Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging

Block layer patches

# gpg: Signature made Fri 18 Dec 2015 13:41:03 GMT using RSA key ID C88F2FD6
# gpg: Good signature from "Kevin Wolf <kwolf@redhat.com>"

* remotes/kevin/tags/for-upstream: (48 commits)
  block/qapi: allow best-effort query
  qemu-img: abort when full_backing_filename not present
  block/qapi: explicitly warn if !has_full_backing_filename
  block/qapi: always report full_backing_filename
  block/qapi: do not redundantly print "actual path"
  qemu-iotests: s390x: fix test 068
  qemu-iotests: s390x: fix test 051
  qemu-iotests: refine common.config
  block: fix bdrv_ioctl called from coroutine
  block: use drained section around bdrv_snapshot_delete
  iotests: Update comments for bdrv_swap() in 094
  block: Remove prototype of bdrv_swap from header
  raw-posix: Make aio=native option binding
  qcow2: insert assert into qcow2_get_specific_info()
  iotests: Extend test 112 for qemu-img amend
  qcow2: Point to amend function in check
  qcow2: Invoke refcount order amendment function
  qcow2: Add function for refcount order amendment
  qcow2: Use intermediate helper CB for amend
  qcow2: Split upgrade/downgrade paths for amend
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

block/blkdebug.c

@@ -674,17 +674,15 @@ static int blkdebug_truncate(BlockDriverState *bs, int64_t offset)
     return bdrv_truncate(bs->file->bs, offset);
 }
 
-static void blkdebug_refresh_filename(BlockDriverState *bs)
+static void blkdebug_refresh_filename(BlockDriverState *bs, QDict *options)
 {
     QDict *opts;
     const QDictEntry *e;
     bool force_json = false;
 
-    for (e = qdict_first(bs->options); e; e = qdict_next(bs->options, e)) {
+    for (e = qdict_first(options); e; e = qdict_next(options, e)) {
         if (strcmp(qdict_entry_key(e), "config") &&
-            strcmp(qdict_entry_key(e), "x-image") &&
-            strcmp(qdict_entry_key(e), "image") &&
-            strncmp(qdict_entry_key(e), "image.", strlen("image.")))
+            strcmp(qdict_entry_key(e), "x-image"))
         {
             force_json = true;
             break;
@@ -700,7 +698,7 @@ static void blkdebug_refresh_filename(BlockDriverState *bs)
     if (!force_json && bs->file->bs->exact_filename[0]) {
         snprintf(bs->exact_filename, sizeof(bs->exact_filename),
                  "blkdebug:%s:%s",
-                 qdict_get_try_str(bs->options, "config") ?: "",
+                 qdict_get_try_str(options, "config") ?: "",
                  bs->file->bs->exact_filename);
     }
 
@@ -710,11 +708,8 @@ static void blkdebug_refresh_filename(BlockDriverState *bs)
     QINCREF(bs->file->bs->full_open_options);
     qdict_put_obj(opts, "image", QOBJECT(bs->file->bs->full_open_options));
 
-    for (e = qdict_first(bs->options); e; e = qdict_next(bs->options, e)) {
-        if (strcmp(qdict_entry_key(e), "x-image") &&
-            strcmp(qdict_entry_key(e), "image") &&
-            strncmp(qdict_entry_key(e), "image.", strlen("image.")))
-        {
+    for (e = qdict_first(options); e; e = qdict_next(options, e)) {
+        if (strcmp(qdict_entry_key(e), "x-image")) {
             qobject_incref(qdict_entry_value(e));
             qdict_put_obj(opts, qdict_entry_key(e), qdict_entry_value(e));
         }
@@ -723,6 +718,12 @@ static void blkdebug_refresh_filename(BlockDriverState *bs)
     bs->full_open_options = opts;
 }
 
+static int blkdebug_reopen_prepare(BDRVReopenState *reopen_state,
+                                   BlockReopenQueue *queue, Error **errp)
+{
+    return 0;
+}
+
 static BlockDriver bdrv_blkdebug = {
     .format_name            = "blkdebug",
     .protocol_name          = "blkdebug",
@@ -731,6 +732,7 @@ static BlockDriver bdrv_blkdebug = {
     .bdrv_parse_filename    = blkdebug_parse_filename,
     .bdrv_file_open         = blkdebug_open,
     .bdrv_close             = blkdebug_close,
+    .bdrv_reopen_prepare    = blkdebug_reopen_prepare,
     .bdrv_getlength         = blkdebug_getlength,
     .bdrv_truncate          = blkdebug_truncate,
     .bdrv_refresh_filename  = blkdebug_refresh_filename,

block/blkverify.c

@@ -307,7 +307,7 @@ static void blkverify_attach_aio_context(BlockDriverState *bs,
     bdrv_attach_aio_context(s->test_file->bs, new_context);
 }
 
-static void blkverify_refresh_filename(BlockDriverState *bs)
+static void blkverify_refresh_filename(BlockDriverState *bs, QDict *options)
 {
     BDRVBlkverifyState *s = bs->opaque;
 

block/io.c

@@ -2614,10 +2614,11 @@ int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
         bdrv_co_ioctl_entry(&data);
     } else {
         Coroutine *co = qemu_coroutine_create(bdrv_co_ioctl_entry);
+
         qemu_coroutine_enter(co, &data);
-    }
-    while (data.ret == -EINPROGRESS) {
-        aio_poll(bdrv_get_aio_context(bs), true);
+        while (data.ret == -EINPROGRESS) {
+            aio_poll(bdrv_get_aio_context(bs), true);
+        }
     }
     return data.ret;
 }

block/mirror.c

@@ -18,6 +18,7 @@
 #include "qapi/qmp/qerror.h"
 #include "qemu/ratelimit.h"
 #include "qemu/bitmap.h"
+#include "qemu/error-report.h"
 
 #define SLICE_TIME    100000000ULL /* ns */
 #define MAX_IN_FLIGHT 16
@@ -370,11 +371,22 @@ static void mirror_exit(BlockJob *job, void *opaque)
         if (s->to_replace) {
             to_replace = s->to_replace;
         }
+
+        /* This was checked in mirror_start_job(), but meanwhile one of the
+         * nodes could have been newly attached to a BlockBackend. */
+        if (to_replace->blk && s->target->blk) {
+            error_report("block job: Can't create node with two BlockBackends");
+            data->ret = -EINVAL;
+            goto out;
+        }
+
         if (bdrv_get_flags(s->target) != bdrv_get_flags(to_replace)) {
             bdrv_reopen(s->target, bdrv_get_flags(to_replace), NULL);
         }
         bdrv_replace_in_backing_chain(to_replace, s->target);
     }
+
+out:
     if (s->to_replace) {
         bdrv_op_unblock_all(s->to_replace, s->replace_blocker);
         error_free(s->replace_blocker);
@@ -640,7 +652,7 @@ static void mirror_complete(BlockJob *job, Error **errp)
     Error *local_err = NULL;
     int ret;
 
-    ret = bdrv_open_backing_file(s->target, NULL, &local_err);
+    ret = bdrv_open_backing_file(s->target, NULL, "backing", &local_err);
     if (ret < 0) {
         error_propagate(errp, local_err);
         return;
@@ -705,6 +717,7 @@ static void mirror_start_job(BlockDriverState *bs, BlockDriverState *target,
                              bool is_none_mode, BlockDriverState *base)
 {
     MirrorBlockJob *s;
+    BlockDriverState *replaced_bs;
 
     if (granularity == 0) {
         granularity = bdrv_get_default_bitmap_granularity(target);
@@ -728,6 +741,21 @@ static void mirror_start_job(BlockDriverState *bs, BlockDriverState *target,
         buf_size = DEFAULT_MIRROR_BUF_SIZE;
     }
 
