Upgrade version to 5.2upstream/5.2.0sandbox/wangbiao/qemu_5.2

Change-Id: I79199e4f7e9f27b498cd41113833a6158d4f07a2
Signed-off-by: biao716.wang <biao716.wang@samsung.com>

1 files changed, 5634 insertions, 3011 deletions

diff --git a/block.c b/block.c
index 01b66d802..f1cedac36 100644
--- a/block.c
+++ b/block.c
@@ -21,23 +21,36 @@
  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  * THE SOFTWARE.
  */
-#include "config-host.h"
-#include "qemu-common.h"
-#include "trace.h"
-#include "monitor/monitor.h"
+
+#include "qemu/osdep.h"
+#include "block/trace.h"
 #include "block/block_int.h"
 #include "block/blockjob.h"
+#include "block/nbd.h"
+#include "block/qdict.h"
+#include "qemu/error-report.h"
+#include "block/module_block.h"
+#include "qemu/main-loop.h"
 #include "qemu/module.h"
+#include "qapi/error.h"
+#include "qapi/qmp/qdict.h"
 #include "qapi/qmp/qjson.h"
+#include "qapi/qmp/qnull.h"
+#include "qapi/qmp/qstring.h"
+#include "qapi/qobject-output-visitor.h"
+#include "qapi/qapi-visit-block-core.h"
+#include "sysemu/block-backend.h"
 #include "sysemu/sysemu.h"
 #include "qemu/notify.h"
-#include "block/coroutine.h"
-#include "qmp-commands.h"
+#include "qemu/option.h"
+#include "qemu/coroutine.h"
+#include "block/qapi.h"
 #include "qemu/timer.h"
+#include "qemu/cutils.h"
+#include "qemu/id.h"
+#include "block/coroutines.h"
 
 #ifdef CONFIG_BSD
-#include <sys/types.h>
-#include <sys/stat.h>
 #include <sys/ioctl.h>
 #include <sys/queue.h>
 #ifndef __DragonFly__
@@ -51,54 +64,23 @@
 
 #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */
 
-typedef enum {
-    BDRV_REQ_COPY_ON_READ = 0x1,
-    BDRV_REQ_ZERO_WRITE   = 0x2,
-} BdrvRequestFlags;
-
-static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load);
-static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
-        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
-        BlockDriverCompletionFunc *cb, void *opaque);
-static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
-        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
-        BlockDriverCompletionFunc *cb, void *opaque);
-static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs,
-                                         int64_t sector_num, int nb_sectors,
-                                         QEMUIOVector *iov);
-static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs,
-                                         int64_t sector_num, int nb_sectors,
-                                         QEMUIOVector *iov);
-static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs,
-    int64_t sector_num, int nb_sectors, QEMUIOVector *qiov,
-    BdrvRequestFlags flags);
-static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs,
-    int64_t sector_num, int nb_sectors, QEMUIOVector *qiov,
-    BdrvRequestFlags flags);
-static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs,
-                                               int64_t sector_num,
-                                               QEMUIOVector *qiov,
-                                               int nb_sectors,
-                                               BlockDriverCompletionFunc *cb,
-                                               void *opaque,
-                                               bool is_write);
-static void coroutine_fn bdrv_co_do_rw(void *opaque);
-static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs,
-    int64_t sector_num, int nb_sectors);
-
-static bool bdrv_exceed_bps_limits(BlockDriverState *bs, int nb_sectors,
-        bool is_write, double elapsed_time, uint64_t *wait);
-static bool bdrv_exceed_iops_limits(BlockDriverState *bs, bool is_write,
-        double elapsed_time, uint64_t *wait);
-static bool bdrv_exceed_io_limits(BlockDriverState *bs, int nb_sectors,
-        bool is_write, int64_t *wait);
-
-static QTAILQ_HEAD(, BlockDriverState) bdrv_states =
-    QTAILQ_HEAD_INITIALIZER(bdrv_states);
+static QTAILQ_HEAD(, BlockDriverState) graph_bdrv_states =
+    QTAILQ_HEAD_INITIALIZER(graph_bdrv_states);
+
+static QTAILQ_HEAD(, BlockDriverState) all_bdrv_states =
+    QTAILQ_HEAD_INITIALIZER(all_bdrv_states);
 
 static QLIST_HEAD(, BlockDriver) bdrv_drivers =
     QLIST_HEAD_INITIALIZER(bdrv_drivers);
 
+static BlockDriverState *bdrv_open_inherit(const char *filename,
+                                           const char *reference,
+                                           QDict *options, int flags,
+                                           BlockDriverState *parent,
+                                           const BdrvChildClass *child_class,
+                                           BdrvChildRole child_role,
+                                           Error **errp);
+
 /* If non-zero, use only whitelisted block drivers */
 static int use_bdrv_whitelist;
 
@@ -122,75 +104,28 @@ int is_windows_drive(const char *filename)
 }
 #endif
 
-/* throttling disk I/O limits */
-void bdrv_io_limits_disable(BlockDriverState *bs)
+size_t bdrv_opt_mem_align(BlockDriverState *bs)
 {
-    bs->io_limits_enabled = false;
-
-    do {} while (qemu_co_enter_next(&bs->throttled_reqs));
-
-    if (bs->block_timer) {
-        qemu_del_timer(bs->block_timer);
-        qemu_free_timer(bs->block_timer);
-        bs->block_timer = NULL;
+    if (!bs || !bs->drv) {
+        /* page size or 4k (hdd sector size) should be on the safe side */
+        return MAX(4096, qemu_real_host_page_size);
     }
 
-    bs->slice_start = 0;
-    bs->slice_end   = 0;
-}
-
-static void bdrv_block_timer(void *opaque)
-{
-    BlockDriverState *bs = opaque;
-
-    qemu_co_enter_next(&bs->throttled_reqs);
-}
-
-void bdrv_io_limits_enable(BlockDriverState *bs)
-{
-    qemu_co_queue_init(&bs->throttled_reqs);
-    bs->block_timer = qemu_new_timer_ns(vm_clock, bdrv_block_timer, bs);
-    bs->io_limits_enabled = true;
+    return bs->bl.opt_mem_alignment;
 }
 
-bool bdrv_io_limits_enabled(BlockDriverState *bs)
+size_t bdrv_min_mem_align(BlockDriverState *bs)
 {
-    BlockIOLimit *io_limits = &bs->io_limits;
-    return io_limits->bps[BLOCK_IO_LIMIT_READ]
-         || io_limits->bps[BLOCK_IO_LIMIT_WRITE]
-         || io_limits->bps[BLOCK_IO_LIMIT_TOTAL]
-         || io_limits->iops[BLOCK_IO_LIMIT_READ]
-         || io_limits->iops[BLOCK_IO_LIMIT_WRITE]
-         || io_limits->iops[BLOCK_IO_LIMIT_TOTAL];
-}
-
-static void bdrv_io_limits_intercept(BlockDriverState *bs,
-                                     bool is_write, int nb_sectors)
-{
-    int64_t wait_time = -1;
-
-    if (!qemu_co_queue_empty(&bs->throttled_reqs)) {
-        qemu_co_queue_wait(&bs->throttled_reqs);
+    if (!bs || !bs->drv) {
+        /* page size or 4k (hdd sector size) should be on the safe side */
+        return MAX(4096, qemu_real_host_page_size);
     }
 
-    /* In fact, we hope to keep each request's timing, in FIFO mode. The next
-     * throttled requests will not be dequeued until the current request is
-     * allowed to be serviced. So if the current request still exceeds the
-     * limits, it will be inserted to the head. All requests followed it will
-     * be still in throttled_reqs queue.
-     */
-
-    while (bdrv_exceed_io_limits(bs, nb_sectors, is_write, &wait_time)) {
-        qemu_mod_timer(bs->block_timer,
-                       wait_time + qemu_get_clock_ns(vm_clock));
-        qemu_co_queue_wait_insert_head(&bs->throttled_reqs);
-    }
-
-    qemu_co_queue_next(&bs->throttled_reqs);
+    return bs->bl.min_mem_alignment;
 }
 
 /* check if the path starts with "<protocol>:" */
-static int path_has_protocol(const char *path)
+int path_has_protocol(const char *path)
 {
     const char *p;
 
@@ -220,119 +155,290 @@ int path_is_absolute(const char *path)
 #endif
 }
 
-/* if filename is absolute, just copy it to dest. Otherwise, build a
+/* if filename is absolute, just return its duplicate. Otherwise, build a
    path to it by considering it is relative to base_path. URL are
    supported. */
-void path_combine(char *dest, int dest_size,
-                  const char *base_path,
-                  const char *filename)
+char *path_combine(const char *base_path, const char *filename)
 {
+    const char *protocol_stripped = NULL;
     const char *p, *p1;
+    char *result;
     int len;
 
-    if (dest_size <= 0)
-        return;
     if (path_is_absolute(filename)) {
-        pstrcpy(dest, dest_size, filename);
-    } else {
-        p = strchr(base_path, ':');
-        if (p)
-            p++;
-        else
-            p = base_path;
-        p1 = strrchr(base_path, '/');
+        return g_strdup(filename);
+    }
+
+    if (path_has_protocol(base_path)) {
+        protocol_stripped = strchr(base_path, ':');
+        if (protocol_stripped) {
+            protocol_stripped++;
+        }
+    }
+    p = protocol_stripped ?: base_path;
+
+    p1 = strrchr(base_path, '/');
 #ifdef _WIN32
-        {
-            const char *p2;
-            p2 = strrchr(base_path, '\\');
-            if (!p1 || p2 > p1)
-                p1 = p2;
+    {
+        const char *p2;
+        p2 = strrchr(base_path, '\\');
+        if (!p1 || p2 > p1) {
+            p1 = p2;
         }
+    }
 #endif
-        if (p1)
-            p1++;
-        else
-            p1 = base_path;
-        if (p1 > p)
-            p = p1;
-        len = p - base_path;
-        if (len > dest_size - 1)
-            len = dest_size - 1;
-        memcpy(dest, base_path, len);
-        dest[len] = '\0';
-        pstrcat(dest, dest_size, filename);
+    if (p1) {
+        p1++;
+    } else {
+        p1 = base_path;
+    }
+    if (p1 > p) {
+        p = p1;
+    }
+    len = p - base_path;
+
+    result = g_malloc(len + strlen(filename) + 1);
+    memcpy(result, base_path, len);
+    strcpy(result + len, filename);
+
+    return result;
+}
+
+/*
+ * Helper function for bdrv_parse_filename() implementations to remove optional
+ * protocol prefixes (especially "file:") from a filename and for putting the
+ * stripped filename into the options QDict if there is such a prefix.
+ */
+void bdrv_parse_filename_strip_prefix(const char *filename, const char *prefix,
+                                      QDict *options)
+{
+    if (strstart(filename, prefix, &filename)) {
+        /* Stripping the explicit protocol prefix may result in a protocol
+         * prefix being (wrongly) detected (if the filename contains a colon) */
+        if (path_has_protocol(filename)) {
+            QString *fat_filename;
+
+            /* This means there is some colon before the first slash; therefore,
+             * this cannot be an absolute path */
+            assert(!path_is_absolute(filename));
+
+            /* And we can thus fix the protocol detection issue by prefixing it
+             * by "./" */
+            fat_filename = qstring_from_str("./");
+            qstring_append(fat_filename, filename);
+
+            assert(!path_has_protocol(qstring_get_str(fat_filename)));
+
+            qdict_put(options, "filename", fat_filename);
+        } else {
+            /* If no protocol prefix was detected, we can use the shortened
+             * filename as-is */
+            qdict_put_str(options, "filename", filename);
+        }
+    }
+}
+
+
+/* Returns whether the image file is opened as read-only. Note that this can
+ * return false and writing to the image file is still not possible because the
+ * image is inactivated. */
+bool bdrv_is_read_only(BlockDriverState *bs)
+{
+    return bs->read_only;
+}
+
+int bdrv_can_set_read_only(BlockDriverState *bs, bool read_only,
+                           bool ignore_allow_rdw, Error **errp)
+{
+    /* Do not set read_only if copy_on_read is enabled */
+    if (bs->copy_on_read && read_only) {
+        error_setg(errp, "Can't set node '%s' to r/o with copy-on-read enabled",
+                   bdrv_get_device_or_node_name(bs));
+        return -EINVAL;
+    }
+
+    /* Do not clear read_only if it is prohibited */
+    if (!read_only && !(bs->open_flags & BDRV_O_ALLOW_RDWR) &&
+        !ignore_allow_rdw)
+    {
+        error_setg(errp, "Node '%s' is read only",
+                   bdrv_get_device_or_node_name(bs));
+        return -EPERM;
+    }
+
+    return 0;
+}
+
+/*
+ * Called by a driver that can only provide a read-only image.
+ *
+ * Returns 0 if the node is already read-only or it could switch the node to
+ * read-only because BDRV_O_AUTO_RDONLY is set.
+ *
+ * Returns -EACCES if the node is read-write and BDRV_O_AUTO_RDONLY is not set
+ * or bdrv_can_set_read_only() forbids making the node read-only. If @errmsg
+ * is not NULL, it is used as the error message for the Error object.
+ */
+int bdrv_apply_auto_read_only(BlockDriverState *bs, const char *errmsg,
+                              Error **errp)
+{
+    int ret = 0;
+
+    if (!(bs->open_flags & BDRV_O_RDWR)) {
+        return 0;
+    }
+    if (!(bs->open_flags & BDRV_O_AUTO_RDONLY)) {
+        goto fail;
+    }
+
+    ret = bdrv_can_set_read_only(bs, true, false, NULL);
+    if (ret < 0) {
+        goto fail;
     }
+
+    bs->read_only = true;
+    bs->open_flags &= ~BDRV_O_RDWR;
+
+    return 0;
+
+fail:
+    error_setg(errp, "%s", errmsg ?: "Image is read-only");
+    return -EACCES;
 }
 
-void bdrv_get_full_backing_filename(BlockDriverState *bs, char *dest, size_t sz)
+/*
+ * If @backing is empty, this function returns NULL without setting
+ * @errp.  In all other cases, NULL will only be returned with @errp
+ * set.
+ *
+ * Therefore, a return value of NULL without @errp set means that
+ * there is no backing file; if @errp is set, there is one but its
+ * absolute filename cannot be generated.
+ */
+char *bdrv_get_full_backing_filename_from_filename(const char *backed,
+                                                   const char *backing,
+                                                   Error **errp)
 {
-    if (bs->backing_file[0] == '\0' || path_has_protocol(bs->backing_file)) {
-        pstrcpy(dest, sz, bs->backing_file);
+    if (backing[0] == '\0') {
+        return NULL;
+    } else if (path_has_protocol(backing) || path_is_absolute(backing)) {
+        return g_strdup(backing);
+    } else if (backed[0] == '\0' || strstart(backed, "json:", NULL)) {
+        error_setg(errp, "Cannot use relative backing file names for '%s'",
+                   backed);
+        return NULL;
     } else {
-        path_combine(dest, sz, bs->filename, bs->backing_file);
+        return path_combine(backed, backing);
     }
 }
 
-void bdrv_register(BlockDriver *bdrv)
+/*
+ * If @filename is empty or NULL, this function returns NULL without
+ * setting @errp.  In all other cases, NULL will only be returned with
+ * @errp set.
+ */
+static char *bdrv_make_absolute_filename(BlockDriverState *relative_to,
+                                         const char *filename, Error **errp)
 {
-    /* Block drivers without coroutine functions need emulation */
-    if (!bdrv->bdrv_co_readv) {
-        bdrv->bdrv_co_readv = bdrv_co_readv_em;
-        bdrv->bdrv_co_writev = bdrv_co_writev_em;
+    char *dir, *full_name;
 
