[Qemu-devel] [PULL v2 21/26] qcow2: Employ metadata overlap checks

[Kevin Wolf <kwolf@redhat.com>]

From: Max Reitz <mreitz@redhat.com>

The pre-write overlap check function is now called before most of the
qcow2 writes (aborting it on collision or other error).

Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
---
 block/qcow2-cache.c    | 17 +++++++++++++++++
 block/qcow2-cluster.c  | 21 +++++++++++++++++++++
 block/qcow2-snapshot.c | 22 ++++++++++++++++++++++
 block/qcow2.c          | 26 ++++++++++++++++++++++++++
 4 files changed, 86 insertions(+)

diff --git a/block/qcow2-cache.c b/block/qcow2-cache.c
index 2f3114e..7bcae09 100644
--- a/block/qcow2-cache.c
+++ b/block/qcow2-cache.c
@@ -115,6 +115,23 @@ static int qcow2_cache_entry_flush(BlockDriverState *bs, Qcow2Cache *c, int i)
     }
 
     if (c == s->refcount_block_cache) {
+        ret = qcow2_pre_write_overlap_check(bs,
+                QCOW2_OL_DEFAULT & ~QCOW2_OL_REFCOUNT_BLOCK,
+                c->entries[i].offset, s->cluster_size);
+    } else if (c == s->l2_table_cache) {
+        ret = qcow2_pre_write_overlap_check(bs,
+                QCOW2_OL_DEFAULT & ~QCOW2_OL_ACTIVE_L2,
+                c->entries[i].offset, s->cluster_size);
+    } else {
+        ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_DEFAULT,
+                c->entries[i].offset, s->cluster_size);
+    }
+
+    if (ret < 0) {
+        return ret;
+    }
+
+    if (c == s->refcount_block_cache) {
         BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_UPDATE_PART);
     } else if (c == s->l2_table_cache) {
         BLKDBG_EVENT(bs->file, BLKDBG_L2_UPDATE);
diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c
index cca76d4..7c248aa 100644
--- a/block/qcow2-cluster.c
+++ b/block/qcow2-cluster.c
@@ -80,6 +80,14 @@ int qcow2_grow_l1_table(BlockDriverState *bs, uint64_t min_size,
         goto fail;
     }
 
+    /* the L1 position has not yet been updated, so these clusters must
+     * indeed be completely free */
+    ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_DEFAULT,
+                                        new_l1_table_offset, new_l1_size2);
+    if (ret < 0) {
+        goto fail;
+    }
+
     BLKDBG_EVENT(bs->file, BLKDBG_L1_GROW_WRITE_TABLE);
     for(i = 0; i < s->l1_size; i++)
         new_l1_table[i] = cpu_to_be64(new_l1_table[i]);
@@ -149,6 +157,13 @@ static int write_l1_entry(BlockDriverState *bs, int l1_index)
         buf[i] = cpu_to_be64(s->l1_table[l1_start_index + i]);
     }
 
+    ret = qcow2_pre_write_overlap_check(bs,
+            QCOW2_OL_DEFAULT & ~QCOW2_OL_ACTIVE_L1,
+            s->l1_table_offset + 8 * l1_start_index, sizeof(buf));
+    if (ret < 0) {
+        return ret;
+    }
+
     BLKDBG_EVENT(bs->file, BLKDBG_L1_UPDATE);
     ret = bdrv_pwrite_sync(bs->file, s->l1_table_offset + 8 * l1_start_index,
         buf, sizeof(buf));
@@ -368,6 +383,12 @@ static int coroutine_fn copy_sectors(BlockDriverState *bs,
                         &s->aes_encrypt_key);
     }
 
+    ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_DEFAULT,
+            cluster_offset + n_start * BDRV_SECTOR_SIZE, n * BDRV_SECTOR_SIZE);
+    if (ret < 0) {
+        goto out;
+    }
+
     BLKDBG_EVENT(bs->file, BLKDBG_COW_WRITE);
     ret = bdrv_co_writev(bs->file, (cluster_offset >> 9) + n_start, n, &qiov);
     if (ret < 0) {
diff --git a/block/qcow2-snapshot.c b/block/qcow2-snapshot.c
index 0caac90..e7e6013 100644
--- a/block/qcow2-snapshot.c
+++ b/block/qcow2-snapshot.c
@@ -189,6 +189,15 @@ static int qcow2_write_snapshots(BlockDriverState *bs)
         return ret;
     }
 