+    /* We can't support this case as long as the block layer can't handle
+     * multiple BlockBackends per BlockDriverState. */
+    if (replaces) {
+        replaced_bs = bdrv_lookup_bs(replaces, replaces, errp);
+        if (replaced_bs == NULL) {
+            return;
+        }
+    } else {
+        replaced_bs = bs;
+    }
+    if (replaced_bs->blk && target->blk) {
+        error_setg(errp, "Can't create node with two BlockBackends");
+        return;
+    }
+
     s = block_job_create(driver, bs, speed, cb, opaque, errp);
     if (!s) {
         return;

block/nbd.c

@@ -342,13 +342,13 @@ static void nbd_attach_aio_context(BlockDriverState *bs,
     nbd_client_attach_aio_context(bs, new_context);
 }
 
-static void nbd_refresh_filename(BlockDriverState *bs)
+static void nbd_refresh_filename(BlockDriverState *bs, QDict *options)
 {
     QDict *opts = qdict_new();
-    const char *path   = qdict_get_try_str(bs->options, "path");
-    const char *host   = qdict_get_try_str(bs->options, "host");
-    const char *port   = qdict_get_try_str(bs->options, "port");
-    const char *export = qdict_get_try_str(bs->options, "export");
+    const char *path   = qdict_get_try_str(options, "path");
+    const char *host   = qdict_get_try_str(options, "host");
+    const char *port   = qdict_get_try_str(options, "port");
+    const char *export = qdict_get_try_str(options, "export");
 
     qdict_put_obj(opts, "driver", QOBJECT(qstring_from_str("nbd")));
 

block/qapi.c

@@ -245,15 +245,17 @@ void bdrv_query_image_info(BlockDriverState *bs,
         info->has_backing_filename = true;
         bdrv_get_full_backing_filename(bs, backing_filename2, PATH_MAX, &err);
         if (err) {
-            error_propagate(errp, err);
-            qapi_free_ImageInfo(info);
+            /* Can't reconstruct the full backing filename, so we must omit
+             * this field and apply a Best Effort to this query. */
             g_free(backing_filename2);
-            return;
+            backing_filename2 = NULL;
+            error_free(err);
         }
 
-        if (strcmp(backing_filename, backing_filename2) != 0) {
-            info->full_backing_filename =
-                        g_strdup(backing_filename2);
+        /* Always report the full_backing_filename if present, even if it's the
+         * same as backing_filename. That they are same is useful info. */
+        if (backing_filename2) {
+            info->full_backing_filename = g_strdup(backing_filename2);
             info->has_full_backing_filename = true;
         }
 
@@ -676,7 +678,10 @@ void bdrv_image_info_dump(fprintf_function func_fprintf, void *f,
 
     if (info->has_backing_filename) {
         func_fprintf(f, "backing file: %s", info->backing_filename);
-        if (info->has_full_backing_filename) {
+        if (!info->has_full_backing_filename) {
+            func_fprintf(f, " (cannot determine actual path)");
+        } else if (strcmp(info->backing_filename,
+                          info->full_backing_filename) != 0) {
             func_fprintf(f, " (actual path: %s)", info->full_backing_filename);
         }
         func_fprintf(f, "\n");

block/qcow2-cluster.c

@@ -1641,7 +1641,8 @@ fail:
 static int expand_zero_clusters_in_l1(BlockDriverState *bs, uint64_t *l1_table,
                                       int l1_size, int64_t *visited_l1_entries,
                                       int64_t l1_entries,
-                                      BlockDriverAmendStatusCB *status_cb)
+                                      BlockDriverAmendStatusCB *status_cb,
+                                      void *cb_opaque)
 {
     BDRVQcow2State *s = bs->opaque;
     bool is_active_l1 = (l1_table == s->l1_table);
@@ -1667,7 +1668,7 @@ static int expand_zero_clusters_in_l1(BlockDriverState *bs, uint64_t *l1_table,
             /* unallocated */
             (*visited_l1_entries)++;
             if (status_cb) {
-                status_cb(bs, *visited_l1_entries, l1_entries);
+                status_cb(bs, *visited_l1_entries, l1_entries, cb_opaque);
             }
             continue;
         }
@@ -1804,7 +1805,7 @@ static int expand_zero_clusters_in_l1(BlockDriverState *bs, uint64_t *l1_table,
 
         (*visited_l1_entries)++;
         if (status_cb) {
-            status_cb(bs, *visited_l1_entries, l1_entries);
+            status_cb(bs, *visited_l1_entries, l1_entries, cb_opaque);
         }
     }
 
@@ -1828,7 +1829,8 @@ fail:
  * qcow2 version which doesn't yet support metadata zero clusters.
  */
 int qcow2_expand_zero_clusters(BlockDriverState *bs,
-                               BlockDriverAmendStatusCB *status_cb)
+                               BlockDriverAmendStatusCB *status_cb,
+                               void *cb_opaque)
 {
     BDRVQcow2State *s = bs->opaque;
     uint64_t *l1_table = NULL;
@@ -1845,7 +1847,7 @@ int qcow2_expand_zero_clusters(BlockDriverState *bs,
 
     ret = expand_zero_clusters_in_l1(bs, s->l1_table, s->l1_size,
                                      &visited_l1_entries, l1_entries,
-                                     status_cb);
+                                     status_cb, cb_opaque);
     if (ret < 0) {
         goto fail;
     }
@@ -1881,7 +1883,7 @@ int qcow2_expand_zero_clusters(BlockDriverState *bs,
 
         ret = expand_zero_clusters_in_l1(bs, l1_table, s->snapshots[i].l1_size,
                                          &visited_l1_entries, l1_entries,
-                                         status_cb);
+                                         status_cb, cb_opaque);
         if (ret < 0) {
             goto fail;
         }

block/qcow2-refcount.c

@@ -1345,6 +1345,9 @@ static int inc_refcounts(BlockDriverState *bs,
         if (refcount == s->refcount_max) {
             fprintf(stderr, "ERROR: overflow cluster offset=0x%" PRIx64
                     "\n", cluster_offset);
+            fprintf(stderr, "Use qemu-img amend to increase the refcount entry "
+                    "width or qemu-img convert to create a clean copy if the "
+                    "image cannot be opened for writing\n");
             res->corruptions++;
             continue;
         }
@@ -2467,3 +2470,450 @@ int qcow2_pre_write_overlap_check(BlockDriverState *bs, int ign, int64_t offset,
 