-        /* bdrv_co_readv_em()/brdv_co_writev_em() work in terms of aio, so if
-         * the block driver lacks aio we need to emulate that too.
-         */
-        if (!bdrv->bdrv_aio_readv) {
-            /* add AIO emulation layer */
-            bdrv->bdrv_aio_readv = bdrv_aio_readv_em;
-            bdrv->bdrv_aio_writev = bdrv_aio_writev_em;
-        }
+    if (!filename || filename[0] == '\0') {
+        return NULL;
+    } else if (path_has_protocol(filename) || path_is_absolute(filename)) {
+        return g_strdup(filename);
+    }
+
+    dir = bdrv_dirname(relative_to, errp);
+    if (!dir) {
+        return NULL;
     }
 
+    full_name = g_strconcat(dir, filename, NULL);
+    g_free(dir);
+    return full_name;
+}
+
+char *bdrv_get_full_backing_filename(BlockDriverState *bs, Error **errp)
+{
+    return bdrv_make_absolute_filename(bs, bs->backing_file, errp);
+}
+
+void bdrv_register(BlockDriver *bdrv)
+{
+    assert(bdrv->format_name);
     QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list);
 }
 
-/* create a new block device (by default it is empty) */
-BlockDriverState *bdrv_new(const char *device_name)
+BlockDriverState *bdrv_new(void)
 {
     BlockDriverState *bs;
+    int i;
 
-    bs = g_malloc0(sizeof(BlockDriverState));
-    pstrcpy(bs->device_name, sizeof(bs->device_name), device_name);
-    if (device_name[0] != '\0') {
-        QTAILQ_INSERT_TAIL(&bdrv_states, bs, list);
+    bs = g_new0(BlockDriverState, 1);
+    QLIST_INIT(&bs->dirty_bitmaps);
+    for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) {
+        QLIST_INIT(&bs->op_blockers[i]);
     }
-    bdrv_iostatus_disable(bs);
-    notifier_list_init(&bs->close_notifiers);
     notifier_with_return_list_init(&bs->before_write_notifiers);
+    qemu_co_mutex_init(&bs->reqs_lock);
+    qemu_mutex_init(&bs->dirty_bitmap_mutex);
+    bs->refcnt = 1;
+    bs->aio_context = qemu_get_aio_context();
 
-    return bs;
-}
+    qemu_co_queue_init(&bs->flush_queue);
 
-void bdrv_add_close_notifier(BlockDriverState *bs, Notifier *notify)
-{
-    notifier_list_add(&bs->close_notifiers, notify);
+    for (i = 0; i < bdrv_drain_all_count; i++) {
+        bdrv_drained_begin(bs);
+    }
+
+    QTAILQ_INSERT_TAIL(&all_bdrv_states, bs, bs_list);
+
+    return bs;
 }
 
-BlockDriver *bdrv_find_format(const char *format_name)
+static BlockDriver *bdrv_do_find_format(const char *format_name)
 {
     BlockDriver *drv1;
+
     QLIST_FOREACH(drv1, &bdrv_drivers, list) {
         if (!strcmp(drv1->format_name, format_name)) {
             return drv1;
         }
     }
+
     return NULL;
 }
 
-static int bdrv_is_whitelisted(BlockDriver *drv, bool read_only)
+BlockDriver *bdrv_find_format(const char *format_name)
+{
+    BlockDriver *drv1;
+    int i;
+
+    drv1 = bdrv_do_find_format(format_name);
+    if (drv1) {
+        return drv1;
+    }
+
+    /* The driver isn't registered, maybe we need to load a module */
+    for (i = 0; i < (int)ARRAY_SIZE(block_driver_modules); ++i) {
+        if (!strcmp(block_driver_modules[i].format_name, format_name)) {
+            block_module_load_one(block_driver_modules[i].library_name);
+            break;
+        }
+    }
+
+    return bdrv_do_find_format(format_name);
+}
+
+static int bdrv_format_is_whitelisted(const char *format_name, bool read_only)
 {
     static const char *whitelist_rw[] = {
         CONFIG_BDRV_RW_WHITELIST
+        NULL
     };
     static const char *whitelist_ro[] = {
         CONFIG_BDRV_RO_WHITELIST
+        NULL
     };
     const char **p;
 
@@ -341,13 +447,13 @@ static int bdrv_is_whitelisted(BlockDriver *drv, bool read_only)
     }
 
     for (p = whitelist_rw; *p; p++) {
-        if (!strcmp(drv->format_name, *p)) {
+        if (!strcmp(format_name, *p)) {
             return 1;
         }
     }
     if (read_only) {
         for (p = whitelist_ro; *p; p++) {
-            if (!strcmp(drv->format_name, *p)) {
+            if (!strcmp(format_name, *p)) {
                 return 1;
             }
         }
@@ -355,30 +461,40 @@ static int bdrv_is_whitelisted(BlockDriver *drv, bool read_only)
     return 0;
 }
 
-BlockDriver *bdrv_find_whitelisted_format(const char *format_name,
-                                          bool read_only)
+int bdrv_is_whitelisted(BlockDriver *drv, bool read_only)
+{
+    return bdrv_format_is_whitelisted(drv->format_name, read_only);
+}
+
+bool bdrv_uses_whitelist(void)
 {
-    BlockDriver *drv = bdrv_find_format(format_name);
-    return drv && bdrv_is_whitelisted(drv, read_only) ? drv : NULL;
+    return use_bdrv_whitelist;
 }
 
 typedef struct CreateCo {
     BlockDriver *drv;
     char *filename;
-    QEMUOptionParameter *options;
+    QemuOpts *opts;
     int ret;
+    Error *err;
 } CreateCo;
 
 static void coroutine_fn bdrv_create_co_entry(void *opaque)
 {
+    Error *local_err = NULL;
+    int ret;
+
     CreateCo *cco = opaque;
     assert(cco->drv);
 
-    cco->ret = cco->drv->bdrv_create(cco->filename, cco->options);
+    ret = cco->drv->bdrv_co_create_opts(cco->drv,
+                                        cco->filename, cco->opts, &local_err);
+    error_propagate(&cco->err, local_err);
+    cco->ret = ret;
 }
 
 int bdrv_create(BlockDriver *drv, const char* filename,
-    QEMUOptionParameter *options)
+                QemuOpts *opts, Error **errp)
 {
     int ret;
 
@@ -386,11 +502,13 @@ int bdrv_create(BlockDriver *drv, const char* filename,
     CreateCo cco = {
         .drv = drv,
         .filename = g_strdup(filename),
-        .options = options,
+        .opts = opts,
         .ret = NOT_DONE,
+        .err = NULL,
     };
 
-    if (!drv->bdrv_create) {
+    if (!drv->bdrv_co_create_opts) {
+        error_setg(errp, "Driver '%s' does not support image creation", drv->format_name);
         ret = -ENOTSUP;
         goto out;
     }
@@ -399,30 +517,231 @@ int bdrv_create(BlockDriver *drv, const char* filename,
         /* Fast-path if already in coroutine context */
         bdrv_create_co_entry(&cco);
     } else {
-        co = qemu_coroutine_create(bdrv_create_co_entry);
-        qemu_coroutine_enter(co, &cco);
+        co = qemu_coroutine_create(bdrv_create_co_entry, &cco);
+        qemu_coroutine_enter(co);
         while (cco.ret == NOT_DONE) {
-            qemu_aio_wait();
+            aio_poll(qemu_get_aio_context(), true);
         }
     }
 
     ret = cco.ret;
+    if (ret < 0) {
+        if (cco.err) {
+            error_propagate(errp, cco.err);
+        } else {
+            error_setg_errno(errp, -ret, "Could not create image");
+        }
+    }
 
 out:
     g_free(cco.filename);
     return ret;
 }
 
-int bdrv_create_file(const char* filename, QEMUOptionParameter *options)
+/**
+ * Helper function for bdrv_create_file_fallback(): Resize @blk to at
+ * least the given @minimum_size.
+ *
+ * On success, return @blk's actual length.
+ * Otherwise, return -errno.
+ */
+static int64_t create_file_fallback_truncate(BlockBackend *blk,
+                                             int64_t minimum_size, Error **errp)
+{
+    Error *local_err = NULL;
+    int64_t size;
+    int ret;
+
+    ret = blk_truncate(blk, minimum_size, false, PREALLOC_MODE_OFF, 0,
+                       &local_err);
+    if (ret < 0 && ret != -ENOTSUP) {
+        error_propagate(errp, local_err);
+        return ret;
+    }
+
+    size = blk_getlength(blk);
+    if (size < 0) {
+        error_free(local_err);
+        error_setg_errno(errp, -size,
+                         "Failed to inquire the new image file's length");
+        return size;
+    }
+
+    if (size < minimum_size) {
+        /* Need to grow the image, but we failed to do that */
+        error_propagate(errp, local_err);
+        return -ENOTSUP;
+    }
+
+    error_free(local_err);
+    local_err = NULL;
+
+    return size;
+}
+
+/**
+ * Helper function for bdrv_create_file_fallback(): Zero the first
+ * sector to remove any potentially pre-existing image header.
+ */
+static int create_file_fallback_zero_first_sector(BlockBackend *blk,
+                                                  int64_t current_size,
+                                                  Error **errp)
+{
+    int64_t bytes_to_clear;
+    int ret;
+
+    bytes_to_clear = MIN(current_size, BDRV_SECTOR_SIZE);
+    if (bytes_to_clear) {
+        ret = blk_pwrite_zeroes(blk, 0, bytes_to_clear, BDRV_REQ_MAY_UNMAP);
+        if (ret < 0) {
+            error_setg_errno(errp, -ret,
+                             "Failed to clear the new image's first sector");
+            return ret;
+        }
+    }
+
+    return 0;
+}
+
+/**
+ * Simple implementation of bdrv_co_create_opts for protocol drivers
+ * which only support creation via opening a file
+ * (usually existing raw storage device)
+ */
+int coroutine_fn bdrv_co_create_opts_simple(BlockDriver *drv,
+                                            const char *filename,
+                                            QemuOpts *opts,
+                                            Error **errp)
+{
+    BlockBackend *blk;
+    QDict *options;
+    int64_t size = 0;
+    char *buf = NULL;
+    PreallocMode prealloc;
+    Error *local_err = NULL;
+    int ret;
+
+    size = qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0);
+    buf = qemu_opt_get_del(opts, BLOCK_OPT_PREALLOC);
+    prealloc = qapi_enum_parse(&PreallocMode_lookup, buf,
+                               PREALLOC_MODE_OFF, &local_err);
+    g_free(buf);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        return -EINVAL;
+    }
+
+    if (prealloc != PREALLOC_MODE_OFF) {
+        error_setg(errp, "Unsupported preallocation mode '%s'",
+                   PreallocMode_str(prealloc));
+        return -ENOTSUP;
+    }
+
+    options = qdict_new();
+    qdict_put_str(options, "driver", drv->format_name);
+
+    blk = blk_new_open(filename, NULL, options,
+                       BDRV_O_RDWR | BDRV_O_RESIZE, errp);
+    if (!blk) {
+        error_prepend(errp, "Protocol driver '%s' does not support image "
+                      "creation, and opening the image failed: ",
+                      drv->format_name);
+        return -EINVAL;
+    }
+
+    size = create_file_fallback_truncate(blk, size, errp);
+    if (size < 0) {
+        ret = size;
+        goto out;
+    }
+
+    ret = create_file_fallback_zero_first_sector(blk, size, errp);
+    if (ret < 0) {
+        goto out;
+    }
+
+    ret = 0;
+out:
+    blk_unref(blk);
+    return ret;
+}
+
+int bdrv_create_file(const char *filename, QemuOpts *opts, Error **errp)
 {
     BlockDriver *drv;
 
-    drv = bdrv_find_protocol(filename, true);
+    drv = bdrv_find_protocol(filename, true, errp);
     if (drv == NULL) {
         return -ENOENT;
     }
 
-    return bdrv_create(drv, filename, options);
+    return bdrv_create(drv, filename, opts, errp);
+}
+
+int coroutine_fn bdrv_co_delete_file(BlockDriverState *bs, Error **errp)
+{
+    Error *local_err = NULL;
+    int ret;
+
+    assert(bs != NULL);
+
+    if (!bs->drv) {
+        error_setg(errp, "Block node '%s' is not opened", bs->filename);
+        return -ENOMEDIUM;
+    }
+
+    if (!bs->drv->bdrv_co_delete_file) {
+        error_setg(errp, "Driver '%s' does not support image deletion",
+                   bs->drv->format_name);
+        return -ENOTSUP;
+    }
+
+    ret = bs->drv->bdrv_co_delete_file(bs, &local_err);
+    if (ret < 0) {
+        error_propagate(errp, local_err);
+    }
+
+    return ret;
+}
+
+/**
+ * Try to get @bs's logical and physical block size.
+ * On success, store them in @bsz struct and return 0.
+ * On failure return -errno.
+ * @bs must not be empty.
+ */
+int bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz)
+{
+    BlockDriver *drv = bs->drv;
+    BlockDriverState *filtered = bdrv_filter_bs(bs);
+
+    if (drv && drv->bdrv_probe_blocksizes) {
+        return drv->bdrv_probe_blocksizes(bs, bsz);
+    } else if (filtered) {
+        return bdrv_probe_blocksizes(filtered, bsz);
+    }
+
+    return -ENOTSUP;
+}
+
+/**
+ * Try to get @bs's geometry (cyls, heads, sectors).
+ * On success, store them in @geo struct and return 0.
+ * On failure return -errno.
+ * @bs must not be empty.
+ */
+int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo)
+{
+    BlockDriver *drv = bs->drv;
+    BlockDriverState *filtered = bdrv_filter_bs(bs);
+
+    if (drv && drv->bdrv_probe_geometry) {
+        return drv->bdrv_probe_geometry(bs, geo);
+    } else if (filtered) {
+        return bdrv_probe_geometry(filtered, geo);
+    }
+
+    return -ENOTSUP;
 }
 
 /*
@@ -443,8 +762,9 @@ int get_tmp_filename(char *filename, int size)
     int fd;
     const char *tmpdir;
     tmpdir = getenv("TMPDIR");
-    if (!tmpdir)
-        tmpdir = "/tmp";
+    if (!tmpdir) {
+        tmpdir = "/var/tmp";
+    }
     if (snprintf(filename, size, "%s/vl.XXXXXX", tmpdir) >= size) {
         return -EOVERFLOW;
     }
@@ -482,13 +802,28 @@ static BlockDriver *find_hdev_driver(const char *filename)
     return drv;
 }
 
+static BlockDriver *bdrv_do_find_protocol(const char *protocol)
+{
+    BlockDriver *drv1;
+
+    QLIST_FOREACH(drv1, &bdrv_drivers, list) {
+        if (drv1->protocol_name && !strcmp(drv1->protocol_name, protocol)) {
+            return drv1;
+        }
+    }
+
+    return NULL;
+}
+
 BlockDriver *bdrv_find_protocol(const char *filename,
-                                bool allow_protocol_prefix)
+                                bool allow_protocol_prefix,
+                                Error **errp)
 {
     BlockDriver *drv1;
     char protocol[128];
     int len;
     const char *p;
+    int i;
 
