[Qemu-devel] [PATCH 2/5] qcow2: Metadata overlap checks

[Max Reitz]

Two new functions are added; the first one checks a given range in the
image file for overlaps with metadata (main header, L1 tables, L2
tables, refcount table and blocks).

The second one should be used immediately before writing to the image
file as it calls the first function and, upon collision, marks the
image as corrupt and makes the BDS unusable, thereby preventing
further access.

Both functions take a bitmask argument specifying the structures which
should be checked for overlaps, making it possible to also check
metadata writes against colliding with other structures.

Signed-off-by: Max Reitz <address@hidden>
---
 block/qcow2-refcount.c | 142 +++++++++++++++++++++++++++++++++++++++++++++++++
 block/qcow2.h          |  28 ++++++++++
 2 files changed, 170 insertions(+)

diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index 1244693..c8141c8 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -25,6 +25,7 @@
 #include "qemu-common.h"
 #include "block/block_int.h"
 #include "block/qcow2.h"
+#include "qemu/range.h"
 
 static int64_t alloc_clusters_noref(BlockDriverState *bs, int64_t size);
 static int QEMU_WARN_UNUSED_RESULT update_refcount(BlockDriverState *bs,
@@ -1372,3 +1373,144 @@ fail:
     return ret;
 }
 
+/*
+ * Checks if the given offset into the image file is actually free to use by
+ * looking for overlaps with important metadata sections (L1/L2 tables etc.),
+ * i.e. a sanity check without relying on the refcount tables.
+ *
+ * The chk parameter specifies exactly what checks to perform.
+ *
+ * Returns:
+ * - 0 if writing to this offset will not affect the mentioned metadata
+ * - a positive QCow2MetadataOverlap value indicating one overlapping section
+ * - a negative value (-errno) indicating an error while performing a check,
+ *   e.g. when bdrv_read failed on QCOW2_OL_INACTIVE_L2
+ */
+int qcow2_check_metadata_overlap(BlockDriverState *bs, QCow2MetadataOverlap 
chk,
+                                 int64_t offset, int64_t size)
+{
+    BDRVQcowState *s = bs->opaque;
+    int i, j;
+
+    if (!size) {
+        return 0;
+    }
+
+    if (chk & QCOW2_OL_MAIN_HEADER) {
+        if (offset < s->cluster_size) {
+            return QCOW2_OL_MAIN_HEADER;
+        }
+    }
+
+    if ((chk & QCOW2_OL_ACTIVE_L1) && s->l1_size) {
+        if (ranges_overlap(offset, size, s->l1_table_offset,
+            s->l1_size * sizeof(uint64_t))) {
+            return QCOW2_OL_ACTIVE_L1;
+        }
+    }
+
+    if ((chk & QCOW2_OL_REFCOUNT_TABLE) && s->refcount_table_size) {
+        if (ranges_overlap(offset, size, s->refcount_table_offset,
+            s->refcount_table_size * sizeof(uint64_t))) {
+            return QCOW2_OL_REFCOUNT_TABLE;
+        }
+    }
+
+    if ((chk & QCOW2_OL_SNAPSHOT_TABLE) && s->snapshots_size) {
+        if (ranges_overlap(offset, size, s->snapshots_offset,
+            s->snapshots_size)) {
+            return QCOW2_OL_SNAPSHOT_TABLE;
+        }
+    }
+
+    if ((chk & QCOW2_OL_INACTIVE_L1) && s->snapshots) {
+        for (i = 0; i < s->nb_snapshots; i++) {
+            if (s->snapshots[i].l1_size &&
+                ranges_overlap(offset, size, s->snapshots[i].l1_table_offset,
+                s->snapshots[i].l1_size * sizeof(uint64_t))) {
+                return QCOW2_OL_INACTIVE_L1;
+            }
+        }
+    }
+
+    if ((chk & QCOW2_OL_ACTIVE_L2) && s->l1_table) {
+        for (i = 0; i < s->l1_size; i++) {
+            if ((s->l1_table[i] & L1E_OFFSET_MASK) &&
+                ranges_overlap(offset, size, s->l1_table[i] & L1E_OFFSET_MASK,
+                s->cluster_size)) {
+                return QCOW2_OL_ACTIVE_L2;
+            }
+        }
+    }
+
+    if ((chk & QCOW2_OL_REFCOUNT_BLOCK) && s->refcount_table) {
+        for (i = 0; i < s->refcount_table_size; i++) {
+            if ((s->refcount_table[i] & REFT_OFFSET_MASK) &&
+                ranges_overlap(offset, size, s->refcount_table[i] &
+                REFT_OFFSET_MASK, s->cluster_size)) {
+                return QCOW2_OL_REFCOUNT_BLOCK;
+            }
+        }
+    }
+
+    if ((chk & QCOW2_OL_INACTIVE_L2) && s->snapshots) {
+        for (i = 0; i < s->nb_snapshots; i++) {
+            uint64_t l1_ofs = s->snapshots[i].l1_table_offset;
+            uint32_t l1_sz  = s->snapshots[i].l1_size;
+            uint64_t *l1 = g_malloc(l1_sz * sizeof(uint64_t));
+            int ret;
+
+            ret = bdrv_read(bs->file, l1_ofs / BDRV_SECTOR_SIZE, (uint8_t *)l1,
+                            l1_sz * sizeof(uint64_t) / BDRV_SECTOR_SIZE);
+
+            if (ret < 0) {
+                g_free(l1);
+                return ret;
+            }
+
+            for (j = 0; j < l1_sz; j++) {
+                if ((l1[j] & L1E_OFFSET_MASK) &&
+                    ranges_overlap(offset, size, l1[j] & L1E_OFFSET_MASK,
+                    s->cluster_size)) {
+                    g_free(l1);
+                    return QCOW2_OL_INACTIVE_L2;
+                }
+            }
+
+            g_free(l1);
+        }
+    }
+
+    return 0;
+}
+
+/*
+ * First performs a check for metadata overlaps (through
+ * qcow2_check_metadata_overlap); if that fails with a negative value (error
+ * while performing a check), it will print a message but otherwise ignore that
+ * error. If an impending overlap is detected, the BDS will be made unusable 
and
+ * the qcow2 file marked corrupt.
+ *
+ * Returns 0 if there were no overlaps (or an error occured while checking for
+ * overlaps) or a positive QCow2MetadataOverlap value on overlap (then, the BDS
+ * will be unusable and the qcow2 file marked corrupt).
+ */
+int qcow2_pre_write_overlap_check(BlockDriverState *bs, QCow2MetadataOverlap 
chk,
+                                  int64_t offset, int64_t size)
+{
+    int ret = qcow2_check_metadata_overlap(bs, chk, offset, size);
+
+    if (ret < 0) {
+        fprintf(stderr, "qcow2: Error while checking for metadata overlaps: "
+                "%s\n", strerror(-ret));
+        return ret;
+    } else if (ret > 0) {
+        fprintf(stderr, "qcow2: Preventing invalid write on metadata; "
+                "image marked as corrupt.\n");
+        qcow2_mark_corrupt(bs);
+        bs->drv = NULL; /* make BDS unusable */
+        return ret;
+    }
+
+    return 0;
+}
diff --git a/block/qcow2.h b/block/qcow2.h
index 4297487..8a55da0 100644
--- a/block/qcow2.h
+++ b/block/qcow2.h
@@ -289,6 +289,29 @@ enum {
     QCOW2_CLUSTER_ZERO
 };
 