     return 0;
 }
+
+/* A pointer to a function of this type is given to walk_over_reftable(). That
+ * function will create refblocks and pass them to a RefblockFinishOp once they
+ * are completed (@refblock). @refblock_empty is set if the refblock is
+ * completely empty.
+ *
+ * Along with the refblock, a corresponding reftable entry is passed, in the
+ * reftable @reftable (which may be reallocated) at @reftable_index.
+ *
+ * @allocated should be set to true if a new cluster has been allocated.
+ */
+typedef int (RefblockFinishOp)(BlockDriverState *bs, uint64_t **reftable,
+                               uint64_t reftable_index, uint64_t *reftable_size,
+                               void *refblock, bool refblock_empty,
+                               bool *allocated, Error **errp);
+
+/**
+ * This "operation" for walk_over_reftable() allocates the refblock on disk (if
+ * it is not empty) and inserts its offset into the new reftable. The size of
+ * this new reftable is increased as required.
+ */
+static int alloc_refblock(BlockDriverState *bs, uint64_t **reftable,
+                          uint64_t reftable_index, uint64_t *reftable_size,
+                          void *refblock, bool refblock_empty, bool *allocated,
+                          Error **errp)
+{
+    BDRVQcow2State *s = bs->opaque;
+    int64_t offset;
+
+    if (!refblock_empty && reftable_index >= *reftable_size) {
+        uint64_t *new_reftable;
+        uint64_t new_reftable_size;
+
+        new_reftable_size = ROUND_UP(reftable_index + 1,
+                                     s->cluster_size / sizeof(uint64_t));
+        if (new_reftable_size > QCOW_MAX_REFTABLE_SIZE / sizeof(uint64_t)) {
+            error_setg(errp,
+                       "This operation would make the refcount table grow "
+                       "beyond the maximum size supported by QEMU, aborting");
+            return -ENOTSUP;
+        }
+
+        new_reftable = g_try_realloc(*reftable, new_reftable_size *
+                                                sizeof(uint64_t));
+        if (!new_reftable) {
+            error_setg(errp, "Failed to increase reftable buffer size");
+            return -ENOMEM;
+        }
+
+        memset(new_reftable + *reftable_size, 0,
+               (new_reftable_size - *reftable_size) * sizeof(uint64_t));
+
+        *reftable      = new_reftable;
+        *reftable_size = new_reftable_size;
+    }
+
+    if (!refblock_empty && !(*reftable)[reftable_index]) {
+        offset = qcow2_alloc_clusters(bs, s->cluster_size);
+        if (offset < 0) {
+            error_setg_errno(errp, -offset, "Failed to allocate refblock");
+            return offset;
+        }
+        (*reftable)[reftable_index] = offset;
+        *allocated = true;
+    }
+
+    return 0;
+}
+
+/**
+ * This "operation" for walk_over_reftable() writes the refblock to disk at the
+ * offset specified by the new reftable's entry. It does not modify the new
+ * reftable or change any refcounts.
+ */
+static int flush_refblock(BlockDriverState *bs, uint64_t **reftable,
+                          uint64_t reftable_index, uint64_t *reftable_size,
+                          void *refblock, bool refblock_empty, bool *allocated,
+                          Error **errp)
+{
+    BDRVQcow2State *s = bs->opaque;
+    int64_t offset;
+    int ret;
+
+    if (reftable_index < *reftable_size && (*reftable)[reftable_index]) {
+        offset = (*reftable)[reftable_index];
+
+        ret = qcow2_pre_write_overlap_check(bs, 0, offset, s->cluster_size);
+        if (ret < 0) {
+            error_setg_errno(errp, -ret, "Overlap check failed");
+            return ret;
+        }
+
+        ret = bdrv_pwrite(bs->file->bs, offset, refblock, s->cluster_size);
+        if (ret < 0) {
+            error_setg_errno(errp, -ret, "Failed to write refblock");
+            return ret;
+        }
+    } else {
+        assert(refblock_empty);
+    }
+
+    return 0;
+}
+
+/**
+ * This function walks over the existing reftable and every referenced refblock;
+ * if @new_set_refcount is non-NULL, it is called for every refcount entry to
+ * create an equal new entry in the passed @new_refblock. Once that
+ * @new_refblock is completely filled, @operation will be called.
+ *
+ * @status_cb and @cb_opaque are used for the amend operation's status callback.
+ * @index is the index of the walk_over_reftable() calls and @total is the total
+ * number of walk_over_reftable() calls per amend operation. Both are used for
+ * calculating the parameters for the status callback.
+ *
+ * @allocated is set to true if a new cluster has been allocated.
+ */
+static int walk_over_reftable(BlockDriverState *bs, uint64_t **new_reftable,
+                              uint64_t *new_reftable_index,
+                              uint64_t *new_reftable_size,
+                              void *new_refblock, int new_refblock_size,
+                              int new_refcount_bits,
+                              RefblockFinishOp *operation, bool *allocated,
+                              Qcow2SetRefcountFunc *new_set_refcount,
+                              BlockDriverAmendStatusCB *status_cb,
+                              void *cb_opaque, int index, int total,
+                              Error **errp)
+{
+    BDRVQcow2State *s = bs->opaque;
+    uint64_t reftable_index;
+    bool new_refblock_empty = true;
+    int refblock_index;
+    int new_refblock_index = 0;
+    int ret;
+
+    for (reftable_index = 0; reftable_index < s->refcount_table_size;
+         reftable_index++)
+    {
+        uint64_t refblock_offset = s->refcount_table[reftable_index]
+                                 & REFT_OFFSET_MASK;
+
+        status_cb(bs, (uint64_t)index * s->refcount_table_size + reftable_index,
+                  (uint64_t)total * s->refcount_table_size, cb_opaque);
+
+        if (refblock_offset) {
+            void *refblock;
+
+            if (offset_into_cluster(s, refblock_offset)) {
+                qcow2_signal_corruption(bs, true, -1, -1, "Refblock offset %#"
+                                        PRIx64 " unaligned (reftable index: %#"
+                                        PRIx64 ")", refblock_offset,
+                                        reftable_index);
+                error_setg(errp,
+                           "Image is corrupt (unaligned refblock offset)");
+                return -EIO;
+            }
+
+            ret = qcow2_cache_get(bs, s->refcount_block_cache, refblock_offset,
+                                  &refblock);
+            if (ret < 0) {
+                error_setg_errno(errp, -ret, "Failed to retrieve refblock");
+                return ret;
+            }
+
+            for (refblock_index = 0; refblock_index < s->refcount_block_size;
+                 refblock_index++)
+            {
+                uint64_t refcount;
+
+                if (new_refblock_index >= new_refblock_size) {
+                    /* new_refblock is now complete */
+                    ret = operation(bs, new_reftable, *new_reftable_index,
+                                    new_reftable_size, new_refblock,
+                                    new_refblock_empty, allocated, errp);
+                    if (ret < 0) {
+                        qcow2_cache_put(bs, s->refcount_block_cache, &refblock);
+                        return ret;
+                    }
+
+                    (*new_reftable_index)++;
+                    new_refblock_index = 0;
+                    new_refblock_empty = true;
+                }
+
+                refcount = s->get_refcount(refblock, refblock_index);
+                if (new_refcount_bits < 64 && refcount >> new_refcount_bits) {
+                    uint64_t offset;
+
+                    qcow2_cache_put(bs, s->refcount_block_cache, &refblock);
+
+                    offset = ((reftable_index << s->refcount_block_bits)
+                              + refblock_index) << s->cluster_bits;
+
+                    error_setg(errp, "Cannot decrease refcount entry width to "
+                               "%i bits: Cluster at offset %#" PRIx64 " has a "
+                               "refcount of %" PRIu64, new_refcount_bits,
+                               offset, refcount);
+                    return -EINVAL;
+                }
+
+                if (new_set_refcount) {
+                    new_set_refcount(new_refblock, new_refblock_index++,
+                                     refcount);
+                } else {
+                    new_refblock_index++;
+                }
+                new_refblock_empty = new_refblock_empty && refcount == 0;
+            }
+
+            qcow2_cache_put(bs, s->refcount_block_cache, &refblock);
+        } else {
+            /* No refblock means every refcount is 0 */