     /* TODO Drivers without bdrv_file_open must be specified explicitly */
 
@@ -505,7 +840,7 @@ BlockDriver *bdrv_find_protocol(const char *filename,
     }
 
     if (!path_has_protocol(filename) || !allow_protocol_prefix) {
-        return bdrv_find_format("file");
+        return &bdrv_file;
     }
 
     p = strchr(filename, ':');
@@ -515,51 +850,85 @@ BlockDriver *bdrv_find_protocol(const char *filename,
         len = sizeof(protocol) - 1;
     memcpy(protocol, filename, len);
     protocol[len] = '\0';
-    QLIST_FOREACH(drv1, &bdrv_drivers, list) {
-        if (drv1->protocol_name &&
-            !strcmp(drv1->protocol_name, protocol)) {
-            return drv1;
+
+    drv1 = bdrv_do_find_protocol(protocol);
+    if (drv1) {
+        return drv1;
+    }
+
+    for (i = 0; i < (int)ARRAY_SIZE(block_driver_modules); ++i) {
+        if (block_driver_modules[i].protocol_name &&
+            !strcmp(block_driver_modules[i].protocol_name, protocol)) {
+            block_module_load_one(block_driver_modules[i].library_name);
+            break;
         }
     }
-    return NULL;
+
+    drv1 = bdrv_do_find_protocol(protocol);
+    if (!drv1) {
+        error_setg(errp, "Unknown protocol '%s'", protocol);
+    }
+    return drv1;
+}
+
+/*
+ * Guess image format by probing its contents.
+ * This is not a good idea when your image is raw (CVE-2008-2004), but
+ * we do it anyway for backward compatibility.
+ *
+ * @buf         contains the image's first @buf_size bytes.
+ * @buf_size    is the buffer size in bytes (generally BLOCK_PROBE_BUF_SIZE,
+ *              but can be smaller if the image file is smaller)
+ * @filename    is its filename.
+ *
+ * For all block drivers, call the bdrv_probe() method to get its
+ * probing score.
+ * Return the first block driver with the highest probing score.
+ */
+BlockDriver *bdrv_probe_all(const uint8_t *buf, int buf_size,
+                            const char *filename)
+{
+    int score_max = 0, score;
+    BlockDriver *drv = NULL, *d;
+
+    QLIST_FOREACH(d, &bdrv_drivers, list) {
+        if (d->bdrv_probe) {
+            score = d->bdrv_probe(buf, buf_size, filename);
+            if (score > score_max) {
+                score_max = score;
+                drv = d;
+            }
+        }
+    }
+
+    return drv;
 }
 
-static int find_image_format(BlockDriverState *bs, const char *filename,
-                             BlockDriver **pdrv)
+static int find_image_format(BlockBackend *file, const char *filename,
+                             BlockDriver **pdrv, Error **errp)
 {
-    int score, score_max;
-    BlockDriver *drv1, *drv;
-    uint8_t buf[2048];
+    BlockDriver *drv;
+    uint8_t buf[BLOCK_PROBE_BUF_SIZE];
     int ret = 0;
 
     /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
-    if (bs->sg || !bdrv_is_inserted(bs) || bdrv_getlength(bs) == 0) {
-        drv = bdrv_find_format("raw");
-        if (!drv) {
-            ret = -ENOENT;
-        }
-        *pdrv = drv;
+    if (blk_is_sg(file) || !blk_is_inserted(file) || blk_getlength(file) == 0) {
+        *pdrv = &bdrv_raw;
         return ret;
     }
 
-    ret = bdrv_pread(bs, 0, buf, sizeof(buf));
+    ret = blk_pread(file, 0, buf, sizeof(buf));
     if (ret < 0) {
+        error_setg_errno(errp, -ret, "Could not read image for determining its "
+                         "format");
         *pdrv = NULL;
         return ret;
     }
 
-    score_max = 0;
-    drv = NULL;
-    QLIST_FOREACH(drv1, &bdrv_drivers, list) {
-        if (drv1->bdrv_probe) {
-            score = drv1->bdrv_probe(buf, ret, filename);
-            if (score > score_max) {
-                score_max = score;
-                drv = drv1;
-            }
-        }
-    }
+    drv = bdrv_probe_all(buf, ret, filename);
     if (!drv) {
+        error_setg(errp, "Could not determine image format: No compatible "
+                   "driver found");
         ret = -ENOENT;
     }
     *pdrv = drv;
@@ -568,13 +937,18 @@ static int find_image_format(BlockDriverState *bs, const char *filename,
 
 /**
  * Set the current 'total_sectors' value
+ * Return 0 on success, -errno on error.
  */
-static int refresh_total_sectors(BlockDriverState *bs, int64_t hint)
+int refresh_total_sectors(BlockDriverState *bs, int64_t hint)
 {
     BlockDriver *drv = bs->drv;
 
+    if (!drv) {
+        return -ENOMEDIUM;
+    }
+
     /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
-    if (bs->sg)
+    if (bdrv_is_sg(bs))
         return 0;
 
     /* query actual device if possible, otherwise just trust the hint */
@@ -583,7 +957,7 @@ static int refresh_total_sectors(BlockDriverState *bs, int64_t hint)
         if (length < 0) {
             return length;
         }
-        hint = length >> BDRV_SECTOR_BITS;
+        hint = DIV_ROUND_UP(length, BDRV_SECTOR_SIZE);
     }
 
     bs->total_sectors = hint;
@@ -591,6 +965,67 @@ static int refresh_total_sectors(BlockDriverState *bs, int64_t hint)
 }
 
 /**
+ * Combines a QDict of new block driver @options with any missing options taken
+ * from @old_options, so that leaving out an option defaults to its old value.
+ */
+static void bdrv_join_options(BlockDriverState *bs, QDict *options,
+                              QDict *old_options)
+{
+    if (bs->drv && bs->drv->bdrv_join_options) {
+        bs->drv->bdrv_join_options(options, old_options);
+    } else {
+        qdict_join(options, old_options, false);
+    }
+}
+
+static BlockdevDetectZeroesOptions bdrv_parse_detect_zeroes(QemuOpts *opts,
+                                                            int open_flags,
+                                                            Error **errp)
+{
+    Error *local_err = NULL;
+    char *value = qemu_opt_get_del(opts, "detect-zeroes");
+    BlockdevDetectZeroesOptions detect_zeroes =
+        qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup, value,
+                        BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF, &local_err);
+    g_free(value);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        return detect_zeroes;
+    }
+
+    if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
+        !(open_flags & BDRV_O_UNMAP))
+    {
+        error_setg(errp, "setting detect-zeroes to unmap is not allowed "
+                   "without setting discard operation to unmap");
+    }
+
+    return detect_zeroes;
+}
+
+/**
+ * Set open flags for aio engine
+ *
+ * Return 0 on success, -1 if the engine specified is invalid
+ */
+int bdrv_parse_aio(const char *mode, int *flags)
+{
+    if (!strcmp(mode, "threads")) {
+        /* do nothing, default */
+    } else if (!strcmp(mode, "native")) {
+        *flags |= BDRV_O_NATIVE_AIO;
+#ifdef CONFIG_LINUX_IO_URING
+    } else if (!strcmp(mode, "io_uring")) {
+        *flags |= BDRV_O_IO_URING;
+#endif
+    } else {
+        return -1;
+    }
+
+    return 0;
+}
+
+/**
  * Set open flags for a given discard mode
  *
  * Return 0 on success, -1 if the discard mode was invalid.
@@ -615,21 +1050,23 @@ int bdrv_parse_discard_flags(const char *mode, int *flags)
  *
  * Return 0 on success, -1 if the cache mode was invalid.
  */
-int bdrv_parse_cache_flags(const char *mode, int *flags)
+int bdrv_parse_cache_mode(const char *mode, int *flags, bool *writethrough)
 {
     *flags &= ~BDRV_O_CACHE_MASK;
 
     if (!strcmp(mode, "off") || !strcmp(mode, "none")) {
-        *flags |= BDRV_O_NOCACHE | BDRV_O_CACHE_WB;
+        *writethrough = false;
+        *flags |= BDRV_O_NOCACHE;
     } else if (!strcmp(mode, "directsync")) {
+        *writethrough = true;
         *flags |= BDRV_O_NOCACHE;
     } else if (!strcmp(mode, "writeback")) {
-        *flags |= BDRV_O_CACHE_WB;
+        *writethrough = false;
     } else if (!strcmp(mode, "unsafe")) {
-        *flags |= BDRV_O_CACHE_WB;
+        *writethrough = false;
         *flags |= BDRV_O_NO_FLUSH;
     } else if (!strcmp(mode, "writethrough")) {
-        /* this is the default */
+        *writethrough = true;
     } else {
         return -1;
     }
@@ -637,85 +1074,654 @@ int bdrv_parse_cache_flags(const char *mode, int *flags)
     return 0;
 }
 
-/**
- * The copy-on-read flag is actually a reference count so multiple users may
- * use the feature without worrying about clobbering its previous state.
- * Copy-on-read stays enabled until all users have called to disable it.
+static char *bdrv_child_get_parent_desc(BdrvChild *c)
+{
+    BlockDriverState *parent = c->opaque;
+    return g_strdup(bdrv_get_device_or_node_name(parent));
+}
+
+static void bdrv_child_cb_drained_begin(BdrvChild *child)
+{
+    BlockDriverState *bs = child->opaque;
+    bdrv_do_drained_begin_quiesce(bs, NULL, false);
+}
+
+static bool bdrv_child_cb_drained_poll(BdrvChild *child)
+{
+    BlockDriverState *bs = child->opaque;
+    return bdrv_drain_poll(bs, false, NULL, false);
+}
+
+static void bdrv_child_cb_drained_end(BdrvChild *child,
+                                      int *drained_end_counter)
+{
+    BlockDriverState *bs = child->opaque;
+    bdrv_drained_end_no_poll(bs, drained_end_counter);
+}
+
+static int bdrv_child_cb_inactivate(BdrvChild *child)
+{
+    BlockDriverState *bs = child->opaque;
+    assert(bs->open_flags & BDRV_O_INACTIVE);
+    return 0;
+}
+
+static bool bdrv_child_cb_can_set_aio_ctx(BdrvChild *child, AioContext *ctx,
+                                          GSList **ignore, Error **errp)
+{
+    BlockDriverState *bs = child->opaque;
+    return bdrv_can_set_aio_context(bs, ctx, ignore, errp);
+}
+
+static void bdrv_child_cb_set_aio_ctx(BdrvChild *child, AioContext *ctx,
+                                      GSList **ignore)
+{
+    BlockDriverState *bs = child->opaque;
+    return bdrv_set_aio_context_ignore(bs, ctx, ignore);
+}
+
+/*
+ * Returns the options and flags that a temporary snapshot should get, based on
+ * the originally requested flags (the originally requested image will have
+ * flags like a backing file)
  */
-void bdrv_enable_copy_on_read(BlockDriverState *bs)
+static void bdrv_temp_snapshot_options(int *child_flags, QDict *child_options,
+                                       int parent_flags, QDict *parent_options)
+{
+    *child_flags = (parent_flags & ~BDRV_O_SNAPSHOT) | BDRV_O_TEMPORARY;
+
+    /* For temporary files, unconditional cache=unsafe is fine */
+    qdict_set_default_str(child_options, BDRV_OPT_CACHE_DIRECT, "off");
+    qdict_set_default_str(child_options, BDRV_OPT_CACHE_NO_FLUSH, "on");
+
+    /* Copy the read-only and discard options from the parent */
+    qdict_copy_default(child_options, parent_options, BDRV_OPT_READ_ONLY);
+    qdict_copy_default(child_options, parent_options, BDRV_OPT_DISCARD);
+
+    /* aio=native doesn't work for cache.direct=off, so disable it for the
+     * temporary snapshot */
+    *child_flags &= ~BDRV_O_NATIVE_AIO;
+}
+
+static void bdrv_backing_attach(BdrvChild *c)
 {
-    bs->copy_on_read++;
+    BlockDriverState *parent = c->opaque;
+    BlockDriverState *backing_hd = c->bs;
+
+    assert(!parent->backing_blocker);
+    error_setg(&parent->backing_blocker,
+               "node is used as backing hd of '%s'",
+               bdrv_get_device_or_node_name(parent));
+
+    bdrv_refresh_filename(backing_hd);
+
+    parent->open_flags &= ~BDRV_O_NO_BACKING;
+
+    bdrv_op_block_all(backing_hd, parent->backing_blocker);
+    /* Otherwise we won't be able to commit or stream */
+    bdrv_op_unblock(backing_hd, BLOCK_OP_TYPE_COMMIT_TARGET,
+                    parent->backing_blocker);
+    bdrv_op_unblock(backing_hd, BLOCK_OP_TYPE_STREAM,
+                    parent->backing_blocker);
+    /*
+     * We do backup in 3 ways:
+     * 1. drive backup
+     *    The target bs is new opened, and the source is top BDS
+     * 2. blockdev backup
+     *    Both the source and the target are top BDSes.
+     * 3. internal backup(used for block replication)
+     *    Both the source and the target are backing file
+     *
+     * In case 1 and 2, neither the source nor the target is the backing file.
+     * In case 3, we will block the top BDS, so there is only one block job
+     * for the top BDS and its backing chain.
+     */
+    bdrv_op_unblock(backing_hd, BLOCK_OP_TYPE_BACKUP_SOURCE,
+                    parent->backing_blocker);
+    bdrv_op_unblock(backing_hd, BLOCK_OP_TYPE_BACKUP_TARGET,
+                    parent->backing_blocker);
 }
 
-void bdrv_disable_copy_on_read(BlockDriverState *bs)
+static void bdrv_backing_detach(BdrvChild *c)
 {
-    assert(bs->copy_on_read > 0);
-    bs->copy_on_read--;
+    BlockDriverState *parent = c->opaque;
+
+    assert(parent->backing_blocker);
+    bdrv_op_unblock_all(c->bs, parent->backing_blocker);
+    error_free(parent->backing_blocker);
+    parent->backing_blocker = NULL;
 }
 
-static int bdrv_open_flags(BlockDriverState *bs, int flags)
+static int bdrv_backing_update_filename(BdrvChild *c, BlockDriverState *base,
+                                        const char *filename, Error **errp)
 {
-    int open_flags = flags | BDRV_O_CACHE_WB;
+    BlockDriverState *parent = c->opaque;
+    bool read_only = bdrv_is_read_only(parent);
+    int ret;
+
+    if (read_only) {
+        ret = bdrv_reopen_set_read_only(parent, false, errp);
+        if (ret < 0) {
+            return ret;
+        }
+    }
+
+    ret = bdrv_change_backing_file(parent, filename,
+                                   base->drv ? base->drv->format_name : "",
+                                   false);
+    if (ret < 0) {
+        error_setg_errno(errp, -ret, "Could not update backing file link");
+    }
+
+    if (read_only) {
+        bdrv_reopen_set_read_only(parent, true, NULL);
+    }
+
+    return ret;
+}
+
+/*
+ * Returns the options and flags that a generic child of a BDS should
+ * get, based on the given options and flags for the parent BDS.
+ */
+static void bdrv_inherited_options(BdrvChildRole role, bool parent_is_format,
+                                   int *child_flags, QDict *child_options,
+                                   int parent_flags, QDict *parent_options)
+{
+    int flags = parent_flags;
 
     /*
-     * Clear flags that are internal to the block layer before opening the
-     * image.
+     * First, decide whether to set, clear, or leave BDRV_O_PROTOCOL.
+     * Generally, the question to answer is: Should this child be
+     * format-probed by default?
+     */
+
+    /*
+     * Pure and non-filtered data children of non-format nodes should
+     * be probed by default (even when the node itself has BDRV_O_PROTOCOL
+     * set).  This only affects a very limited set of drivers (namely
+     * quorum and blkverify when this comment was written).
+     * Force-clear BDRV_O_PROTOCOL then.
+     */
+    if (!parent_is_format &&
+        (role & BDRV_CHILD_DATA) &&
+        !(role & (BDRV_CHILD_METADATA | BDRV_CHILD_FILTERED)))
+    {
+        flags &= ~BDRV_O_PROTOCOL;
+    }
+
+    /*
+     * All children of format nodes (except for COW children) and all
+     * metadata children in general should never be format-probed.
+     * Force-set BDRV_O_PROTOCOL then.
      */
-    open_flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
+    if ((parent_is_format && !(role & BDRV_CHILD_COW)) ||
+        (role & BDRV_CHILD_METADATA))
+    {
+        flags |= BDRV_O_PROTOCOL;
+    }
+
+    /*
+     * If the cache mode isn't explicitly set, inherit direct and no-flush from
+     * the parent.
+     */
+    qdict_copy_default(child_options, parent_options, BDRV_OPT_CACHE_DIRECT);
+    qdict_copy_default(child_options, parent_options, BDRV_OPT_CACHE_NO_FLUSH);
+    qdict_copy_default(child_options, parent_options, BDRV_OPT_FORCE_SHARE);
+
+    if (role & BDRV_CHILD_COW) {
+        /* backing files are opened read-only by default */
+        qdict_set_default_str(child_options, BDRV_OPT_READ_ONLY, "on");
+        qdict_set_default_str(child_options, BDRV_OPT_AUTO_READ_ONLY, "off");
+    } else {
+        /* Inherit the read-only option from the parent if it's not set */
+        qdict_copy_default(child_options, parent_options, BDRV_OPT_READ_ONLY);
+        qdict_copy_default(child_options, parent_options,
+                           BDRV_OPT_AUTO_READ_ONLY);
+    }
 
