ -339,8 +342,6 @@ static int alloc_refcount_block(BlockDriverState *bs,
     uint16_t *new_blocks = g_malloc0(blocks_clusters * s->cluster_size);
     uint64_t *new_table = g_malloc0(table_size * sizeof(uint64_t));
 
-    assert(meta_offset >= (s->free_cluster_index * s->cluster_size));
-
     /* Fill the new refcount table */
     memcpy(new_table, s->refcount_table,
         s->refcount_table_size * sizeof(uint64_t));
@@ -403,18 +404,19 @@ static int alloc_refcount_block(BlockDriverState *bs,
     s->refcount_table_size = table_size;
     s->refcount_table_offset = table_offset;
 
-    /* Free old table. Remember, we must not change free_cluster_index */
-    uint64_t old_free_cluster_index = s->free_cluster_index;
+    /* Free old table. */
     qcow2_free_clusters(bs, old_table_offset, old_table_size * sizeof(uint64_t),
                         QCOW2_DISCARD_OTHER);
-    s->free_cluster_index = old_free_cluster_index;
 
     ret = load_refcount_block(bs, new_block, (void**) refcount_block);
     if (ret < 0) {
         return ret;
     }
 
-    return 0;
+    /* If we were trying to do the initial refcount update for some cluster
+     * allocation, we might have used the same clusters to store newly
+     * allocated metadata. Make the caller search some new space. */
+    return -EAGAIN;
 
 fail_table:
     g_free(new_table);
@@ -660,12 +662,15 @@ int64_t qcow2_alloc_clusters(BlockDriverState *bs, int64_t size)
     int ret;
 
     BLKDBG_EVENT(bs->file, BLKDBG_CLUSTER_ALLOC);
-    offset = alloc_clusters_noref(bs, size);
-    if (offset < 0) {
-        return offset;
-    }
+    do {
+        offset = alloc_clusters_noref(bs, size);
+        if (offset < 0) {
+            return offset;
+        }
+
+        ret = update_refcount(bs, offset, size, 1, QCOW2_DISCARD_NEVER);
+    } while (ret == -EAGAIN);
 
-    ret = update_refcount(bs, offset, size, 1, QCOW2_DISCARD_NEVER);
     if (ret < 0) {
         return ret;
     }
@@ -678,7 +683,6 @@ int qcow2_alloc_clusters_at(BlockDriverState *bs, uint64_t offset,
 {
     BDRVQcowState *s = bs->opaque;
     uint64_t cluster_index;
-    uint64_t old_free_cluster_index;
     uint64_t i;
     int refcount, ret;
 
@@ -687,30 +691,28 @@ int qcow2_alloc_clusters_at(BlockDriverState *bs, uint64_t offset,
         return 0;
     }
 
-    /* Check how many clusters there are free */
-    cluster_index = offset >> s->cluster_bits;
-    for(i = 0; i < nb_clusters; i++) {
-        refcount = get_refcount(bs, cluster_index++);
+    do {
+        /* Check how many clusters there are free */
+        cluster_index = offset >> s->cluster_bits;
+        for(i = 0; i < nb_clusters; i++) {
+            refcount = get_refcount(bs, cluster_index++);
 
-        if (refcount < 0) {
-            return refcount;
-        } else if (refcount != 0) {
-            break;
+            if (refcount < 0) {
+                return refcount;
+            } else if (refcount != 0) {
+                break;
+            }
         }
-    }
 
-    /* And then allocate them */
-    old_free_cluster_index = s->free_cluster_index;
-    s->free_cluster_index = cluster_index + i;
+        /* And then allocate them */
+        ret = update_refcount(bs, offset, i << s->cluster_bits, 1,
+                              QCOW2_DISCARD_NEVER);
+    } while (ret == -EAGAIN);
 
-    ret = update_refcount(bs, offset, i << s->cluster_bits, 1,
-                          QCOW2_DISCARD_NEVER);
     if (ret < 0) {
         return ret;
     }
 
-    s->free_cluster_index = old_free_cluster_index;
-
     return i;
 }
 
diff --git a/block/qcow2.c b/block/qcow2.c
index ffcb36d..c086266 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -1598,7 +1598,7 @@ static int qcow2_create2(const char *filename, int64_t total_size,
      */
     BlockDriverState* bs;
     QCowHeader *header;
-    uint8_t* refcount_table;
+    uint64_t* refcount_table;
     Error *local_err = NULL;
     int ret;
 