+            for (refblock_index = 0; refblock_index < s->refcount_block_size;
+                 refblock_index++)
+            {
+                if (new_refblock_index >= new_refblock_size) {
+                    /* new_refblock is now complete */
+                    ret = operation(bs, new_reftable, *new_reftable_index,
+                                    new_reftable_size, new_refblock,
+                                    new_refblock_empty, allocated, errp);
+                    if (ret < 0) {
+                        return ret;
+                    }
+
+                    (*new_reftable_index)++;
+                    new_refblock_index = 0;
+                    new_refblock_empty = true;
+                }
+
+                if (new_set_refcount) {
+                    new_set_refcount(new_refblock, new_refblock_index++, 0);
+                } else {
+                    new_refblock_index++;
+                }
+            }
+        }
+    }
+
+    if (new_refblock_index > 0) {
+        /* Complete the potentially existing partially filled final refblock */
+        if (new_set_refcount) {
+            for (; new_refblock_index < new_refblock_size;
+                 new_refblock_index++)
+            {
+                new_set_refcount(new_refblock, new_refblock_index, 0);
+            }
+        }
+
+        ret = operation(bs, new_reftable, *new_reftable_index,
+                        new_reftable_size, new_refblock, new_refblock_empty,
+                        allocated, errp);
+        if (ret < 0) {
+            return ret;
+        }
+
+        (*new_reftable_index)++;
+    }
+
+    status_cb(bs, (uint64_t)(index + 1) * s->refcount_table_size,
+              (uint64_t)total * s->refcount_table_size, cb_opaque);
+
+    return 0;
+}
+
+int qcow2_change_refcount_order(BlockDriverState *bs, int refcount_order,
+                                BlockDriverAmendStatusCB *status_cb,
+                                void *cb_opaque, Error **errp)
+{
+    BDRVQcow2State *s = bs->opaque;
+    Qcow2GetRefcountFunc *new_get_refcount;
+    Qcow2SetRefcountFunc *new_set_refcount;
+    void *new_refblock = qemu_blockalign(bs->file->bs, s->cluster_size);
+    uint64_t *new_reftable = NULL, new_reftable_size = 0;
+    uint64_t *old_reftable, old_reftable_size, old_reftable_offset;
+    uint64_t new_reftable_index = 0;
+    uint64_t i;
+    int64_t new_reftable_offset = 0, allocated_reftable_size = 0;
+    int new_refblock_size, new_refcount_bits = 1 << refcount_order;
+    int old_refcount_order;
+    int walk_index = 0;
+    int ret;
+    bool new_allocation;
+
+    assert(s->qcow_version >= 3);
+    assert(refcount_order >= 0 && refcount_order <= 6);
+
+    /* see qcow2_open() */
+    new_refblock_size = 1 << (s->cluster_bits - (refcount_order - 3));
+
+    new_get_refcount = get_refcount_funcs[refcount_order];
+    new_set_refcount = set_refcount_funcs[refcount_order];
+
+
+    do {
+        int total_walks;
+
+        new_allocation = false;
+
+        /* At least we have to do this walk and the one which writes the
+         * refblocks; also, at least we have to do this loop here at least
+         * twice (normally), first to do the allocations, and second to
+         * determine that everything is correctly allocated, this then makes
+         * three walks in total */
+        total_walks = MAX(walk_index + 2, 3);
+
+        /* First, allocate the structures so they are present in the refcount
+         * structures */
+        ret = walk_over_reftable(bs, &new_reftable, &new_reftable_index,
+                                 &new_reftable_size, NULL, new_refblock_size,
+                                 new_refcount_bits, &alloc_refblock,
+                                 &new_allocation, NULL, status_cb, cb_opaque,
+                                 walk_index++, total_walks, errp);
+        if (ret < 0) {
+            goto done;
+        }
+
+        new_reftable_index = 0;
+
+        if (new_allocation) {
+            if (new_reftable_offset) {
+                qcow2_free_clusters(bs, new_reftable_offset,
+                                    allocated_reftable_size * sizeof(uint64_t),
+                                    QCOW2_DISCARD_NEVER);
+            }
+
+            new_reftable_offset = qcow2_alloc_clusters(bs, new_reftable_size *
+                                                           sizeof(uint64_t));
+            if (new_reftable_offset < 0) {
+                error_setg_errno(errp, -new_reftable_offset,
+                                 "Failed to allocate the new reftable");
+                ret = new_reftable_offset;
+                goto done;
+            }
+            allocated_reftable_size = new_reftable_size;
+        }
+    } while (new_allocation);
+
+    /* Second, write the new refblocks */
+    ret = walk_over_reftable(bs, &new_reftable, &new_reftable_index,
+                             &new_reftable_size, new_refblock,
+                             new_refblock_size, new_refcount_bits,
+                             &flush_refblock, &new_allocation, new_set_refcount,
+                             status_cb, cb_opaque, walk_index, walk_index + 1,
+                             errp);
+    if (ret < 0) {
+        goto done;
+    }
+    assert(!new_allocation);
+
+
+    /* Write the new reftable */
+    ret = qcow2_pre_write_overlap_check(bs, 0, new_reftable_offset,
+                                        new_reftable_size * sizeof(uint64_t));
+    if (ret < 0) {
+        error_setg_errno(errp, -ret, "Overlap check failed");
+        goto done;
+    }
+
+    for (i = 0; i < new_reftable_size; i++) {
+        cpu_to_be64s(&new_reftable[i]);
+    }
+
+    ret = bdrv_pwrite(bs->file->bs, new_reftable_offset, new_reftable,
+                      new_reftable_size * sizeof(uint64_t));
+
+    for (i = 0; i < new_reftable_size; i++) {
+        be64_to_cpus(&new_reftable[i]);
+    }
+
+    if (ret < 0) {
+        error_setg_errno(errp, -ret, "Failed to write the new reftable");
+        goto done;
+    }
+
+
+    /* Empty the refcount cache */
+    ret = qcow2_cache_flush(bs, s->refcount_block_cache);
+    if (ret < 0) {
+        error_setg_errno(errp, -ret, "Failed to flush the refblock cache");
+        goto done;
+    }
+
+    /* Update the image header to point to the new reftable; this only updates
+     * the fields which are relevant to qcow2_update_header(); other fields
+     * such as s->refcount_table or s->refcount_bits stay stale for now
+     * (because we have to restore everything if qcow2_update_header() fails) */
+    old_refcount_order  = s->refcount_order;
+    old_reftable_size   = s->refcount_table_size;
+    old_reftable_offset = s->refcount_table_offset;
+
+    s->refcount_order        = refcount_order;
+    s->refcount_table_size   = new_reftable_size;
+    s->refcount_table_offset = new_reftable_offset;
+
+    ret = qcow2_update_header(bs);
+    if (ret < 0) {
+        s->refcount_order        = old_refcount_order;
+        s->refcount_table_size   = old_reftable_size;
+        s->refcount_table_offset = old_reftable_offset;
+        error_setg_errno(errp, -ret, "Failed to update the qcow2 header");
+        goto done;
+    }
+
+    /* Now update the rest of the in-memory information */
+    old_reftable = s->refcount_table;
+    s->refcount_table = new_reftable;
+
+    s->refcount_bits = 1 << refcount_order;
+    s->refcount_max = UINT64_C(1) << (s->refcount_bits - 1);
+    s->refcount_max += s->refcount_max - 1;
+
+    s->refcount_block_bits = s->cluster_bits - (refcount_order - 3);
+    s->refcount_block_size = 1 << s->refcount_block_bits;
+
+    s->get_refcount = new_get_refcount;
+    s->set_refcount = new_set_refcount;
+
+    /* For cleaning up all old refblocks and the old reftable below the "done"
+     * label */
+    new_reftable        = old_reftable;
+    new_reftable_size   = old_reftable_size;
+    new_reftable_offset = old_reftable_offset;
+
+done:
+    if (new_reftable) {
+        /* On success, new_reftable actually points to the old reftable (and
+         * new_reftable_size is the old reftable's size); but that is just
+         * fine */
+        for (i = 0; i < new_reftable_size; i++) {
+            uint64_t offset = new_reftable[i] & REFT_OFFSET_MASK;
+            if (offset) {
+                qcow2_free_clusters(bs, offset, s->cluster_size,
+                                    QCOW2_DISCARD_OTHER);
+            }
+        }
+        g_free(new_reftable);
+
+        if (new_reftable_offset > 0) {
+            qcow2_free_clusters(bs, new_reftable_offset,
+                                new_reftable_size * sizeof(uint64_t),
+                                QCOW2_DISCARD_OTHER);
+        }
+    }
+
+    qemu_vfree(new_refblock);
+    return ret;
+}