     /*
-     * Snapshots should be writable.
+     * bdrv_co_pdiscard() respects unmap policy for the parent, so we
+     * can default to enable it on lower layers regardless of the
+     * parent option.
      */
-    if (bs->is_temporary) {
-        open_flags |= BDRV_O_RDWR;
+    qdict_set_default_str(child_options, BDRV_OPT_DISCARD, "unmap");
+
+    /* Clear flags that only apply to the top layer */
+    flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING | BDRV_O_COPY_ON_READ);
+
+    if (role & BDRV_CHILD_METADATA) {
+        flags &= ~BDRV_O_NO_IO;
+    }
+    if (role & BDRV_CHILD_COW) {
+        flags &= ~BDRV_O_TEMPORARY;
+    }
+
+    *child_flags = flags;
+}
+
+static void bdrv_child_cb_attach(BdrvChild *child)
+{
+    BlockDriverState *bs = child->opaque;
+
+    if (child->role & BDRV_CHILD_COW) {
+        bdrv_backing_attach(child);
+    }
+
+    bdrv_apply_subtree_drain(child, bs);
+}
+
+static void bdrv_child_cb_detach(BdrvChild *child)
+{
+    BlockDriverState *bs = child->opaque;
+
+    if (child->role & BDRV_CHILD_COW) {
+        bdrv_backing_detach(child);
+    }
+
+    bdrv_unapply_subtree_drain(child, bs);
+}
+
+static int bdrv_child_cb_update_filename(BdrvChild *c, BlockDriverState *base,
+                                         const char *filename, Error **errp)
+{
+    if (c->role & BDRV_CHILD_COW) {
+        return bdrv_backing_update_filename(c, base, filename, errp);
     }
+    return 0;
+}
+
+const BdrvChildClass child_of_bds = {
+    .parent_is_bds   = true,
+    .get_parent_desc = bdrv_child_get_parent_desc,
+    .inherit_options = bdrv_inherited_options,
+    .drained_begin   = bdrv_child_cb_drained_begin,
+    .drained_poll    = bdrv_child_cb_drained_poll,
+    .drained_end     = bdrv_child_cb_drained_end,
+    .attach          = bdrv_child_cb_attach,
+    .detach          = bdrv_child_cb_detach,
+    .inactivate      = bdrv_child_cb_inactivate,
+    .can_set_aio_ctx = bdrv_child_cb_can_set_aio_ctx,
+    .set_aio_ctx     = bdrv_child_cb_set_aio_ctx,
+    .update_filename = bdrv_child_cb_update_filename,
+};
+
+static int bdrv_open_flags(BlockDriverState *bs, int flags)
+{
+    int open_flags = flags;
+
+    /*
+     * Clear flags that are internal to the block layer before opening the
+     * image.
+     */
+    open_flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING | BDRV_O_PROTOCOL);
 
     return open_flags;
 }
 
+static void update_flags_from_options(int *flags, QemuOpts *opts)
+{
+    *flags &= ~(BDRV_O_CACHE_MASK | BDRV_O_RDWR | BDRV_O_AUTO_RDONLY);
+
+    if (qemu_opt_get_bool_del(opts, BDRV_OPT_CACHE_NO_FLUSH, false)) {
+        *flags |= BDRV_O_NO_FLUSH;
+    }
+
+    if (qemu_opt_get_bool_del(opts, BDRV_OPT_CACHE_DIRECT, false)) {
+        *flags |= BDRV_O_NOCACHE;
+    }
+
+    if (!qemu_opt_get_bool_del(opts, BDRV_OPT_READ_ONLY, false)) {
+        *flags |= BDRV_O_RDWR;
+    }
+
+    if (qemu_opt_get_bool_del(opts, BDRV_OPT_AUTO_READ_ONLY, false)) {
+        *flags |= BDRV_O_AUTO_RDONLY;
+    }
+}
+
+static void update_options_from_flags(QDict *options, int flags)
+{
+    if (!qdict_haskey(options, BDRV_OPT_CACHE_DIRECT)) {
+        qdict_put_bool(options, BDRV_OPT_CACHE_DIRECT, flags & BDRV_O_NOCACHE);
+    }
+    if (!qdict_haskey(options, BDRV_OPT_CACHE_NO_FLUSH)) {
+        qdict_put_bool(options, BDRV_OPT_CACHE_NO_FLUSH,
+                       flags & BDRV_O_NO_FLUSH);
+    }
+    if (!qdict_haskey(options, BDRV_OPT_READ_ONLY)) {
+        qdict_put_bool(options, BDRV_OPT_READ_ONLY, !(flags & BDRV_O_RDWR));
+    }
+    if (!qdict_haskey(options, BDRV_OPT_AUTO_READ_ONLY)) {
+        qdict_put_bool(options, BDRV_OPT_AUTO_READ_ONLY,
+                       flags & BDRV_O_AUTO_RDONLY);
+    }
+}
+
+static void bdrv_assign_node_name(BlockDriverState *bs,
+                                  const char *node_name,
+                                  Error **errp)
+{
+    char *gen_node_name = NULL;
+
+    if (!node_name) {
+        node_name = gen_node_name = id_generate(ID_BLOCK);
+    } else if (!id_wellformed(node_name)) {
+        /*
+         * Check for empty string or invalid characters, but not if it is
+         * generated (generated names use characters not available to the user)
+         */
+        error_setg(errp, "Invalid node name");
+        return;
+    }
+
+    /* takes care of avoiding namespaces collisions */
+    if (blk_by_name(node_name)) {
+        error_setg(errp, "node-name=%s is conflicting with a device id",
+                   node_name);
+        goto out;
+    }
+
+    /* takes care of avoiding duplicates node names */
+    if (bdrv_find_node(node_name)) {
+        error_setg(errp, "Duplicate node name");
+        goto out;
+    }
+
+    /* Make sure that the node name isn't truncated */
+    if (strlen(node_name) >= sizeof(bs->node_name)) {
+        error_setg(errp, "Node name too long");
+        goto out;
+    }
+
+    /* copy node name into the bs and insert it into the graph list */
+    pstrcpy(bs->node_name, sizeof(bs->node_name), node_name);
+    QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs, node_list);
+out:
+    g_free(gen_node_name);
+}
+
+static int bdrv_open_driver(BlockDriverState *bs, BlockDriver *drv,
+                            const char *node_name, QDict *options,
+                            int open_flags, Error **errp)
+{
+    Error *local_err = NULL;
+    int i, ret;
+
+    bdrv_assign_node_name(bs, node_name, &local_err);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        return -EINVAL;
+    }
+
+    bs->drv = drv;
+    bs->read_only = !(bs->open_flags & BDRV_O_RDWR);
+    bs->opaque = g_malloc0(drv->instance_size);
+
+    if (drv->bdrv_file_open) {
+        assert(!drv->bdrv_needs_filename || bs->filename[0]);
+        ret = drv->bdrv_file_open(bs, options, open_flags, &local_err);
+    } else if (drv->bdrv_open) {
+        ret = drv->bdrv_open(bs, options, open_flags, &local_err);
+    } else {
+        ret = 0;
+    }
+
+    if (ret < 0) {
+        if (local_err) {
+            error_propagate(errp, local_err);
+        } else if (bs->filename[0]) {
+            error_setg_errno(errp, -ret, "Could not open '%s'", bs->filename);
+        } else {
+            error_setg_errno(errp, -ret, "Could not open image");
+        }
+        goto open_failed;
+    }
+
+    ret = refresh_total_sectors(bs, bs->total_sectors);
+    if (ret < 0) {
+        error_setg_errno(errp, -ret, "Could not refresh total sector count");
+        return ret;
+    }
+
+    bdrv_refresh_limits(bs, &local_err);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        return -EINVAL;
+    }
+
+    assert(bdrv_opt_mem_align(bs) != 0);
+    assert(bdrv_min_mem_align(bs) != 0);
+    assert(is_power_of_2(bs->bl.request_alignment));
+
+    for (i = 0; i < bs->quiesce_counter; i++) {
+        if (drv->bdrv_co_drain_begin) {
+            drv->bdrv_co_drain_begin(bs);
+        }
+    }
+
+    return 0;
+open_failed:
+    bs->drv = NULL;
+    if (bs->file != NULL) {
+        bdrv_unref_child(bs, bs->file);
+        bs->file = NULL;
+    }
+    g_free(bs->opaque);
+    bs->opaque = NULL;
+    return ret;
+}
+
+BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name,
+                                       int flags, Error **errp)
+{
+    BlockDriverState *bs;
+    int ret;
+
+    bs = bdrv_new();
+    bs->open_flags = flags;
+    bs->explicit_options = qdict_new();
+    bs->options = qdict_new();
+    bs->opaque = NULL;
+
+    update_options_from_flags(bs->options, flags);
+
+    ret = bdrv_open_driver(bs, drv, node_name, bs->options, flags, errp);
+    if (ret < 0) {
+        qobject_unref(bs->explicit_options);
+        bs->explicit_options = NULL;
+        qobject_unref(bs->options);
+        bs->options = NULL;
+        bdrv_unref(bs);
+        return NULL;
+    }
+
+    return bs;
+}
+
+QemuOptsList bdrv_runtime_opts = {
+    .name = "bdrv_common",
+    .head = QTAILQ_HEAD_INITIALIZER(bdrv_runtime_opts.head),
+    .desc = {
+        {
+            .name = "node-name",
+            .type = QEMU_OPT_STRING,
+            .help = "Node name of the block device node",
+        },
+        {
+            .name = "driver",
+            .type = QEMU_OPT_STRING,
+            .help = "Block driver to use for the node",
+        },
+        {
+            .name = BDRV_OPT_CACHE_DIRECT,
+            .type = QEMU_OPT_BOOL,
+            .help = "Bypass software writeback cache on the host",
+        },
+        {
+            .name = BDRV_OPT_CACHE_NO_FLUSH,
+            .type = QEMU_OPT_BOOL,
+            .help = "Ignore flush requests",
+        },
+        {
+            .name = BDRV_OPT_READ_ONLY,
+            .type = QEMU_OPT_BOOL,
+            .help = "Node is opened in read-only mode",
+        },
+        {
+            .name = BDRV_OPT_AUTO_READ_ONLY,
+            .type = QEMU_OPT_BOOL,
+            .help = "Node can become read-only if opening read-write fails",
+        },
+        {
+            .name = "detect-zeroes",
+            .type = QEMU_OPT_STRING,
+            .help = "try to optimize zero writes (off, on, unmap)",
+        },
+        {
+            .name = BDRV_OPT_DISCARD,
+            .type = QEMU_OPT_STRING,
+            .help = "discard operation (ignore/off, unmap/on)",
+        },
+        {
+            .name = BDRV_OPT_FORCE_SHARE,
+            .type = QEMU_OPT_BOOL,
+            .help = "always accept other writers (default: off)",
+        },
+        { /* end of list */ }
+    },
+};
+
+QemuOptsList bdrv_create_opts_simple = {
+    .name = "simple-create-opts",
+    .head = QTAILQ_HEAD_INITIALIZER(bdrv_create_opts_simple.head),
+    .desc = {
+        {
+            .name = BLOCK_OPT_SIZE,
+            .type = QEMU_OPT_SIZE,
+            .help = "Virtual disk size"
+        },
+        {
+            .name = BLOCK_OPT_PREALLOC,
+            .type = QEMU_OPT_STRING,
+            .help = "Preallocation mode (allowed values: off)"
+        },
+        { /* end of list */ }
+    }
+};
+
 /*
  * Common part for opening disk images and files
  *
  * Removes all processed options from *options.
  */
-static int bdrv_open_common(BlockDriverState *bs, BlockDriverState *file,
-    QDict *options, int flags, BlockDriver *drv)
+static int bdrv_open_common(BlockDriverState *bs, BlockBackend *file,
+                            QDict *options, Error **errp)
 {
     int ret, open_flags;
     const char *filename;
+    const char *driver_name = NULL;
+    const char *node_name = NULL;
+    const char *discard;
+    QemuOpts *opts;
+    BlockDriver *drv;
+    Error *local_err = NULL;
 
-    assert(drv != NULL);
     assert(bs->file == NULL);
     assert(options != NULL && bs->options != options);
 
+    opts = qemu_opts_create(&bdrv_runtime_opts, NULL, 0, &error_abort);
+    if (!qemu_opts_absorb_qdict(opts, options, errp)) {
+        ret = -EINVAL;
+        goto fail_opts;
+    }
+
+    update_flags_from_options(&bs->open_flags, opts);
+
+    driver_name = qemu_opt_get(opts, "driver");
+    drv = bdrv_find_format(driver_name);
+    assert(drv != NULL);
+
+    bs->force_share = qemu_opt_get_bool(opts, BDRV_OPT_FORCE_SHARE, false);
+
+    if (bs->force_share && (bs->open_flags & BDRV_O_RDWR)) {
+        error_setg(errp,
+                   BDRV_OPT_FORCE_SHARE
+                   "=on can only be used with read-only images");
+        ret = -EINVAL;
+        goto fail_opts;
+    }
+
     if (file != NULL) {
-        filename = file->filename;
+        bdrv_refresh_filename(blk_bs(file));
+        filename = blk_bs(file)->filename;
     } else {
+        /*
+         * Caution: while qdict_get_try_str() is fine, getting
+         * non-string types would require more care.  When @options
+         * come from -blockdev or blockdev_add, its members are typed
+         * according to the QAPI schema, but when they come from
+         * -drive, they're all QString.
+         */
         filename = qdict_get_try_str(options, "filename");
     }
 
-    trace_bdrv_open_common(bs, filename ?: "", flags, drv->format_name);
-
-    /* bdrv_open() with directly using a protocol as drv. This layer is already
-     * opened, so assign it to bs (while file becomes a closed BlockDriverState)
-     * and return immediately. */
-    if (file != NULL && drv->bdrv_file_open) {
-        bdrv_swap(file, bs);
-        return 0;
+    if (drv->bdrv_needs_filename && (!filename || !filename[0])) {
+        error_setg(errp, "The '%s' block driver requires a file name",
+                   drv->format_name);
+        ret = -EINVAL;
+        goto fail_opts;
     }
 
-    bs->open_flags = flags;
-    bs->buffer_alignment = 512;
-    open_flags = bdrv_open_flags(bs, flags);
-    bs->read_only = !(open_flags & BDRV_O_RDWR);
+    trace_bdrv_open_common(bs, filename ?: "", bs->open_flags,
+                           drv->format_name);
+
+    bs->read_only = !(bs->open_flags & BDRV_O_RDWR);
 
     if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv, bs->read_only)) {
-        return -ENOTSUP;
+        if (!bs->read_only && bdrv_is_whitelisted(drv, true)) {
+            ret = bdrv_apply_auto_read_only(bs, NULL, NULL);
+        } else {
+            ret = -ENOTSUP;
+        }
+        if (ret < 0) {
+            error_setg(errp,
+                       !bs->read_only && bdrv_is_whitelisted(drv, true)
+                       ? "Driver '%s' can only be used for read-only devices"
+                       : "Driver '%s' is not whitelisted",
+                       drv->format_name);
+            goto fail_opts;
+        }
+    }
+
+    /* bdrv_new() and bdrv_close() make it so */
+    assert(qatomic_read(&bs->copy_on_read) == 0);
+
+    if (bs->open_flags & BDRV_O_COPY_ON_READ) {
+        if (!bs->read_only) {
+            bdrv_enable_copy_on_read(bs);
+        } else {
+            error_setg(errp, "Can't use copy-on-read on read-only device");
+            ret = -EINVAL;
+            goto fail_opts;
+        }
+    }
+
+    discard = qemu_opt_get(opts, BDRV_OPT_DISCARD);
+    if (discard != NULL) {
+        if (bdrv_parse_discard_flags(discard, &bs->open_flags) != 0) {
+            error_setg(errp, "Invalid discard option");
+            ret = -EINVAL;
+            goto fail_opts;
+        }
     }
 