+typedef enum QCow2MetadataOverlap {
+    QCOW2_OL_NONE           = 0,
+    QCOW2_OL_MAIN_HEADER    = (1 << 0),
+    QCOW2_OL_ACTIVE_L1      = (1 << 1),
+    QCOW2_OL_ACTIVE_L2      = (1 << 2),
+    QCOW2_OL_REFCOUNT_TABLE = (1 << 3),
+    QCOW2_OL_REFCOUNT_BLOCK = (1 << 4),
+    QCOW2_OL_SNAPSHOT_TABLE = (1 << 5),
+    QCOW2_OL_INACTIVE_L1    = (1 << 6),
+    /* NOTE: Checking overlaps with inactive L2 tables will result in bdrv
+     * reads. */
+    QCOW2_OL_INACTIVE_L2    = (1 << 7),
+} QCow2MetadataOverlap;
+
+/* Perform all overlap checks which don't require disk access */
+#define QCOW2_OL_CACHED \
+    (QCOW2_OL_MAIN_HEADER | QCOW2_OL_ACTIVE_L1 | QCOW2_OL_ACTIVE_L2 | \
+     QCOW2_OL_REFCOUNT_TABLE | QCOW2_OL_REFCOUNT_BLOCK | \
+     QCOW2_OL_SNAPSHOT_TABLE | QCOW2_OL_INACTIVE_L1)
+
+/* The default checks to perform */
+#define QCOW2_OL_DEFAULT QCOW2_OL_CACHED
+
 #define L1E_OFFSET_MASK 0x00ffffffffffff00ULL
 #define L2E_OFFSET_MASK 0x00ffffffffffff00ULL
 #define L2E_COMPRESSED_OFFSET_SIZE_MASK 0x3fffffffffffffffULL
@@ -390,6 +413,11 @@ int qcow2_check_refcounts(BlockDriverState *bs, 
BdrvCheckResult *res,
 
 void qcow2_process_discards(BlockDriverState *bs, int ret);
 
+int qcow2_check_metadata_overlap(BlockDriverState *bs, QCow2MetadataOverlap 
chk,
+                                 int64_t offset, int64_t size);
+int qcow2_pre_write_overlap_check(BlockDriverState *bs, QCow2MetadataOverlap 
chk,
+                                  int64_t offset, int64_t size);
+
 /* qcow2-cluster.c functions */
 int qcow2_grow_l1_table(BlockDriverState *bs, uint64_t min_size,
                         bool exact_size);
-- 
1.8.3.1