block/qcow2.c

@@ -1282,6 +1282,52 @@ static void qcow2_reopen_abort(BDRVReopenState *state)
     g_free(state->opaque);
 }
 
+static void qcow2_join_options(QDict *options, QDict *old_options)
+{
+    bool has_new_overlap_template =
+        qdict_haskey(options, QCOW2_OPT_OVERLAP) ||
+        qdict_haskey(options, QCOW2_OPT_OVERLAP_TEMPLATE);
+    bool has_new_total_cache_size =
+        qdict_haskey(options, QCOW2_OPT_CACHE_SIZE);
+    bool has_all_cache_options;
+
+    /* New overlap template overrides all old overlap options */
+    if (has_new_overlap_template) {
+        qdict_del(old_options, QCOW2_OPT_OVERLAP);
+        qdict_del(old_options, QCOW2_OPT_OVERLAP_TEMPLATE);
+        qdict_del(old_options, QCOW2_OPT_OVERLAP_MAIN_HEADER);
+        qdict_del(old_options, QCOW2_OPT_OVERLAP_ACTIVE_L1);
+        qdict_del(old_options, QCOW2_OPT_OVERLAP_ACTIVE_L2);
+        qdict_del(old_options, QCOW2_OPT_OVERLAP_REFCOUNT_TABLE);
+        qdict_del(old_options, QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK);
+        qdict_del(old_options, QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE);
+        qdict_del(old_options, QCOW2_OPT_OVERLAP_INACTIVE_L1);
+        qdict_del(old_options, QCOW2_OPT_OVERLAP_INACTIVE_L2);
+    }
+
+    /* New total cache size overrides all old options */
+    if (qdict_haskey(options, QCOW2_OPT_CACHE_SIZE)) {
+        qdict_del(old_options, QCOW2_OPT_L2_CACHE_SIZE);
+        qdict_del(old_options, QCOW2_OPT_REFCOUNT_CACHE_SIZE);
+    }
+
+    qdict_join(options, old_options, false);
+
+    /*
+     * If after merging all cache size options are set, an old total size is
+     * overwritten. Do keep all options, however, if all three are new. The
+     * resulting error message is what we want to happen.
+     */
+    has_all_cache_options =
+        qdict_haskey(options, QCOW2_OPT_CACHE_SIZE) ||
+        qdict_haskey(options, QCOW2_OPT_L2_CACHE_SIZE) ||
+        qdict_haskey(options, QCOW2_OPT_REFCOUNT_CACHE_SIZE);
+
+    if (has_all_cache_options && !has_new_total_cache_size) {
+        qdict_del(options, QCOW2_OPT_CACHE_SIZE);
+    }
+}
+
 static int64_t coroutine_fn qcow2_co_get_block_status(BlockDriverState *bs,
         int64_t sector_num, int nb_sectors, int *pnum)
 {
@@ -2757,6 +2803,10 @@ static ImageInfoSpecific *qcow2_get_specific_info(BlockDriverState *bs)
             .has_corrupt        = true,
             .refcount_bits      = s->refcount_bits,
         };
+    } else {
+        /* if this assertion fails, this probably means a new version was
+         * added without having it covered here */
+        assert(false);
     }
 
     return spec_info;
@@ -2824,7 +2874,7 @@ static int qcow2_load_vmstate(BlockDriverState *bs, uint8_t *buf,
  * have to be removed.
  */
 static int qcow2_downgrade(BlockDriverState *bs, int target_version,
-                           BlockDriverAmendStatusCB *status_cb)
+                           BlockDriverAmendStatusCB *status_cb, void *cb_opaque)
 {
     BDRVQcow2State *s = bs->opaque;
     int current_version = s->qcow_version;
@@ -2839,13 +2889,7 @@ static int qcow2_downgrade(BlockDriverState *bs, int target_version,
     }
 
     if (s->refcount_order != 4) {
-        /* we would have to convert the image to a refcount_order == 4 image
-         * here; however, since qemu (at the time of writing this) does not
-         * support anything different than 4 anyway, there is no point in doing
-         * so right now; however, we should error out (if qemu supports this in
-         * the future and this code has not been adapted) */
-        error_report("qcow2_downgrade: Image refcount orders other than 4 are "
-                     "currently not supported.");
+        error_report("compat=0.10 requires refcount_bits=16");
         return -ENOTSUP;
     }
 
@@ -2873,7 +2917,7 @@ static int qcow2_downgrade(BlockDriverState *bs, int target_version,
     /* clearing autoclear features is trivial */
     s->autoclear_features = 0;
 
-    ret = qcow2_expand_zero_clusters(bs, status_cb);
+    ret = qcow2_expand_zero_clusters(bs, status_cb, cb_opaque);
     if (ret < 0) {
         return ret;
     }
@@ -2887,8 +2931,79 @@ static int qcow2_downgrade(BlockDriverState *bs, int target_version,
     return 0;
 }
 
+typedef enum Qcow2AmendOperation {
+    /* This is the value Qcow2AmendHelperCBInfo::last_operation will be
+     * statically initialized to so that the helper CB can discern the first
+     * invocation from an operation change */
+    QCOW2_NO_OPERATION = 0,
+
+    QCOW2_CHANGING_REFCOUNT_ORDER,
+    QCOW2_DOWNGRADING,
+} Qcow2AmendOperation;
+
+typedef struct Qcow2AmendHelperCBInfo {
+    /* The code coordinating the amend operations should only modify
+     * these four fields; the rest will be managed by the CB */
+    BlockDriverAmendStatusCB *original_status_cb;
+    void *original_cb_opaque;
+
+    Qcow2AmendOperation current_operation;
+
+    /* Total number of operations to perform (only set once) */
+    int total_operations;
+
+    /* The following fields are managed by the CB */
+
+    /* Number of operations completed */
+    int operations_completed;
+
+    /* Cumulative offset of all completed operations */
+    int64_t offset_completed;
+
+    Qcow2AmendOperation last_operation;
+    int64_t last_work_size;
+} Qcow2AmendHelperCBInfo;
+
+static void qcow2_amend_helper_cb(BlockDriverState *bs,
+                                  int64_t operation_offset,
+                                  int64_t operation_work_size, void *opaque)
+{
+    Qcow2AmendHelperCBInfo *info = opaque;
+    int64_t current_work_size;
+    int64_t projected_work_size;
+
+    if (info->current_operation != info->last_operation) {
+        if (info->last_operation != QCOW2_NO_OPERATION) {
+            info->offset_completed += info->last_work_size;
+            info->operations_completed++;
+        }
+
+        info->last_operation = info->current_operation;
+    }
+
+    assert(info->total_operations > 0);
+    assert(info->operations_completed < info->total_operations);
+
+    info->last_work_size = operation_work_size;
+
+    current_work_size = info->offset_completed + operation_work_size;
+
+    /* current_work_size is the total work size for (operations_completed + 1)
+     * operations (which includes this one), so multiply it by the number of
+     * operations not covered and divide it by the number of operations
+     * covered to get a projection for the operations not covered */
+    projected_work_size = current_work_size * (info->total_operations -
+                                               info->operations_completed - 1)
+                                            / (info->operations_completed + 1);
+
+    info->original_status_cb(bs, info->offset_completed + operation_offset,
+                             current_work_size + projected_work_size,
+                             info->original_cb_opaque);
+}
+
 static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
-                               BlockDriverAmendStatusCB *status_cb)
+                               BlockDriverAmendStatusCB *status_cb,
+                               void *cb_opaque)
 {
     BDRVQcow2State *s = bs->opaque;
     int old_version = s->qcow_version, new_version = old_version;
@@ -2898,8 +3013,10 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
     const char *compat = NULL;
     uint64_t cluster_size = s->cluster_size;
     bool encrypt;
+    int refcount_bits = s->refcount_bits;
     int ret;
     QemuOptDesc *desc = opts->list->desc;
+    Qcow2AmendHelperCBInfo helper_cb_info;
 