-    assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */
-    if (!bs->read_only && (flags & BDRV_O_COPY_ON_READ)) {
-        bdrv_enable_copy_on_read(bs);
+    bs->detect_zeroes =
+        bdrv_parse_detect_zeroes(opts, bs->open_flags, &local_err);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        ret = -EINVAL;
+        goto fail_opts;
     }
 
     if (filename != NULL) {
@@ -723,235 +1729,1488 @@ static int bdrv_open_common(BlockDriverState *bs, BlockDriverState *file,
     } else {
         bs->filename[0] = '\0';
     }
+    pstrcpy(bs->exact_filename, sizeof(bs->exact_filename), bs->filename);
 
-    bs->drv = drv;
-    bs->opaque = g_malloc0(drv->instance_size);
+    /* Open the image, either directly or using a protocol */
+    open_flags = bdrv_open_flags(bs, bs->open_flags);
+    node_name = qemu_opt_get(opts, "node-name");
 
-    bs->enable_write_cache = !!(flags & BDRV_O_CACHE_WB);
+    assert(!drv->bdrv_file_open || file == NULL);
+    ret = bdrv_open_driver(bs, drv, node_name, options, open_flags, errp);
+    if (ret < 0) {
+        goto fail_opts;
+    }
 
-    /* Open the image, either directly or using a protocol */
-    if (drv->bdrv_file_open) {
-        assert(file == NULL);
-        assert(drv->bdrv_parse_filename || filename != NULL);
-        ret = drv->bdrv_file_open(bs, options, open_flags);
-    } else {
-        if (file == NULL) {
-            qerror_report(ERROR_CLASS_GENERIC_ERROR, "Can't use '%s' as a "
-                          "block driver for the protocol level",
-                          drv->format_name);
-            ret = -EINVAL;
-            goto free_and_fail;
+    qemu_opts_del(opts);
+    return 0;
+
+fail_opts:
+    qemu_opts_del(opts);
+    return ret;
+}
+
+static QDict *parse_json_filename(const char *filename, Error **errp)
+{
+    QObject *options_obj;
+    QDict *options;
+    int ret;
+
+    ret = strstart(filename, "json:", &filename);
+    assert(ret);
+
+    options_obj = qobject_from_json(filename, errp);
+    if (!options_obj) {
+        error_prepend(errp, "Could not parse the JSON options: ");
+        return NULL;
+    }
+
+    options = qobject_to(QDict, options_obj);
+    if (!options) {
+        qobject_unref(options_obj);
+        error_setg(errp, "Invalid JSON object given");
+        return NULL;
+    }
+
+    qdict_flatten(options);
+
+    return options;
+}
+
+static void parse_json_protocol(QDict *options, const char **pfilename,
+                                Error **errp)
+{
+    QDict *json_options;
+    Error *local_err = NULL;
+
+    /* Parse json: pseudo-protocol */
+    if (!*pfilename || !g_str_has_prefix(*pfilename, "json:")) {
+        return;
+    }
+
+    json_options = parse_json_filename(*pfilename, &local_err);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        return;
+    }
+
+    /* Options given in the filename have lower priority than options
+     * specified directly */
+    qdict_join(options, json_options, false);
+    qobject_unref(json_options);
+    *pfilename = NULL;
+}
+
+/*
+ * Fills in default options for opening images and converts the legacy
+ * filename/flags pair to option QDict entries.
+ * The BDRV_O_PROTOCOL flag in *flags will be set or cleared accordingly if a
+ * block driver has been specified explicitly.
+ */
+static int bdrv_fill_options(QDict **options, const char *filename,
+                             int *flags, Error **errp)
+{
+    const char *drvname;
+    bool protocol = *flags & BDRV_O_PROTOCOL;
+    bool parse_filename = false;
+    BlockDriver *drv = NULL;
+    Error *local_err = NULL;
+
+    /*
+     * Caution: while qdict_get_try_str() is fine, getting non-string
+     * types would require more care.  When @options come from
+     * -blockdev or blockdev_add, its members are typed according to
+     * the QAPI schema, but when they come from -drive, they're all
+     * QString.
+     */
+    drvname = qdict_get_try_str(*options, "driver");
+    if (drvname) {
+        drv = bdrv_find_format(drvname);
+        if (!drv) {
+            error_setg(errp, "Unknown driver '%s'", drvname);
+            return -ENOENT;
         }
-        assert(file != NULL);
-        bs->file = file;
-        ret = drv->bdrv_open(bs, options, open_flags);
+        /* If the user has explicitly specified the driver, this choice should
+         * override the BDRV_O_PROTOCOL flag */
+        protocol = drv->bdrv_file_open;
     }
 
-    if (ret < 0) {
-        goto free_and_fail;
+    if (protocol) {
+        *flags |= BDRV_O_PROTOCOL;
+    } else {
+        *flags &= ~BDRV_O_PROTOCOL;
     }
 
-    ret = refresh_total_sectors(bs, bs->total_sectors);
-    if (ret < 0) {
-        goto free_and_fail;
+    /* Translate cache options from flags into options */
+    update_options_from_flags(*options, *flags);
+
+    /* Fetch the file name from the options QDict if necessary */
+    if (protocol && filename) {
+        if (!qdict_haskey(*options, "filename")) {
+            qdict_put_str(*options, "filename", filename);
+            parse_filename = true;
+        } else {
+            error_setg(errp, "Can't specify 'file' and 'filename' options at "
+                             "the same time");
+            return -EINVAL;
+        }
     }
 
-#ifndef _WIN32
-    if (bs->is_temporary) {
-        assert(filename != NULL);
-        unlink(filename);
+    /* Find the right block driver */
+    /* See cautionary note on accessing @options above */
+    filename = qdict_get_try_str(*options, "filename");
+
+    if (!drvname && protocol) {
+        if (filename) {
+            drv = bdrv_find_protocol(filename, parse_filename, errp);
+            if (!drv) {
+                return -EINVAL;
+            }
+
+            drvname = drv->format_name;
+            qdict_put_str(*options, "driver", drvname);
+        } else {
+            error_setg(errp, "Must specify either driver or file");
+            return -EINVAL;
+        }
     }
-#endif
+
+    assert(drv || !protocol);
+
+    /* Driver-specific filename parsing */
+    if (drv && drv->bdrv_parse_filename && parse_filename) {
+        drv->bdrv_parse_filename(filename, *options, &local_err);
+        if (local_err) {
+            error_propagate(errp, local_err);
+            return -EINVAL;
+        }
+
+        if (!drv->bdrv_needs_filename) {
+            qdict_del(*options, "filename");
+        }
+    }
+
     return 0;
+}
 
-free_and_fail:
-    bs->file = NULL;
-    g_free(bs->opaque);
-    bs->opaque = NULL;
-    bs->drv = NULL;
-    return ret;
+static int bdrv_child_check_perm(BdrvChild *c, BlockReopenQueue *q,
+                                 uint64_t perm, uint64_t shared,
+                                 GSList *ignore_children,
+                                 bool *tighten_restrictions, Error **errp);
+static void bdrv_child_abort_perm_update(BdrvChild *c);
+static void bdrv_child_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared);
+
+typedef struct BlockReopenQueueEntry {
+     bool prepared;
+     bool perms_checked;
+     BDRVReopenState state;
+     QTAILQ_ENTRY(BlockReopenQueueEntry) entry;
+} BlockReopenQueueEntry;
+
+/*
+ * Return the flags that @bs will have after the reopens in @q have
+ * successfully completed. If @q is NULL (or @bs is not contained in @q),
+ * return the current flags.
+ */
+static int bdrv_reopen_get_flags(BlockReopenQueue *q, BlockDriverState *bs)
+{
+    BlockReopenQueueEntry *entry;
+
+    if (q != NULL) {
+        QTAILQ_FOREACH(entry, q, entry) {
+            if (entry->state.bs == bs) {
+                return entry->state.flags;
+            }
+        }
+    }
+
+    return bs->open_flags;
+}
+
+/* Returns whether the image file can be written to after the reopen queue @q
+ * has been successfully applied, or right now if @q is NULL. */
+static bool bdrv_is_writable_after_reopen(BlockDriverState *bs,
+                                          BlockReopenQueue *q)
+{
+    int flags = bdrv_reopen_get_flags(q, bs);
+
+    return (flags & (BDRV_O_RDWR | BDRV_O_INACTIVE)) == BDRV_O_RDWR;
 }
 
 /*
- * Opens a file using a protocol (file, host_device, nbd, ...)
+ * Return whether the BDS can be written to.  This is not necessarily
+ * the same as !bdrv_is_read_only(bs), as inactivated images may not
+ * be written to but do not count as read-only images.
+ */
+bool bdrv_is_writable(BlockDriverState *bs)
+{
+    return bdrv_is_writable_after_reopen(bs, NULL);
+}
+
+static void bdrv_child_perm(BlockDriverState *bs, BlockDriverState *child_bs,
+                            BdrvChild *c, BdrvChildRole role,
+                            BlockReopenQueue *reopen_queue,
+                            uint64_t parent_perm, uint64_t parent_shared,
+                            uint64_t *nperm, uint64_t *nshared)
+{
+    assert(bs->drv && bs->drv->bdrv_child_perm);
+    bs->drv->bdrv_child_perm(bs, c, role, reopen_queue,
+                             parent_perm, parent_shared,
+                             nperm, nshared);
+    /* TODO Take force_share from reopen_queue */
+    if (child_bs && child_bs->force_share) {
+        *nshared = BLK_PERM_ALL;
+    }
+}
+
+/*
+ * Check whether permissions on this node can be changed in a way that
+ * @cumulative_perms and @cumulative_shared_perms are the new cumulative
+ * permissions of all its parents. This involves checking whether all necessary
+ * permission changes to child nodes can be performed.
  *
- * options is a QDict of options to pass to the block drivers, or NULL for an
- * empty set of options. The reference to the QDict belongs to the block layer
- * after the call (even on failure), so if the caller intends to reuse the
- * dictionary, it needs to use QINCREF() before calling bdrv_file_open.
+ * Will set *tighten_restrictions to true if and only if new permissions have to
+ * be taken or currently shared permissions are to be unshared.  Otherwise,
+ * errors are not fatal as long as the caller accepts that the restrictions
+ * remain tighter than they need to be.  The caller still has to abort the
+ * transaction.
+ * @tighten_restrictions cannot be used together with @q: When reopening, we may
+ * encounter fatal errors even though no restrictions are to be tightened.  For
+ * example, changing a node from RW to RO will fail if the WRITE permission is
+ * to be kept.
+ *
+ * A call to this function must always be followed by a call to bdrv_set_perm()
+ * or bdrv_abort_perm_update().
  */
-int bdrv_file_open(BlockDriverState **pbs, const char *filename,
-                   QDict *options, int flags)
+static int bdrv_check_perm(BlockDriverState *bs, BlockReopenQueue *q,
+                           uint64_t cumulative_perms,
+                           uint64_t cumulative_shared_perms,
+                           GSList *ignore_children,
+                           bool *tighten_restrictions, Error **errp)
 {
-    BlockDriverState *bs;
-    BlockDriver *drv;
-    const char *drvname;
-    bool allow_protocol_prefix = false;
+    BlockDriver *drv = bs->drv;
+    BdrvChild *c;
     int ret;
 
-    /* NULL means an empty set of options */
-    if (options == NULL) {
-        options = qdict_new();
+    assert(!q || !tighten_restrictions);
+
+    if (tighten_restrictions) {
+        uint64_t current_perms, current_shared;
+        uint64_t added_perms, removed_shared_perms;
+
+        bdrv_get_cumulative_perm(bs, &current_perms, &current_shared);
+
+        added_perms = cumulative_perms & ~current_perms;
+        removed_shared_perms = current_shared & ~cumulative_shared_perms;
+
+        *tighten_restrictions = added_perms || removed_shared_perms;
     }
 
-    bs = bdrv_new("");
-    bs->options = options;
-    options = qdict_clone_shallow(options);
+    /* Write permissions never work with read-only images */
+    if ((cumulative_perms & (BLK_PERM_WRITE | BLK_PERM_WRITE_UNCHANGED)) &&
+        !bdrv_is_writable_after_reopen(bs, q))
+    {
+        if (!bdrv_is_writable_after_reopen(bs, NULL)) {
+            error_setg(errp, "Block node is read-only");
+        } else {
+            uint64_t current_perms, current_shared;
+            bdrv_get_cumulative_perm(bs, &current_perms, &current_shared);
+            if (current_perms & (BLK_PERM_WRITE | BLK_PERM_WRITE_UNCHANGED)) {
+                error_setg(errp, "Cannot make block node read-only, there is "
+                           "a writer on it");
+            } else {
+                error_setg(errp, "Cannot make block node read-only and create "
+                           "a writer on it");
+            }
+        }
 
-    /* Fetch the file name from the options QDict if necessary */
-    if (!filename) {
-        filename = qdict_get_try_str(options, "filename");
-    } else if (filename && !qdict_haskey(options, "filename")) {
-        qdict_put(options, "filename", qstring_from_str(filename));
-        allow_protocol_prefix = true;
-    } else {
-        qerror_report(ERROR_CLASS_GENERIC_ERROR, "Can't specify 'file' and "
-                      "'filename' options at the same time");
-        ret = -EINVAL;
-        goto fail;
+        return -EPERM;
     }
 
-    /* Find the right block driver */
-    drvname = qdict_get_try_str(options, "driver");
-    if (drvname) {
-        drv = bdrv_find_whitelisted_format(drvname, !(flags & BDRV_O_RDWR));
-        qdict_del(options, "driver");
-    } else if (filename) {
-        drv = bdrv_find_protocol(filename, allow_protocol_prefix);
-        if (!drv) {
-            qerror_report(ERROR_CLASS_GENERIC_ERROR, "Unknown protocol");
+    /*
+     * Unaligned requests will automatically be aligned to bl.request_alignment
+     * and without RESIZE we can't extend requests to write to space beyond the
+     * end of the image, so it's required that the image size is aligned.
+     */
+    if ((cumulative_perms & (BLK_PERM_WRITE | BLK_PERM_WRITE_UNCHANGED)) &&
+        !(cumulative_perms & BLK_PERM_RESIZE))
+    {
+        if ((bs->total_sectors * BDRV_SECTOR_SIZE) % bs->bl.request_alignment) {
+            error_setg(errp, "Cannot get 'write' permission without 'resize': "
+                             "Image size is not a multiple of request "
+                             "alignment");
+            return -EPERM;
         }
-    } else {
-        qerror_report(ERROR_CLASS_GENERIC_ERROR,
-                      "Must specify either driver or file");
-        drv = NULL;
     }
 
+    /* Check this node */
     if (!drv) {
-        ret = -ENOENT;
-        goto fail;
+        return 0;
     }
 