+    /* The snapshot list position has not yet been updated, so these clusters
+     * must indeed be completely free */
+    ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_DEFAULT, offset,
+                                        s->snapshots_size);
+    if (ret < 0) {
+        return ret;
+    }
+
+
     /* Write all snapshots to the new list */
     for(i = 0; i < s->nb_snapshots; i++) {
         sn = s->snapshots + i;
@@ -363,6 +372,12 @@ int qcow2_snapshot_create(BlockDriverState *bs, QEMUSnapshotInfo *sn_info)
         l1_table[i] = cpu_to_be64(s->l1_table[i]);
     }
 
+    ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_DEFAULT,
+            sn->l1_table_offset, s->l1_size * sizeof(uint64_t));
+    if (ret < 0) {
+        goto fail;
+    }
+
     ret = bdrv_pwrite(bs->file, sn->l1_table_offset, l1_table,
                       s->l1_size * sizeof(uint64_t));
     if (ret < 0) {
@@ -475,6 +490,13 @@ int qcow2_snapshot_goto(BlockDriverState *bs, const char *snapshot_id)
         goto fail;
     }
 
+    ret = qcow2_pre_write_overlap_check(bs,
+            QCOW2_OL_DEFAULT & ~QCOW2_OL_ACTIVE_L1,
+            s->l1_table_offset, cur_l1_bytes);
+    if (ret < 0) {
+        goto fail;
+    }
+
     ret = bdrv_pwrite_sync(bs->file, s->l1_table_offset, sn_l1_table,
                            cur_l1_bytes);
     if (ret < 0) {
diff --git a/block/qcow2.c b/block/qcow2.c
index fe91568..05e002d 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -624,6 +624,8 @@ static int qcow2_open(BlockDriverState *bs, QDict *options, int flags)
     qcow2_free_snapshots(bs);
     qcow2_refcount_close(bs);
     g_free(s->l1_table);
+    /* else pre-write overlap checks in cache_destroy may crash */
+    s->l1_table = NULL;
     if (s->l2_table_cache) {
         qcow2_cache_destroy(bs, s->l2_table_cache);
     }
@@ -923,6 +925,13 @@ static coroutine_fn int qcow2_co_writev(BlockDriverState *bs,
                 cur_nr_sectors * 512);
         }
 
+        ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_DEFAULT,
+                cluster_offset + index_in_cluster * BDRV_SECTOR_SIZE,
+                cur_nr_sectors * BDRV_SECTOR_SIZE);
+        if (ret < 0) {
+            goto fail;
+        }
+
         qemu_co_mutex_unlock(&s->lock);
         BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO);
         trace_qcow2_writev_data(qemu_coroutine_self(),
@@ -989,6 +998,8 @@ static void qcow2_close(BlockDriverState *bs)
 {
     BDRVQcowState *s = bs->opaque;
     g_free(s->l1_table);
+    /* else pre-write overlap checks in cache_destroy may crash */
+    s->l1_table = NULL;
 
     qcow2_cache_flush(bs, s->l2_table_cache);
     qcow2_cache_flush(bs, s->refcount_block_cache);
@@ -1668,6 +1679,14 @@ static int qcow2_write_compressed(BlockDriverState *bs, int64_t sector_num,
 
     if (ret != Z_STREAM_END || out_len >= s->cluster_size) {
         /* could not compress: write normal cluster */
+
+        ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_DEFAULT,
+                sector_num * BDRV_SECTOR_SIZE,
+                s->cluster_sectors * BDRV_SECTOR_SIZE);
+        if (ret < 0) {
+            goto fail;
+        }
+
         ret = bdrv_write(bs, sector_num, buf, s->cluster_sectors);
         if (ret < 0) {
             goto fail;
@@ -1680,6 +1699,13 @@ static int qcow2_write_compressed(BlockDriverState *bs, int64_t sector_num,
             goto fail;
         }
         cluster_offset &= s->cluster_offset_mask;
+
+        ret = qcow2_pre_write_overlap_check(bs, QCOW2_OL_DEFAULT,
+                cluster_offset, out_len);
+        if (ret < 0) {
+            goto fail;
+        }
+
         BLKDBG_EVENT(bs->file, BLKDBG_WRITE_COMPRESSED);
         ret = bdrv_pwrite(bs->file, cluster_offset, out_buf, out_len);
         if (ret < 0) {
-- 
1.8.1.4