     while (desc && desc->name) {
         if (!qemu_opt_find(opts, desc->name)) {
@@ -2917,11 +3034,11 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
             } else if (!strcmp(compat, "1.1")) {
                 new_version = 3;
             } else {
-                fprintf(stderr, "Unknown compatibility level %s.\n", compat);
+                error_report("Unknown compatibility level %s", compat);
                 return -EINVAL;
             }
         } else if (!strcmp(desc->name, BLOCK_OPT_PREALLOC)) {
-            fprintf(stderr, "Cannot change preallocation mode.\n");
+            error_report("Cannot change preallocation mode");
             return -ENOTSUP;
         } else if (!strcmp(desc->name, BLOCK_OPT_SIZE)) {
             new_size = qemu_opt_get_size(opts, BLOCK_OPT_SIZE, 0);
@@ -2934,47 +3051,74 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
                                         !!s->cipher);
 
             if (encrypt != !!s->cipher) {
-                fprintf(stderr, "Changing the encryption flag is not "
-                        "supported.\n");
+                error_report("Changing the encryption flag is not supported");
                 return -ENOTSUP;
             }
         } else if (!strcmp(desc->name, BLOCK_OPT_CLUSTER_SIZE)) {
             cluster_size = qemu_opt_get_size(opts, BLOCK_OPT_CLUSTER_SIZE,
                                              cluster_size);
             if (cluster_size != s->cluster_size) {
-                fprintf(stderr, "Changing the cluster size is not "
-                        "supported.\n");
+                error_report("Changing the cluster size is not supported");
                 return -ENOTSUP;
             }
         } else if (!strcmp(desc->name, BLOCK_OPT_LAZY_REFCOUNTS)) {
             lazy_refcounts = qemu_opt_get_bool(opts, BLOCK_OPT_LAZY_REFCOUNTS,
                                                lazy_refcounts);
         } else if (!strcmp(desc->name, BLOCK_OPT_REFCOUNT_BITS)) {
-            error_report("Cannot change refcount entry width");
-            return -ENOTSUP;
+            refcount_bits = qemu_opt_get_number(opts, BLOCK_OPT_REFCOUNT_BITS,
+                                                refcount_bits);
+
+            if (refcount_bits <= 0 || refcount_bits > 64 ||
+                !is_power_of_2(refcount_bits))
+            {
+                error_report("Refcount width must be a power of two and may "
+                             "not exceed 64 bits");
+                return -EINVAL;
+            }
         } else {
-            /* if this assertion fails, this probably means a new option was
+            /* if this point is reached, this probably means a new option was
              * added without having it covered here */
-            assert(false);
+            abort();
         }
 
         desc++;
     }
 
-    if (new_version != old_version) {
-        if (new_version > old_version) {
-            /* Upgrade */
-            s->qcow_version = new_version;
-            ret = qcow2_update_header(bs);
-            if (ret < 0) {
-                s->qcow_version = old_version;
-                return ret;
-            }
-        } else {
-            ret = qcow2_downgrade(bs, new_version, status_cb);
-            if (ret < 0) {
-                return ret;
-            }
+    helper_cb_info = (Qcow2AmendHelperCBInfo){
+        .original_status_cb = status_cb,
+        .original_cb_opaque = cb_opaque,
+        .total_operations = (new_version < old_version)
+                          + (s->refcount_bits != refcount_bits)
+    };
+
+    /* Upgrade first (some features may require compat=1.1) */
+    if (new_version > old_version) {
+        s->qcow_version = new_version;
+        ret = qcow2_update_header(bs);
+        if (ret < 0) {
+            s->qcow_version = old_version;
+            return ret;
+        }
+    }
+
+    if (s->refcount_bits != refcount_bits) {
+        int refcount_order = ctz32(refcount_bits);
+        Error *local_error = NULL;
+
+        if (new_version < 3 && refcount_bits != 16) {
+            error_report("Different refcount widths than 16 bits require "
+                         "compatibility level 1.1 or above (use compat=1.1 or "
+                         "greater)");
+            return -EINVAL;
+        }
+
+        helper_cb_info.current_operation = QCOW2_CHANGING_REFCOUNT_ORDER;
+        ret = qcow2_change_refcount_order(bs, refcount_order,
+                                          &qcow2_amend_helper_cb,
+                                          &helper_cb_info, &local_error);
+        if (ret < 0) {
+            error_report_err(local_error);
+            return ret;
         }
     }
 
@@ -2989,9 +3133,9 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
 
     if (s->use_lazy_refcounts != lazy_refcounts) {
         if (lazy_refcounts) {
-            if (s->qcow_version < 3) {
-                fprintf(stderr, "Lazy refcounts only supported with compatibility "
-                        "level 1.1 and above (use compat=1.1 or greater)\n");
+            if (new_version < 3) {
+                error_report("Lazy refcounts only supported with compatibility "
+                             "level 1.1 and above (use compat=1.1 or greater)");
                 return -EINVAL;
             }
             s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS;
@@ -3025,6 +3169,16 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
         }
     }
 
+    /* Downgrade last (so unsupported features can be removed before) */
+    if (new_version < old_version) {
+        helper_cb_info.current_operation = QCOW2_DOWNGRADING;
+        ret = qcow2_downgrade(bs, new_version, &qcow2_amend_helper_cb,
+                              &helper_cb_info);
+        if (ret < 0) {
+            return ret;
+        }
+    }
+
     return 0;
 }
 
@@ -3145,6 +3299,7 @@ BlockDriver bdrv_qcow2 = {
     .bdrv_reopen_prepare  = qcow2_reopen_prepare,
     .bdrv_reopen_commit   = qcow2_reopen_commit,
     .bdrv_reopen_abort    = qcow2_reopen_abort,
+    .bdrv_join_options    = qcow2_join_options,
     .bdrv_create        = qcow2_create,
     .bdrv_has_zero_init = bdrv_has_zero_init_1,
     .bdrv_co_get_block_status = qcow2_co_get_block_status,

block/qcow2.h

@@ -529,6 +529,10 @@ int qcow2_check_metadata_overlap(BlockDriverState *bs, int ign, int64_t offset,
 int qcow2_pre_write_overlap_check(BlockDriverState *bs, int ign, int64_t offset,
                                   int64_t size);
 