-    /* Parse the filename and open it */
-    if (drv->bdrv_parse_filename && filename) {
-        Error *local_err = NULL;
-        drv->bdrv_parse_filename(filename, options, &local_err);
-        if (error_is_set(&local_err)) {
-            qerror_report_err(local_err);
-            error_free(local_err);
-            ret = -EINVAL;
-            goto fail;
+    if (drv->bdrv_check_perm) {
+        return drv->bdrv_check_perm(bs, cumulative_perms,
+                                    cumulative_shared_perms, errp);
+    }
+
+    /* Drivers that never have children can omit .bdrv_child_perm() */
+    if (!drv->bdrv_child_perm) {
+        assert(QLIST_EMPTY(&bs->children));
+        return 0;
+    }
+
+    /* Check all children */
+    QLIST_FOREACH(c, &bs->children, next) {
+        uint64_t cur_perm, cur_shared;
+        bool child_tighten_restr;
+
+        bdrv_child_perm(bs, c->bs, c, c->role, q,
+                        cumulative_perms, cumulative_shared_perms,
+                        &cur_perm, &cur_shared);
+        ret = bdrv_child_check_perm(c, q, cur_perm, cur_shared, ignore_children,
+                                    tighten_restrictions ? &child_tighten_restr
+                                                         : NULL,
+                                    errp);
+        if (tighten_restrictions) {
+            *tighten_restrictions |= child_tighten_restr;
         }
-        qdict_del(options, "filename");
-    } else if (!drv->bdrv_parse_filename && !filename) {
-        qerror_report(ERROR_CLASS_GENERIC_ERROR,
-                      "The '%s' block driver requires a file name",
-                      drv->format_name);
-        ret = -EINVAL;
-        goto fail;
+        if (ret < 0) {
+            return ret;
+        }
+    }
+
+    return 0;
+}
+
+/*
+ * Notifies drivers that after a previous bdrv_check_perm() call, the
+ * permission update is not performed and any preparations made for it (e.g.
+ * taken file locks) need to be undone.
+ *
+ * This function recursively notifies all child nodes.
+ */
+static void bdrv_abort_perm_update(BlockDriverState *bs)
+{
+    BlockDriver *drv = bs->drv;
+    BdrvChild *c;
+
+    if (!drv) {
+        return;
+    }
+
+    if (drv->bdrv_abort_perm_update) {
+        drv->bdrv_abort_perm_update(bs);
+    }
+
+    QLIST_FOREACH(c, &bs->children, next) {
+        bdrv_child_abort_perm_update(c);
+    }
+}
+
+static void bdrv_set_perm(BlockDriverState *bs, uint64_t cumulative_perms,
+                          uint64_t cumulative_shared_perms)
+{
+    BlockDriver *drv = bs->drv;
+    BdrvChild *c;
+
+    if (!drv) {
+        return;
+    }
+
+    /* Update this node */
+    if (drv->bdrv_set_perm) {
+        drv->bdrv_set_perm(bs, cumulative_perms, cumulative_shared_perms);
+    }
+
+    /* Drivers that never have children can omit .bdrv_child_perm() */
+    if (!drv->bdrv_child_perm) {
+        assert(QLIST_EMPTY(&bs->children));
+        return;
+    }
+
+    /* Update all children */
+    QLIST_FOREACH(c, &bs->children, next) {
+        uint64_t cur_perm, cur_shared;
+        bdrv_child_perm(bs, c->bs, c, c->role, NULL,
+                        cumulative_perms, cumulative_shared_perms,
+                        &cur_perm, &cur_shared);
+        bdrv_child_set_perm(c, cur_perm, cur_shared);
+    }
+}
+
+void bdrv_get_cumulative_perm(BlockDriverState *bs, uint64_t *perm,
+                              uint64_t *shared_perm)
+{
+    BdrvChild *c;
+    uint64_t cumulative_perms = 0;
+    uint64_t cumulative_shared_perms = BLK_PERM_ALL;
+
+    QLIST_FOREACH(c, &bs->parents, next_parent) {
+        cumulative_perms |= c->perm;
+        cumulative_shared_perms &= c->shared_perm;
+    }
+
+    *perm = cumulative_perms;
+    *shared_perm = cumulative_shared_perms;
+}
+
+static char *bdrv_child_user_desc(BdrvChild *c)
+{
+    if (c->klass->get_parent_desc) {
+        return c->klass->get_parent_desc(c);
+    }
+
+    return g_strdup("another user");
+}
+
+char *bdrv_perm_names(uint64_t perm)
+{
+    struct perm_name {
+        uint64_t perm;
+        const char *name;
+    } permissions[] = {
+        { BLK_PERM_CONSISTENT_READ, "consistent read" },
+        { BLK_PERM_WRITE,           "write" },
+        { BLK_PERM_WRITE_UNCHANGED, "write unchanged" },
+        { BLK_PERM_RESIZE,          "resize" },
+        { BLK_PERM_GRAPH_MOD,       "change children" },
+        { 0, NULL }
+    };
+
+    GString *result = g_string_sized_new(30);
+    struct perm_name *p;
+
+    for (p = permissions; p->name; p++) {
+        if (perm & p->perm) {
+            if (result->len > 0) {
+                g_string_append(result, ", ");
+            }
+            g_string_append(result, p->name);
+        }
+    }
+
+    return g_string_free(result, FALSE);
+}
+
+/*
+ * Checks whether a new reference to @bs can be added if the new user requires
+ * @new_used_perm/@new_shared_perm as its permissions. If @ignore_children is
+ * set, the BdrvChild objects in this list are ignored in the calculations;
+ * this allows checking permission updates for an existing reference.
+ *
+ * See bdrv_check_perm() for the semantics of @tighten_restrictions.
+ *
+ * Needs to be followed by a call to either bdrv_set_perm() or
+ * bdrv_abort_perm_update(). */
+static int bdrv_check_update_perm(BlockDriverState *bs, BlockReopenQueue *q,
+                                  uint64_t new_used_perm,
+                                  uint64_t new_shared_perm,
+                                  GSList *ignore_children,
+                                  bool *tighten_restrictions,
+                                  Error **errp)
+{
+    BdrvChild *c;
+    uint64_t cumulative_perms = new_used_perm;
+    uint64_t cumulative_shared_perms = new_shared_perm;
+
+    assert(!q || !tighten_restrictions);
+
+    /* There is no reason why anyone couldn't tolerate write_unchanged */
+    assert(new_shared_perm & BLK_PERM_WRITE_UNCHANGED);
+
+    QLIST_FOREACH(c, &bs->parents, next_parent) {
+        if (g_slist_find(ignore_children, c)) {
+            continue;
+        }
+
+        if ((new_used_perm & c->shared_perm) != new_used_perm) {
+            char *user = bdrv_child_user_desc(c);
+            char *perm_names = bdrv_perm_names(new_used_perm & ~c->shared_perm);
+
+            if (tighten_restrictions) {
+                *tighten_restrictions = true;
+            }
+
+            error_setg(errp, "Conflicts with use by %s as '%s', which does not "
+                             "allow '%s' on %s",
+                       user, c->name, perm_names, bdrv_get_node_name(c->bs));
+            g_free(user);
+            g_free(perm_names);
+            return -EPERM;
+        }
+
+        if ((c->perm & new_shared_perm) != c->perm) {
+            char *user = bdrv_child_user_desc(c);
+            char *perm_names = bdrv_perm_names(c->perm & ~new_shared_perm);
+
+            if (tighten_restrictions) {
+                *tighten_restrictions = true;
+            }
+
+            error_setg(errp, "Conflicts with use by %s as '%s', which uses "
+                             "'%s' on %s",
+                       user, c->name, perm_names, bdrv_get_node_name(c->bs));
+            g_free(user);
+            g_free(perm_names);
+            return -EPERM;
+        }
+
+        cumulative_perms |= c->perm;
+        cumulative_shared_perms &= c->shared_perm;
     }
 
-    ret = bdrv_open_common(bs, NULL, options, flags, drv);
+    return bdrv_check_perm(bs, q, cumulative_perms, cumulative_shared_perms,
+                           ignore_children, tighten_restrictions, errp);
+}
+
+/* Needs to be followed by a call to either bdrv_child_set_perm() or
+ * bdrv_child_abort_perm_update(). */
+static int bdrv_child_check_perm(BdrvChild *c, BlockReopenQueue *q,
+                                 uint64_t perm, uint64_t shared,
+                                 GSList *ignore_children,
+                                 bool *tighten_restrictions, Error **errp)
+{
+    int ret;
+
+    ignore_children = g_slist_prepend(g_slist_copy(ignore_children), c);
+    ret = bdrv_check_update_perm(c->bs, q, perm, shared, ignore_children,
+                                 tighten_restrictions, errp);
+    g_slist_free(ignore_children);
+
     if (ret < 0) {
-        goto fail;
+        return ret;
     }
 
-    /* Check if any unknown options were used */
-    if (qdict_size(options) != 0) {
-        const QDictEntry *entry = qdict_first(options);
-        qerror_report(ERROR_CLASS_GENERIC_ERROR, "Block protocol '%s' doesn't "
-                      "support the option '%s'",
-                      drv->format_name, entry->key);
-        ret = -EINVAL;
-        goto fail;
+    if (!c->has_backup_perm) {
+        c->has_backup_perm = true;
+        c->backup_perm = c->perm;
+        c->backup_shared_perm = c->shared_perm;
     }
-    QDECREF(options);
+    /*
+     * Note: it's OK if c->has_backup_perm was already set, as we can find the
+     * same child twice during check_perm procedure
+     */
+
+    c->perm = perm;
+    c->shared_perm = shared;
 