@@ -1650,9 +1650,10 @@ static int qcow2_create2(const char *filename, int64_t total_size,
         goto out;
     }
 
-    /* Write an empty refcount table */
-    refcount_table = g_malloc0(cluster_size);
-    ret = bdrv_pwrite(bs, cluster_size, refcount_table, cluster_size);
+    /* Write a refcount table with one refcount block */
+    refcount_table = g_malloc0(2 * cluster_size);
+    refcount_table[0] = cpu_to_be64(2 * cluster_size);
+    ret = bdrv_pwrite(bs, cluster_size, refcount_table, 2 * cluster_size);
     g_free(refcount_table);
 
     if (ret < 0) {
@@ -1677,7 +1678,7 @@ static int qcow2_create2(const char *filename, int64_t total_size,
         goto out;
     }
 
-    ret = qcow2_alloc_clusters(bs, 2 * cluster_size);
+    ret = qcow2_alloc_clusters(bs, 3 * cluster_size);
     if (ret < 0) {
         error_setg_errno(errp, -ret, "Could not allocate clusters for qcow2 "
                          "header and refcount table");
diff --git a/tests/qemu-iotests/044.out b/tests/qemu-iotests/044.out
index 5c5aa92..4789a53 100644
--- a/tests/qemu-iotests/044.out
+++ b/tests/qemu-iotests/044.out
@@ -1,6 +1,6 @@
 No errors were found on the image.
 7292415/33554432 = 21.73% allocated, 0.00% fragmented, 0.00% compressed clusters
-Image end offset: 4296448000
+Image end offset: 4296152064
 .
 ----------------------------------------------------------------------
 Ran 1 tests
diff --git a/tests/qemu-iotests/080 b/tests/qemu-iotests/080
index f3091a9..56f8903 100755
--- a/tests/qemu-iotests/080
+++ b/tests/qemu-iotests/080
@@ -56,6 +56,8 @@ offset_header_size=100
 offset_ext_magic=$header_size
 offset_ext_size=$((header_size + 4))
 
+offset_l2_table_0=$((0x40000))
+
 echo
 echo "== Huge header size =="
 _make_test_img 64M
@@ -143,6 +145,15 @@ poke_file "$TEST_IMG" "$offset_backing_file_offset" "\x00\x00\x00\x00\x00\x00\x1
 poke_file "$TEST_IMG" "$offset_backing_file_size" "\xff\xff\xff\xff"
 { $QEMU_IO -c "read 0 512" $TEST_IMG; } 2>&1 | _filter_qemu_io | _filter_testdir
 
+echo
+echo "== Invalid L2 entry (huge physical offset) =="
+_make_test_img 64M
+{ $QEMU_IO -c "write 0 512" $TEST_IMG; } 2>&1 | _filter_qemu_io | _filter_testdir
+poke_file "$TEST_IMG" "$offset_l2_table_0" "\xbf\xff\xff\xff\xff\xff\x00\x00"
+{ $QEMU_IMG snapshot -c test $TEST_IMG; } 2>&1 | _filter_qemu_io | _filter_testdir
+poke_file "$TEST_IMG" "$offset_l2_table_0" "\x80\x00\x00\xff\xff\xff\x00\x00"
+{ $QEMU_IMG snapshot -c test $TEST_IMG; } 2>&1 | _filter_qemu_io | _filter_testdir
+
 # success, all done
 echo "*** done"
 rm -f $seq.full
diff --git a/tests/qemu-iotests/080.out b/tests/qemu-iotests/080.out
index 8103211..303d6c3 100644
--- a/tests/qemu-iotests/080.out
+++ b/tests/qemu-iotests/080.out
@@ -63,4 +63,11 @@ no file open, try 'help open'
 Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=67108864 
 qemu-io: can't open device TEST_DIR/t.qcow2: Backing file name too long
 no file open, try 'help open'
+
+== Invalid L2 entry (huge physical offset) ==
+Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=67108864 
+wrote 512/512 bytes at offset 0
+512 bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
+qemu-img: Could not create snapshot 'test': -27 (File too large)
+qemu-img: Could not create snapshot 'test': -11 (Resource temporarily unavailable)
 *** done
-- 
1.8.5.3