+int qcow2_change_refcount_order(BlockDriverState *bs, int refcount_order,
+                                BlockDriverAmendStatusCB *status_cb,
+                                void *cb_opaque, Error **errp);
+
 /* qcow2-cluster.c functions */
 int qcow2_grow_l1_table(BlockDriverState *bs, uint64_t min_size,
                         bool exact_size);
@@ -553,7 +557,8 @@ int qcow2_discard_clusters(BlockDriverState *bs, uint64_t offset,
 int qcow2_zero_clusters(BlockDriverState *bs, uint64_t offset, int nb_sectors);
 
 int qcow2_expand_zero_clusters(BlockDriverState *bs,
-                               BlockDriverAmendStatusCB *status_cb);
+                               BlockDriverAmendStatusCB *status_cb,
+                               void *cb_opaque);
 
 /* qcow2-snapshot.c functions */
 int qcow2_snapshot_create(BlockDriverState *bs, QEMUSnapshotInfo *sn_info);

block/quorum.c

@@ -997,7 +997,7 @@ static void quorum_attach_aio_context(BlockDriverState *bs,
     }
 }
 
-static void quorum_refresh_filename(BlockDriverState *bs)
+static void quorum_refresh_filename(BlockDriverState *bs, QDict *options)
 {
     BDRVQuorumState *s = bs->opaque;
     QDict *opts;

block/raw-posix.c

@@ -500,21 +500,17 @@ static int raw_open_common(BlockDriverState *bs, QDict *options,
         goto fail;
     }
     if (!s->use_aio && (bdrv_flags & BDRV_O_NATIVE_AIO)) {
-        error_printf("WARNING: aio=native was specified for '%s', but "
-                     "it requires cache.direct=on, which was not "
-                     "specified. Falling back to aio=threads.\n"
-                     "         This will become an error condition in "
-                     "future QEMU versions.\n",
-                     bs->filename);
+        error_setg(errp, "aio=native was specified, but it requires "
+                         "cache.direct=on, which was not specified.");
+        ret = -EINVAL;
+        goto fail;
     }
 #else
     if (bdrv_flags & BDRV_O_NATIVE_AIO) {
-        error_printf("WARNING: aio=native was specified for '%s', but "
-                     "is not supported in this build. Falling back to "
-                     "aio=threads.\n"
-                     "         This will become an error condition in "
-                     "future QEMU versions.\n",
-                     bs->filename);
+        error_setg(errp, "aio=native was specified, but is not supported "
+                         "in this build.");
+        ret = -EINVAL;
+        goto fail;
     }
 #endif /* !defined(CONFIG_LINUX_AIO) */
 

block/snapshot.c

@@ -229,6 +229,8 @@ int bdrv_snapshot_delete(BlockDriverState *bs,
                          Error **errp)
 {
     BlockDriver *drv = bs->drv;
+    int ret;
+
     if (!drv) {
         error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, bdrv_get_device_name(bs));
         return -ENOMEDIUM;
@@ -239,18 +241,21 @@ int bdrv_snapshot_delete(BlockDriverState *bs,
     }
 
     /* drain all pending i/o before deleting snapshot */
-    bdrv_drain(bs);
+    bdrv_drained_begin(bs);
 
     if (drv->bdrv_snapshot_delete) {
-        return drv->bdrv_snapshot_delete(bs, snapshot_id, name, errp);
+        ret = drv->bdrv_snapshot_delete(bs, snapshot_id, name, errp);
+    } else if (bs->file) {
+        ret = bdrv_snapshot_delete(bs->file->bs, snapshot_id, name, errp);
+    } else {
+        error_setg(errp, "Block format '%s' used by device '%s' "
+                   "does not support internal snapshot deletion",
+                   drv->format_name, bdrv_get_device_name(bs));
+        ret = -ENOTSUP;
     }
-    if (bs->file) {
-        return bdrv_snapshot_delete(bs->file->bs, snapshot_id, name, errp);
-    }
-    error_setg(errp, "Block format '%s' used by device '%s' "
-               "does not support internal snapshot deletion",
-               drv->format_name, bdrv_get_device_name(bs));
-    return -ENOTSUP;
+
+    bdrv_drained_end(bs);
+    return ret;
 }
 
 int bdrv_snapshot_delete_by_id_or_name(BlockDriverState *bs,

blockdev.c

@@ -387,16 +387,6 @@ static void extract_common_blockdev_options(QemuOpts *opts, int *bdrv_flags,
             }
         }
 
-        if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_WB, true)) {
-            *bdrv_flags |= BDRV_O_CACHE_WB;
-        }
-        if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_DIRECT, false)) {
-            *bdrv_flags |= BDRV_O_NOCACHE;
-        }
-        if (qemu_opt_get_bool(opts, BDRV_OPT_CACHE_NO_FLUSH, false)) {
-            *bdrv_flags |= BDRV_O_NO_FLUSH;
-        }
-
         if ((aio = qemu_opt_get(opts, "aio")) != NULL) {
             if (!strcmp(aio, "native")) {
                 *bdrv_flags |= BDRV_O_NATIVE_AIO;
@@ -490,7 +480,6 @@ static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
     QDict *interval_dict = NULL;
     QList *interval_list = NULL;
     const char *id;
-    bool has_driver_specific_opts;
     BlockdevDetectZeroesOptions detect_zeroes =
         BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
     const char *throttling_group = NULL;
@@ -514,8 +503,6 @@ static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
         qdict_del(bs_opts, "id");
     }
 
-    has_driver_specific_opts = !!qdict_size(bs_opts);
-
     /* extract parameters */
     snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
 
@@ -572,13 +559,11 @@ static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
     }
 
     if (snapshot) {
-        /* always use cache=unsafe with snapshot */
-        bdrv_flags &= ~BDRV_O_CACHE_MASK;
-        bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
+        bdrv_flags |= BDRV_O_SNAPSHOT;
     }
 
     /* init */
-    if ((!file || !*file) && !has_driver_specific_opts) {
+    if ((!file || !*file) && !qdict_size(bs_opts)) {
         BlockBackendRootState *blk_rs;
 
         blk = blk_new(qemu_opts_id(opts), errp);
@@ -606,6 +591,20 @@ static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
             file = NULL;
         }
 
+        /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
+         * with other callers) rather than what we want as the real defaults.
+         * Apply the defaults here instead. */
+        qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_WB, "on");
+        qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
+        qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
+
+        if (snapshot) {
+            /* always use cache=unsafe with snapshot */
+            qdict_put(bs_opts, BDRV_OPT_CACHE_WB, qstring_from_str("on"));
+            qdict_put(bs_opts, BDRV_OPT_CACHE_DIRECT, qstring_from_str("off"));
+            qdict_put(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, qstring_from_str("on"));
+        }
+
         blk = blk_new_open(qemu_opts_id(opts), file, NULL, bs_opts, bdrv_flags,
                            errp);
         if (!blk) {
@@ -3872,18 +3871,6 @@ QemuOptsList qemu_common_drive_opts = {
             .name = "discard",
             .type = QEMU_OPT_STRING,
             .help = "discard operation (ignore/off, unmap/on)",
-        },{
-            .name = BDRV_OPT_CACHE_WB,
-            .type = QEMU_OPT_BOOL,
-            .help = "enables writeback mode for any caches",
-        },{
-            .name = BDRV_OPT_CACHE_DIRECT,
-            .type = QEMU_OPT_BOOL,
-            .help = "enables use of O_DIRECT (bypass the host page cache)",
-        },{
-            .name = BDRV_OPT_CACHE_NO_FLUSH,
-            .type = QEMU_OPT_BOOL,
-            .help = "ignore any flush requests for the device",
         },{
             .name = "aio",
             .type = QEMU_OPT_STRING,
@@ -3991,18 +3978,6 @@ static QemuOptsList qemu_root_bds_opts = {
             .name = "discard",
             .type = QEMU_OPT_STRING,
             .help = "discard operation (ignore/off, unmap/on)",
-        },{
-            .name = "cache.writeback",
-            .type = QEMU_OPT_BOOL,
-            .help = "enables writeback mode for any caches",
-        },{
-            .name = "cache.direct",
-            .type = QEMU_OPT_BOOL,
-            .help = "enables use of O_DIRECT (bypass the host page cache)",
-        },{
-            .name = "cache.no-flush",
-            .type = QEMU_OPT_BOOL,
-            .help = "ignore any flush requests for the device",
         },{
             .name = "aio",
             .type = QEMU_OPT_STRING,

include/block/block.h

@@ -150,6 +150,7 @@ typedef struct BDRVReopenState {
     BlockDriverState *bs;
     int flags;
     QDict *options;
+    QDict *explicit_options;
     void *opaque;
 } BDRVReopenState;
 