-    bs->growable = 1;
-    *pbs = bs;
     return 0;
+}
 
-fail:
-    QDECREF(options);
-    if (!bs->drv) {
-        QDECREF(bs->options);
+static void bdrv_child_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared)
+{
+    uint64_t cumulative_perms, cumulative_shared_perms;
+
+    c->has_backup_perm = false;
+
+    c->perm = perm;
+    c->shared_perm = shared;
+
+    bdrv_get_cumulative_perm(c->bs, &cumulative_perms,
+                             &cumulative_shared_perms);
+    bdrv_set_perm(c->bs, cumulative_perms, cumulative_shared_perms);
+}
+
+static void bdrv_child_abort_perm_update(BdrvChild *c)
+{
+    if (c->has_backup_perm) {
+        c->perm = c->backup_perm;
+        c->shared_perm = c->backup_shared_perm;
+        c->has_backup_perm = false;
     }
-    bdrv_delete(bs);
-    return ret;
+
+    bdrv_abort_perm_update(c->bs);
+}
+
+int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared,
+                            Error **errp)
+{
+    Error *local_err = NULL;
+    int ret;
+    bool tighten_restrictions;
+
+    ret = bdrv_child_check_perm(c, NULL, perm, shared, NULL,
+                                &tighten_restrictions, &local_err);
+    if (ret < 0) {
+        bdrv_child_abort_perm_update(c);
+        if (tighten_restrictions) {
+            error_propagate(errp, local_err);
+        } else {
+            /*
+             * Our caller may intend to only loosen restrictions and
+             * does not expect this function to fail.  Errors are not
+             * fatal in such a case, so we can just hide them from our
+             * caller.
+             */
+            error_free(local_err);
+            ret = 0;
+        }
+        return ret;
+    }
+
+    bdrv_child_set_perm(c, perm, shared);
+
+    return 0;
+}
+
+int bdrv_child_refresh_perms(BlockDriverState *bs, BdrvChild *c, Error **errp)
+{
+    uint64_t parent_perms, parent_shared;
+    uint64_t perms, shared;
+
+    bdrv_get_cumulative_perm(bs, &parent_perms, &parent_shared);
+    bdrv_child_perm(bs, c->bs, c, c->role, NULL,
+                    parent_perms, parent_shared, &perms, &shared);
+
+    return bdrv_child_try_set_perm(c, perms, shared, errp);
+}
+
+/*
+ * Default implementation for .bdrv_child_perm() for block filters:
+ * Forward CONSISTENT_READ, WRITE, WRITE_UNCHANGED, and RESIZE to the
+ * filtered child.
+ */
+static void bdrv_filter_default_perms(BlockDriverState *bs, BdrvChild *c,
+                                      BdrvChildRole role,
+                                      BlockReopenQueue *reopen_queue,
+                                      uint64_t perm, uint64_t shared,
+                                      uint64_t *nperm, uint64_t *nshared)
+{
+    *nperm = perm & DEFAULT_PERM_PASSTHROUGH;
+    *nshared = (shared & DEFAULT_PERM_PASSTHROUGH) | DEFAULT_PERM_UNCHANGED;
+}
+
+static void bdrv_default_perms_for_cow(BlockDriverState *bs, BdrvChild *c,
+                                       BdrvChildRole role,
+                                       BlockReopenQueue *reopen_queue,
+                                       uint64_t perm, uint64_t shared,
+                                       uint64_t *nperm, uint64_t *nshared)
+{
+    assert(role & BDRV_CHILD_COW);
+
+    /*
+     * We want consistent read from backing files if the parent needs it.
+     * No other operations are performed on backing files.
+     */
+    perm &= BLK_PERM_CONSISTENT_READ;
+
+    /*
+     * If the parent can deal with changing data, we're okay with a
+     * writable and resizable backing file.
+     * TODO Require !(perm & BLK_PERM_CONSISTENT_READ), too?
+     */
+    if (shared & BLK_PERM_WRITE) {
+        shared = BLK_PERM_WRITE | BLK_PERM_RESIZE;
+    } else {
+        shared = 0;
+    }
+
+    shared |= BLK_PERM_CONSISTENT_READ | BLK_PERM_GRAPH_MOD |
+              BLK_PERM_WRITE_UNCHANGED;
+
+    if (bs->open_flags & BDRV_O_INACTIVE) {
+        shared |= BLK_PERM_WRITE | BLK_PERM_RESIZE;
+    }
+
+    *nperm = perm;
+    *nshared = shared;
+}
+
+static void bdrv_default_perms_for_storage(BlockDriverState *bs, BdrvChild *c,
+                                           BdrvChildRole role,
+                                           BlockReopenQueue *reopen_queue,
+                                           uint64_t perm, uint64_t shared,
+                                           uint64_t *nperm, uint64_t *nshared)
+{
+    int flags;
+
+    assert(role & (BDRV_CHILD_METADATA | BDRV_CHILD_DATA));
+
+    flags = bdrv_reopen_get_flags(reopen_queue, bs);
+
+    /*
+     * Apart from the modifications below, the same permissions are
+     * forwarded and left alone as for filters
+     */
+    bdrv_filter_default_perms(bs, c, role, reopen_queue,
+                              perm, shared, &perm, &shared);
+
+    if (role & BDRV_CHILD_METADATA) {
+        /* Format drivers may touch metadata even if the guest doesn't write */
+        if (bdrv_is_writable_after_reopen(bs, reopen_queue)) {
+            perm |= BLK_PERM_WRITE | BLK_PERM_RESIZE;
+        }
+
+        /*
+         * bs->file always needs to be consistent because of the
+         * metadata. We can never allow other users to resize or write
+         * to it.
+         */
+        if (!(flags & BDRV_O_NO_IO)) {
+            perm |= BLK_PERM_CONSISTENT_READ;
+        }
+        shared &= ~(BLK_PERM_WRITE | BLK_PERM_RESIZE);
+    }
+
+    if (role & BDRV_CHILD_DATA) {
+        /*
+         * Technically, everything in this block is a subset of the
+         * BDRV_CHILD_METADATA path taken above, and so this could
+         * be an "else if" branch.  However, that is not obvious, and
+         * this function is not performance critical, therefore we let
+         * this be an independent "if".
+         */
+
+        /*
+         * We cannot allow other users to resize the file because the
+         * format driver might have some assumptions about the size
+         * (e.g. because it is stored in metadata, or because the file
+         * is split into fixed-size data files).
+         */
+        shared &= ~BLK_PERM_RESIZE;
+
+        /*
+         * WRITE_UNCHANGED often cannot be performed as such on the
+         * data file.  For example, the qcow2 driver may still need to
+         * write copied clusters on copy-on-read.
+         */
+        if (perm & BLK_PERM_WRITE_UNCHANGED) {
+            perm |= BLK_PERM_WRITE;
+        }
+
+        /*
+         * If the data file is written to, the format driver may
+         * expect to be able to resize it by writing beyond the EOF.
+         */
+        if (perm & BLK_PERM_WRITE) {
+            perm |= BLK_PERM_RESIZE;
+        }
+    }
+
+    if (bs->open_flags & BDRV_O_INACTIVE) {
+        shared |= BLK_PERM_WRITE | BLK_PERM_RESIZE;
+    }
+
+    *nperm = perm;
+    *nshared = shared;
+}
+
+void bdrv_default_perms(BlockDriverState *bs, BdrvChild *c,
+                        BdrvChildRole role, BlockReopenQueue *reopen_queue,
+                        uint64_t perm, uint64_t shared,
+                        uint64_t *nperm, uint64_t *nshared)
+{
+    if (role & BDRV_CHILD_FILTERED) {
+        assert(!(role & (BDRV_CHILD_DATA | BDRV_CHILD_METADATA |
+                         BDRV_CHILD_COW)));
+        bdrv_filter_default_perms(bs, c, role, reopen_queue,
+                                  perm, shared, nperm, nshared);
+    } else if (role & BDRV_CHILD_COW) {
+        assert(!(role & (BDRV_CHILD_DATA | BDRV_CHILD_METADATA)));
+        bdrv_default_perms_for_cow(bs, c, role, reopen_queue,
+                                   perm, shared, nperm, nshared);
+    } else if (role & (BDRV_CHILD_METADATA | BDRV_CHILD_DATA)) {
+        bdrv_default_perms_for_storage(bs, c, role, reopen_queue,
+                                       perm, shared, nperm, nshared);
+    } else {
+        g_assert_not_reached();
+    }
+}
+
+uint64_t bdrv_qapi_perm_to_blk_perm(BlockPermission qapi_perm)
+{
+    static const uint64_t permissions[] = {
+        [BLOCK_PERMISSION_CONSISTENT_READ]  = BLK_PERM_CONSISTENT_READ,
+        [BLOCK_PERMISSION_WRITE]            = BLK_PERM_WRITE,
+        [BLOCK_PERMISSION_WRITE_UNCHANGED]  = BLK_PERM_WRITE_UNCHANGED,
+        [BLOCK_PERMISSION_RESIZE]           = BLK_PERM_RESIZE,
+        [BLOCK_PERMISSION_GRAPH_MOD]        = BLK_PERM_GRAPH_MOD,
+    };
+
+    QEMU_BUILD_BUG_ON(ARRAY_SIZE(permissions) != BLOCK_PERMISSION__MAX);
+    QEMU_BUILD_BUG_ON(1UL << ARRAY_SIZE(permissions) != BLK_PERM_ALL + 1);
+
+    assert(qapi_perm < BLOCK_PERMISSION__MAX);
+
+    return permissions[qapi_perm];
+}
+
+static void bdrv_replace_child_noperm(BdrvChild *child,
+                                      BlockDriverState *new_bs)
+{
+    BlockDriverState *old_bs = child->bs;
+    int new_bs_quiesce_counter;
+    int drain_saldo;
+
+    assert(!child->frozen);
+
+    if (old_bs && new_bs) {
+        assert(bdrv_get_aio_context(old_bs) == bdrv_get_aio_context(new_bs));
+    }
+
+    new_bs_quiesce_counter = (new_bs ? new_bs->quiesce_counter : 0);
+    drain_saldo = new_bs_quiesce_counter - child->parent_quiesce_counter;
+
+    /*
+     * If the new child node is drained but the old one was not, flush
+     * all outstanding requests to the old child node.
+     */
+    while (drain_saldo > 0 && child->klass->drained_begin) {
+        bdrv_parent_drained_begin_single(child, true);
+        drain_saldo--;
+    }
+
+    if (old_bs) {
+        /* Detach first so that the recursive drain sections coming from @child
+         * are already gone and we only end the drain sections that came from
+         * elsewhere. */
+        if (child->klass->detach) {
+            child->klass->detach(child);
+        }
+        QLIST_REMOVE(child, next_parent);
+    }
+
+    child->bs = new_bs;
+
+    if (new_bs) {
+        QLIST_INSERT_HEAD(&new_bs->parents, child, next_parent);
+
+        /*
+         * Detaching the old node may have led to the new node's
+         * quiesce_counter having been decreased.  Not a problem, we
+         * just need to recognize this here and then invoke
+         * drained_end appropriately more often.
+         */
+        assert(new_bs->quiesce_counter <= new_bs_quiesce_counter);
+        drain_saldo += new_bs->quiesce_counter - new_bs_quiesce_counter;
+
+        /* Attach only after starting new drained sections, so that recursive
+         * drain sections coming from @child don't get an extra .drained_begin
+         * callback. */
+        if (child->klass->attach) {
+            child->klass->attach(child);
+        }
+    }
+
+    /*
+     * If the old child node was drained but the new one is not, allow
+     * requests to come in only after the new node has been attached.
+     */
+    while (drain_saldo < 0 && child->klass->drained_end) {
+        bdrv_parent_drained_end_single(child);
+        drain_saldo++;
+    }
+}
+
+/*
+ * Updates @child to change its reference to point to @new_bs, including
+ * checking and applying the necessary permission updates both to the old node
+ * and to @new_bs.
+ *
+ * NULL is passed as @new_bs for removing the reference before freeing @child.
+ *
+ * If @new_bs is not NULL, bdrv_check_perm() must be called beforehand, as this
+ * function uses bdrv_set_perm() to update the permissions according to the new
+ * reference that @new_bs gets.
+ *
+ * Callers must ensure that child->frozen is false.
+ */
+static void bdrv_replace_child(BdrvChild *child, BlockDriverState *new_bs)
+{
+    BlockDriverState *old_bs = child->bs;
+    uint64_t perm, shared_perm;
+
+    /* Asserts that child->frozen == false */
+    bdrv_replace_child_noperm(child, new_bs);
+
+    /*
+     * Start with the new node's permissions.  If @new_bs is a (direct
+     * or indirect) child of @old_bs, we must complete the permission
+     * update on @new_bs before we loosen the restrictions on @old_bs.
+     * Otherwise, bdrv_check_perm() on @old_bs would re-initiate
+     * updating the permissions of @new_bs, and thus not purely loosen
+     * restrictions.
+     */
+    if (new_bs) {
+        bdrv_get_cumulative_perm(new_bs, &perm, &shared_perm);
+        bdrv_set_perm(new_bs, perm, shared_perm);
+    }
+
+    if (old_bs) {
+        /* Update permissions for old node. This is guaranteed to succeed
+         * because we're just taking a parent away, so we're loosening
+         * restrictions. */
+        bool tighten_restrictions;
+        int ret;
+
+        bdrv_get_cumulative_perm(old_bs, &perm, &shared_perm);
+        ret = bdrv_check_perm(old_bs, NULL, perm, shared_perm, NULL,
+                              &tighten_restrictions, NULL);
+        assert(tighten_restrictions == false);
+        if (ret < 0) {
+            /* We only tried to loosen restrictions, so errors are not fatal */
+            bdrv_abort_perm_update(old_bs);
+        } else {
+            bdrv_set_perm(old_bs, perm, shared_perm);
+        }
+
+        /* When the parent requiring a non-default AioContext is removed, the
+         * node moves back to the main AioContext */
+        bdrv_try_set_aio_context(old_bs, qemu_get_aio_context(), NULL);
+    }
+}
+
+/*
+ * This function steals the reference to child_bs from the caller.
+ * That reference is later dropped by bdrv_root_unref_child().
+ *
+ * On failure NULL is returned, errp is set and the reference to
+ * child_bs is also dropped.
+ *
+ * The caller must hold the AioContext lock @child_bs, but not that of @ctx
+ * (unless @child_bs is already in @ctx).
+ */
+BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
+                                  const char *child_name,
+                                  const BdrvChildClass *child_class,
+                                  BdrvChildRole child_role,
+                                  AioContext *ctx,
+                                  uint64_t perm, uint64_t shared_perm,
+                                  void *opaque, Error **errp)
+{
+    BdrvChild *child;
+    Error *local_err = NULL;
+    int ret;
+
+    ret = bdrv_check_update_perm(child_bs, NULL, perm, shared_perm, NULL, NULL,
+                                 errp);
+    if (ret < 0) {
+        bdrv_abort_perm_update(child_bs);
+        bdrv_unref(child_bs);
+        return NULL;
+    }
+
+    child = g_new(BdrvChild, 1);
+    *child = (BdrvChild) {
+        .bs             = NULL,
+        .name           = g_strdup(child_name),
+        .klass          = child_class,
+        .role           = child_role,
+        .perm           = perm,
+        .shared_perm    = shared_perm,
+        .opaque         = opaque,
+    };
+
+    /* If the AioContexts don't match, first try to move the subtree of
+     * child_bs into the AioContext of the new parent. If this doesn't work,
+     * try moving the parent into the AioContext of child_bs instead. */
+    if (bdrv_get_aio_context(child_bs) != ctx) {
+        ret = bdrv_try_set_aio_context(child_bs, ctx, &local_err);
+        if (ret < 0 && child_class->can_set_aio_ctx) {
+            GSList *ignore = g_slist_prepend(NULL, child);
+            ctx = bdrv_get_aio_context(child_bs);
+            if (child_class->can_set_aio_ctx(child, ctx, &ignore, NULL)) {
+                error_free(local_err);
+                ret = 0;
+                g_slist_free(ignore);
+                ignore = g_slist_prepend(NULL, child);
+                child_class->set_aio_ctx(child, ctx, &ignore);
+            }
+            g_slist_free(ignore);
+        }
+        if (ret < 0) {
+            error_propagate(errp, local_err);
+            g_free(child);
+            bdrv_abort_perm_update(child_bs);
+            bdrv_unref(child_bs);
+            return NULL;
+        }
+    }
+
+    /* This performs the matching bdrv_set_perm() for the above check. */
+    bdrv_replace_child(child, child_bs);
+
+    return child;
+}
+
+/*
+ * This function transfers the reference to child_bs from the caller
+ * to parent_bs. That reference is later dropped by parent_bs on
+ * bdrv_close() or if someone calls bdrv_unref_child().
+ *
+ * On failure NULL is returned, errp is set and the reference to
+ * child_bs is also dropped.
+ *
+ * If @parent_bs and @child_bs are in different AioContexts, the caller must
+ * hold the AioContext lock for @child_bs, but not for @parent_bs.
+ */
+BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
+                             BlockDriverState *child_bs,
+                             const char *child_name,
+                             const BdrvChildClass *child_class,
+                             BdrvChildRole child_role,
+                             Error **errp)
+{
+    BdrvChild *child;
+    uint64_t perm, shared_perm;
+
+    bdrv_get_cumulative_perm(parent_bs, &perm, &shared_perm);
+
+    assert(parent_bs->drv);
+    bdrv_child_perm(parent_bs, child_bs, NULL, child_role, NULL,
+                    perm, shared_perm, &perm, &shared_perm);
+
+    child = bdrv_root_attach_child(child_bs, child_name, child_class,
+                                   child_role, bdrv_get_aio_context(parent_bs),
+                                   perm, shared_perm, parent_bs, errp);
+    if (child == NULL) {
+        return NULL;
+    }
+
+    QLIST_INSERT_HEAD(&parent_bs->children, child, next);
+    return child;
+}
+
+static void bdrv_detach_child(BdrvChild *child)
+{
+    QLIST_SAFE_REMOVE(child, next);
+
+    bdrv_replace_child(child, NULL);
+
+    g_free(child->name);
+    g_free(child);
+}
+
+/* Callers must ensure that child->frozen is false. */
+void bdrv_root_unref_child(BdrvChild *child)
+{
+    BlockDriverState *child_bs;
+
+    child_bs = child->bs;
+    bdrv_detach_child(child);
+    bdrv_unref(child_bs);
+}
+
+/**
+ * Clear all inherits_from pointers from children and grandchildren of
+ * @root that point to @root, where necessary.
+ */
+static void bdrv_unset_inherits_from(BlockDriverState *root, BdrvChild *child)
+{
+    BdrvChild *c;
+
+    if (child->bs->inherits_from == root) {
+        /*
+         * Remove inherits_from only when the last reference between root and
+         * child->bs goes away.
+         */
+        QLIST_FOREACH(c, &root->children, next) {
+            if (c != child && c->bs == child->bs) {
+                break;
+            }
+        }
+        if (c == NULL) {
+            child->bs->inherits_from = NULL;
+        }
+    }
+
+    QLIST_FOREACH(c, &child->bs->children, next) {
+        bdrv_unset_inherits_from(root, c);
+    }
+}
+
+/* Callers must ensure that child->frozen is false. */
+void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child)
+{
+    if (child == NULL) {
+        return;
+    }
+
+    bdrv_unset_inherits_from(parent, child);
+    bdrv_root_unref_child(child);
+}
+
+
+static void bdrv_parent_cb_change_media(BlockDriverState *bs, bool load)
+{
+    BdrvChild *c;
+    QLIST_FOREACH(c, &bs->parents, next_parent) {
+        if (c->klass->change_media) {
+            c->klass->change_media(c, load);
+        }
+    }
+}
+
+/* Return true if you can reach parent going through child->inherits_from
+ * recursively. If parent or child are NULL, return false */
+static bool bdrv_inherits_from_recursive(BlockDriverState *child,
+                                         BlockDriverState *parent)
+{
+    while (child && child != parent) {
+        child = child->inherits_from;
+    }
+
+    return child != NULL;
+}
+
+/*
+ * Return the BdrvChildRole for @bs's backing child.  bs->backing is
+ * mostly used for COW backing children (role = COW), but also for
+ * filtered children (role = FILTERED | PRIMARY).
+ */
+static BdrvChildRole bdrv_backing_role(BlockDriverState *bs)
+{
+    if (bs->drv && bs->drv->is_filter) {
+        return BDRV_CHILD_FILTERED | BDRV_CHILD_PRIMARY;
+    } else {
+        return BDRV_CHILD_COW;
+    }
+}
+
+/*
+ * Sets the bs->backing link of a BDS. A new reference is created; callers
+ * which don't need their own reference any more must call bdrv_unref().
+ */
+void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd,
+                         Error **errp)
+{
+    bool update_inherits_from = bdrv_chain_contains(bs, backing_hd) &&
+        bdrv_inherits_from_recursive(backing_hd, bs);
+
+    if (bdrv_is_backing_chain_frozen(bs, child_bs(bs->backing), errp)) {
+        return;
+    }
+
+    if (backing_hd) {
+        bdrv_ref(backing_hd);
+    }
+
+    if (bs->backing) {
+        /* Cannot be frozen, we checked that above */
+        bdrv_unref_child(bs, bs->backing);
+        bs->backing = NULL;
+    }
+
+    if (!backing_hd) {
+        goto out;
+    }
+
+    bs->backing = bdrv_attach_child(bs, backing_hd, "backing", &child_of_bds,
+                                    bdrv_backing_role(bs), errp);
+    /* If backing_hd was already part of bs's backing chain, and
+     * inherits_from pointed recursively to bs then let's update it to
+     * point directly to bs (else it will become NULL). */
+    if (bs->backing && update_inherits_from) {
+        backing_hd->inherits_from = bs;
+    }
+
+out:
+    bdrv_refresh_limits(bs, NULL);
 }
 
 /*
  * Opens the backing file for a BlockDriverState if not yet open
  *
- * options is a QDict of options to pass to the block drivers, or NULL for an
- * empty set of options. The reference to the QDict is transferred to this
- * function (even on failure), so if the caller intends to reuse the dictionary,
- * it needs to use QINCREF() before calling bdrv_file_open.
+ * bdref_key specifies the key for the image's BlockdevRef in the options QDict.
+ * That QDict has to be flattened; therefore, if the BlockdevRef is a QDict
+ * itself, all options starting with "${bdref_key}." are considered part of the
+ * BlockdevRef.
+ *
+ * TODO Can this be unified with bdrv_open_image()?
  */
-int bdrv_open_backing_file(BlockDriverState *bs, QDict *options)
+int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
+                           const char *bdref_key, Error **errp)
 {
-    char backing_filename[PATH_MAX];
-    int back_flags, ret;
-    BlockDriver *back_drv = NULL;
+    char *backing_filename = NULL;
+    char *bdref_key_dot;
+    const char *reference = NULL;
+    int ret = 0;
+    bool implicit_backing = false;
+    BlockDriverState *backing_hd;
+    QDict *options;
+    QDict *tmp_parent_options = NULL;
+    Error *local_err = NULL;
 