@@ -197,7 +198,6 @@ int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp);
 BlockDriverState *bdrv_new_root(void);
 BlockDriverState *bdrv_new(void);
 void bdrv_make_anon(BlockDriverState *bs);
-void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old);
 void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top);
 void bdrv_replace_in_backing_chain(BlockDriverState *old,
                                    BlockDriverState *new);
@@ -210,7 +210,8 @@ BdrvChild *bdrv_open_child(const char *filename,
                            const BdrvChildRole *child_role,
                            bool allow_none, Error **errp);
 void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd);
-int bdrv_open_backing_file(BlockDriverState *bs, QDict *options, Error **errp);
+int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
+                           const char *bdref_key, Error **errp);
 int bdrv_append_temp_snapshot(BlockDriverState *bs, int flags, Error **errp);
 int bdrv_open(BlockDriverState **pbs, const char *filename,
               const char *reference, QDict *options, int flags, Error **errp);
@@ -304,9 +305,9 @@ int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix);
  * block driver; total_work_size may change during the course of the amendment
  * operation */
 typedef void BlockDriverAmendStatusCB(BlockDriverState *bs, int64_t offset,
-                                      int64_t total_work_size);
+                                      int64_t total_work_size, void *opaque);
 int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts,
-                       BlockDriverAmendStatusCB *status_cb);
+                       BlockDriverAmendStatusCB *status_cb, void *cb_opaque);
 
 /* external snapshots */
 bool bdrv_recurse_is_first_non_filter(BlockDriverState *bs,

include/block/block_int.h

@@ -121,6 +121,7 @@ struct BlockDriver {
                                BlockReopenQueue *queue, Error **errp);
     void (*bdrv_reopen_commit)(BDRVReopenState *reopen_state);
     void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state);
+    void (*bdrv_join_options)(QDict *options, QDict *old_options);
 
     int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags,
                      Error **errp);
@@ -135,7 +136,7 @@ struct BlockDriver {
     int (*bdrv_set_key)(BlockDriverState *bs, const char *key);
     int (*bdrv_make_empty)(BlockDriverState *bs);
 
-    void (*bdrv_refresh_filename)(BlockDriverState *bs);
+    void (*bdrv_refresh_filename)(BlockDriverState *bs, QDict *options);
 
     /* aio */
     BlockAIOCB *(*bdrv_aio_readv)(BlockDriverState *bs,
@@ -242,7 +243,8 @@ struct BlockDriver {
         BdrvCheckMode fix);
 
     int (*bdrv_amend_options)(BlockDriverState *bs, QemuOpts *opts,
-                              BlockDriverAmendStatusCB *status_cb);
+                              BlockDriverAmendStatusCB *status_cb,
+                              void *cb_opaque);
 
     void (*bdrv_debug_event)(BlockDriverState *bs, BlkdebugEvent event);
 
@@ -342,7 +344,8 @@ typedef struct BdrvAioNotifier {
 } BdrvAioNotifier;
 
 struct BdrvChildRole {
-    int (*inherit_flags)(int parent_flags);
+    void (*inherit_options)(int *child_flags, QDict *child_options,
+                            int parent_flags, QDict *parent_options);
 };
 
 extern const BdrvChildRole child_file;
@@ -350,6 +353,7 @@ extern const BdrvChildRole child_format;
 
 struct BdrvChild {
     BlockDriverState *bs;
+    char *name;
     const BdrvChildRole *role;
     QLIST_ENTRY(BdrvChild) next;
     QLIST_ENTRY(BdrvChild) next_parent;
@@ -456,6 +460,7 @@ struct BlockDriverState {
     QLIST_HEAD(, BdrvChild) parents;
 
     QDict *options;
+    QDict *explicit_options;
     BlockdevDetectZeroesOptions detect_zeroes;
 
     /* The error object in use for blocking operations on backing_hd */

qemu-img.c

@@ -2040,7 +2040,10 @@ static ImageInfoList *collect_image_info_list(const char *filename,
             if (info->has_full_backing_filename) {
                 filename = info->full_backing_filename;
             } else if (info->has_backing_filename) {
-                filename = info->backing_filename;
+                error_report("Could not determine absolute backing filename,"
+                             " but backing filename '%s' present",
+                             info->backing_filename);
+                goto err;
             }
             if (info->has_backing_filename_format) {
                 fmt = info->backing_filename_format;
@@ -2896,7 +2899,8 @@ out:
 }
 
 static void amend_status_cb(BlockDriverState *bs,
-                            int64_t offset, int64_t total_work_size)
+                            int64_t offset, int64_t total_work_size,
+                            void *opaque)
 {
     qemu_progress_print(100.f * offset / total_work_size, 0);
 }
@@ -3020,7 +3024,7 @@ static int img_amend(int argc, char **argv)
 
     /* In case the driver does not call amend_status_cb() */
     qemu_progress_print(0.f, 0);
-    ret = bdrv_amend_options(bs, opts, &amend_status_cb);
+    ret = bdrv_amend_options(bs, opts, &amend_status_cb, NULL);
     qemu_progress_print(100.f, 0);
     if (ret < 0) {
         error_report("Error while amending options: %s", strerror(-ret));

tests/hd-geo-test.c

@@ -206,13 +206,13 @@ static int setup_ide(int argc, char *argv[], int argv_sz,
 {
     char *s1, *s2, *s3;
 
-    s1 = g_strdup_printf("-drive id=drive%d,if=%s,format=raw",
+    s1 = g_strdup_printf("-drive id=drive%d,if=%s",
                          ide_idx, dev ? "none" : "ide");
     s2 = dev ? g_strdup("") : g_strdup_printf(",index=%d", ide_idx);
 
     if (img_secs[img_idx] >= 0) {
         setup_mbr(img_idx, mbr);
-        s3 = g_strdup_printf(",file=%s", img_file_name[img_idx]);
+        s3 = g_strdup_printf(",format=raw,file=%s", img_file_name[img_idx]);
     } else {
         s3 = g_strdup(",media=cdrom");
     }

tests/qemu-iotests/043.out

@@ -44,6 +44,7 @@ cluster_size: 65536
         "filename": "TEST_DIR/t.IMGFMT",
         "cluster-size": 65536,
         "format": "IMGFMT",
+        "full-backing-filename": "TEST_DIR/t.IMGFMT.2.base",
         "backing-filename": "TEST_DIR/t.IMGFMT.2.base",
         "dirty-flag": false
     },
@@ -52,6 +53,7 @@ cluster_size: 65536
         "filename": "TEST_DIR/t.IMGFMT.2.base",
         "cluster-size": 65536,
         "format": "IMGFMT",
+        "full-backing-filename": "TEST_DIR/t.IMGFMT.1.base",
         "backing-filename": "TEST_DIR/t.IMGFMT.1.base",
         "dirty-flag": false
     },