-    if (bs->backing_hd != NULL) {
-        QDECREF(options);
-        return 0;
+    if (bs->backing != NULL) {
+        goto free_exit;
     }
 
     /* NULL means an empty set of options */
-    if (options == NULL) {
-        options = qdict_new();
+    if (parent_options == NULL) {
+        tmp_parent_options = qdict_new();
+        parent_options = tmp_parent_options;
     }
 
     bs->open_flags &= ~BDRV_O_NO_BACKING;
-    if (qdict_haskey(options, "file.filename")) {
-        backing_filename[0] = '\0';
+
+    bdref_key_dot = g_strdup_printf("%s.", bdref_key);
+    qdict_extract_subqdict(parent_options, &options, bdref_key_dot);
+    g_free(bdref_key_dot);
+
+    /*
+     * Caution: while qdict_get_try_str() is fine, getting non-string
+     * types would require more care.  When @parent_options come from
+     * -blockdev or blockdev_add, its members are typed according to
+     * the QAPI schema, but when they come from -drive, they're all
+     * QString.
+     */
+    reference = qdict_get_try_str(parent_options, bdref_key);
+    if (reference || qdict_haskey(options, "file.filename")) {
+        /* keep backing_filename NULL */
     } else if (bs->backing_file[0] == '\0' && qdict_size(options) == 0) {
-        QDECREF(options);
-        return 0;
-    }
+        qobject_unref(options);
+        goto free_exit;
+    } else {
+        if (qdict_size(options) == 0) {
+            /* If the user specifies options that do not modify the
+             * backing file's behavior, we might still consider it the
+             * implicit backing file.  But it's easier this way, and
+             * just specifying some of the backing BDS's options is
+             * only possible with -drive anyway (otherwise the QAPI
+             * schema forces the user to specify everything). */
+            implicit_backing = !strcmp(bs->auto_backing_file, bs->backing_file);
+        }
 
-    bs->backing_hd = bdrv_new("");
-    bdrv_get_full_backing_filename(bs, backing_filename,
-                                   sizeof(backing_filename));
+        backing_filename = bdrv_get_full_backing_filename(bs, &local_err);
+        if (local_err) {
+            ret = -EINVAL;
+            error_propagate(errp, local_err);
+            qobject_unref(options);
+            goto free_exit;
+        }
+    }
 
-    if (bs->backing_format[0] != '\0') {
-        back_drv = bdrv_find_format(bs->backing_format);
+    if (!bs->drv || !bs->drv->supports_backing) {
+        ret = -EINVAL;
+        error_setg(errp, "Driver doesn't support backing files");
+        qobject_unref(options);
+        goto free_exit;
     }
 
-    /* backing files always opened read-only */
-    back_flags = bs->open_flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT);
+    if (!reference &&
+        bs->backing_format[0] != '\0' && !qdict_haskey(options, "driver")) {
+        qdict_put_str(options, "driver", bs->backing_format);
+    }
 
-    ret = bdrv_open(bs->backing_hd,
-                    *backing_filename ? backing_filename : NULL, options,
-                    back_flags, back_drv);
-    if (ret < 0) {
-        bdrv_delete(bs->backing_hd);
-        bs->backing_hd = NULL;
+    backing_hd = bdrv_open_inherit(backing_filename, reference, options, 0, bs,
+                                   &child_of_bds, bdrv_backing_role(bs), errp);
+    if (!backing_hd) {
         bs->open_flags |= BDRV_O_NO_BACKING;
-        return ret;
+        error_prepend(errp, "Could not open backing file: ");
+        ret = -EINVAL;
+        goto free_exit;
     }
-    return 0;
+
+    if (implicit_backing) {
+        bdrv_refresh_filename(backing_hd);
+        pstrcpy(bs->auto_backing_file, sizeof(bs->auto_backing_file),
+                backing_hd->filename);
+    }
+
+    /* Hook up the backing file link; drop our reference, bs owns the
+     * backing_hd reference now */
+    bdrv_set_backing_hd(bs, backing_hd, &local_err);
+    bdrv_unref(backing_hd);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        ret = -EINVAL;
+        goto free_exit;
+    }
+
+    qdict_del(parent_options, bdref_key);
+
+free_exit:
+    g_free(backing_filename);
+    qobject_unref(tmp_parent_options);
+    return ret;
 }
 
-static void extract_subqdict(QDict *src, QDict **dst, const char *start)
+static BlockDriverState *
+bdrv_open_child_bs(const char *filename, QDict *options, const char *bdref_key,
+                   BlockDriverState *parent, const BdrvChildClass *child_class,
+                   BdrvChildRole child_role, bool allow_none, Error **errp)
 {
-    const QDictEntry *entry, *next;
-    const char *p;
+    BlockDriverState *bs = NULL;
+    QDict *image_options;
+    char *bdref_key_dot;
+    const char *reference;
+
+    assert(child_class != NULL);
 
-    *dst = qdict_new();
-    entry = qdict_first(src);
+    bdref_key_dot = g_strdup_printf("%s.", bdref_key);
+    qdict_extract_subqdict(options, &image_options, bdref_key_dot);
+    g_free(bdref_key_dot);
 
-    while (entry != NULL) {
-        next = qdict_next(src, entry);
-        if (strstart(entry->key, start, &p)) {
-            qobject_incref(entry->value);
-            qdict_put_obj(*dst, p, entry->value);
-            qdict_del(src, entry->key);
+    /*
+     * Caution: while qdict_get_try_str() is fine, getting non-string
+     * types would require more care.  When @options come from
+     * -blockdev or blockdev_add, its members are typed according to
+     * the QAPI schema, but when they come from -drive, they're all
+     * QString.
+     */
+    reference = qdict_get_try_str(options, bdref_key);
+    if (!filename && !reference && !qdict_size(image_options)) {
+        if (!allow_none) {
+            error_setg(errp, "A block device must be specified for \"%s\"",
+                       bdref_key);
         }
-        entry = next;
+        qobject_unref(image_options);
+        goto done;
+    }
+
+    bs = bdrv_open_inherit(filename, reference, image_options, 0,
+                           parent, child_class, child_role, errp);
+    if (!bs) {
+        goto done;
+    }
+
+done:
+    qdict_del(options, bdref_key);
+    return bs;
+}
+
+/*
+ * Opens a disk image whose options are given as BlockdevRef in another block
+ * device's options.
+ *
+ * If allow_none is true, no image will be opened if filename is false and no
+ * BlockdevRef is given. NULL will be returned, but errp remains unset.
+ *
+ * bdrev_key specifies the key for the image's BlockdevRef in the options QDict.
+ * That QDict has to be flattened; therefore, if the BlockdevRef is a QDict
+ * itself, all options starting with "${bdref_key}." are considered part of the
+ * BlockdevRef.
+ *
+ * The BlockdevRef will be removed from the options QDict.
+ */
+BdrvChild *bdrv_open_child(const char *filename,
+                           QDict *options, const char *bdref_key,
+                           BlockDriverState *parent,
+                           const BdrvChildClass *child_class,
+                           BdrvChildRole child_role,
+                           bool allow_none, Error **errp)
+{
+    BlockDriverState *bs;
+
+    bs = bdrv_open_child_bs(filename, options, bdref_key, parent, child_class,
+                            child_role, allow_none, errp);
+    if (bs == NULL) {
+        return NULL;
+    }
+
+    return bdrv_attach_child(parent, bs, bdref_key, child_class, child_role,
+                             errp);
+}
+
+/*
+ * TODO Future callers may need to specify parent/child_class in order for
+ * option inheritance to work. Existing callers use it for the root node.
+ */
+BlockDriverState *bdrv_open_blockdev_ref(BlockdevRef *ref, Error **errp)
+{
+    BlockDriverState *bs = NULL;
+    QObject *obj = NULL;
+    QDict *qdict = NULL;
+    const char *reference = NULL;
+    Visitor *v = NULL;
+
+    if (ref->type == QTYPE_QSTRING) {
+        reference = ref->u.reference;
+    } else {
+        BlockdevOptions *options = &ref->u.definition;
+        assert(ref->type == QTYPE_QDICT);
+
+        v = qobject_output_visitor_new(&obj);
+        visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
+        visit_complete(v, &obj);
+
+        qdict = qobject_to(QDict, obj);
+        qdict_flatten(qdict);
+
+        /* bdrv_open_inherit() 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(qdict, BDRV_OPT_CACHE_DIRECT, "off");
+        qdict_set_default_str(qdict, BDRV_OPT_CACHE_NO_FLUSH, "off");
+        qdict_set_default_str(qdict, BDRV_OPT_READ_ONLY, "off");
+        qdict_set_default_str(qdict, BDRV_OPT_AUTO_READ_ONLY, "off");
+
+    }
+
+    bs = bdrv_open_inherit(NULL, reference, qdict, 0, NULL, NULL, 0, errp);
+    obj = NULL;
+    qobject_unref(obj);
+    visit_free(v);
+    return bs;
+}
+
+static BlockDriverState *bdrv_append_temp_snapshot(BlockDriverState *bs,
+                                                   int flags,
+                                                   QDict *snapshot_options,
+                                                   Error **errp)
+{
+    /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */
+    char *tmp_filename = g_malloc0(PATH_MAX + 1);
+    int64_t total_size;
+    QemuOpts *opts = NULL;
+    BlockDriverState *bs_snapshot = NULL;
+    Error *local_err = NULL;
+    int ret;
+
+    /* if snapshot, we create a temporary backing file and open it
+       instead of opening 'filename' directly */
+
+    /* Get the required size from the image */
+    total_size = bdrv_getlength(bs);
+    if (total_size < 0) {
+        error_setg_errno(errp, -total_size, "Could not get image size");
+        goto out;
+    }
+
+    /* Create the temporary image */
+    ret = get_tmp_filename(tmp_filename, PATH_MAX + 1);
+    if (ret < 0) {
+        error_setg_errno(errp, -ret, "Could not get temporary filename");
+        goto out;
     }
+
+    opts = qemu_opts_create(bdrv_qcow2.create_opts, NULL, 0,
+                            &error_abort);
+    qemu_opt_set_number(opts, BLOCK_OPT_SIZE, total_size, &error_abort);
+    ret = bdrv_create(&bdrv_qcow2, tmp_filename, opts, errp);
+    qemu_opts_del(opts);
+    if (ret < 0) {
+        error_prepend(errp, "Could not create temporary overlay '%s': ",
+                      tmp_filename);
+        goto out;
+    }
+
+    /* Prepare options QDict for the temporary file */
+    qdict_put_str(snapshot_options, "file.driver", "file");
+    qdict_put_str(snapshot_options, "file.filename", tmp_filename);
+    qdict_put_str(snapshot_options, "driver", "qcow2");
+
+    bs_snapshot = bdrv_open(NULL, NULL, snapshot_options, flags, errp);
+    snapshot_options = NULL;
+    if (!bs_snapshot) {
+        goto out;
+    }
+
+    /* bdrv_append() consumes a strong reference to bs_snapshot
+     * (i.e. it will call bdrv_unref() on it) even on error, so in
+     * order to be able to return one, we have to increase
+     * bs_snapshot's refcount here */
+    bdrv_ref(bs_snapshot);
+    bdrv_append(bs_snapshot, bs, &local_err);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        bs_snapshot = NULL;
+        goto out;
+    }
+
+out:
+    qobject_unref(snapshot_options);
+    g_free(tmp_filename);
+    return bs_snapshot;
 }
 
 /*
@@ -960,195 +3219,381 @@ static void extract_subqdict(QDict *src, QDict **dst, const char *start)
  * options is a QDict of options to pass to the block drivers, or NULL for an
  * empty set of options. The reference to the QDict belongs to the block layer
  * after the call (even on failure), so if the caller intends to reuse the
- * dictionary, it needs to use QINCREF() before calling bdrv_open.
+ * dictionary, it needs to use qobject_ref() before calling bdrv_open.
+ *
+ * If *pbs is NULL, a new BDS will be created with a pointer to it stored there.
+ * If it is not NULL, the referenced BDS will be reused.
+ *
+ * The reference parameter may be used to specify an existing block device which
+ * should be opened. If specified, neither options nor a filename may be given,
+ * nor can an existing BDS be reused (that is, *pbs has to be NULL).
  */
-int bdrv_open(BlockDriverState *bs, const char *filename, QDict *options,
-              int flags, BlockDriver *drv)
+static BlockDriverState *bdrv_open_inherit(const char *filename,
+                                           const char *reference,
+                                           QDict *options, int flags,
+                                           BlockDriverState *parent,
+                                           const BdrvChildClass *child_class,
+                                           BdrvChildRole child_role,
+                                           Error **errp)
 {
     int ret;
-    /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */
-    char tmp_filename[PATH_MAX + 1];
-    BlockDriverState *file = NULL;
-    QDict *file_options = NULL;
+    BlockBackend *file = NULL;
+    BlockDriverState *bs;
+    BlockDriver *drv = NULL;
+    BdrvChild *child;
     const char *drvname;
+    const char *backing;
+    Error *local_err = NULL;
+    QDict *snapshot_options = NULL;
+    int snapshot_flags = 0;
 
-    /* NULL means an empty set of options */
-    if (options == NULL) {
-        options = qdict_new();
-    }
+    assert(!child_class || !flags);
+    assert(!child_class == !parent);
 
-    bs->options = options;
-    options = qdict_clone_shallow(options);
+    if (reference) {
+        bool options_non_empty = options ? qdict_size(options) : false;
+        qobject_unref(options);
 
-    /* For snapshot=on, create a temporary qcow2 overlay */
-    if (flags & BDRV_O_SNAPSHOT) {
-        BlockDriverState *bs1;
-        int64_t total_size;
-        BlockDriver *bdrv_qcow2;
-        QEMUOptionParameter *create_options;
-        char backing_filename[PATH_MAX];
-
-        if (qdict_size(options) != 0) {
-            error_report("Can't use snapshot=on with driver-specific options");
-            ret = -EINVAL;
-            goto fail;
+        if (filename || options_non_empty) {
+            error_setg(errp, "Cannot reference an existing block device with "
+                       "additional options or a new filename");
+            return NULL;
         }
-        assert(filename != NULL);
-
-        /* if snapshot, we create a temporary backing file and open it
-           instead of opening 'filename' directly */
 
-        /* if there is a backing file, use it */
-        bs1 = bdrv_new("");
-        ret = bdrv_open(bs1, filename, NULL, 0, drv);
-        if (ret < 0) {
-            bdrv_delete(bs1);
-            goto fail;
+        bs = bdrv_lookup_bs(reference, reference, errp);
+        if (!bs) {
+            return NULL;
         }
-        total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK;
 
-        bdrv_delete(bs1);
+        bdrv_ref(bs);
+        return bs;
+    }
 
-        ret = get_tmp_filename(tmp_filename, sizeof(tmp_filename));
-        if (ret < 0) {
-            goto fail;
-        }
+    bs = bdrv_new();
 
-        /* Real path is meaningless for protocols */
-        if (path_has_protocol(filename)) {
-            snprintf(backing_filename, sizeof(backing_filename),
-                     "%s", filename);
-        } else if (!realpath(filename, backing_filename)) {
-            ret = -errno;
-            goto fail;
-        }
+    /* NULL means an empty set of options */
+    if (options == NULL) {
+        options = qdict_new();
+    }
+
+    /* json: syntax counts as explicit options, as if in the QDict */
+    parse_json_protocol(options, &filename, &local_err);
+    if (local_err) {
+        goto fail;
+    }
 
-        bdrv_qcow2 = bdrv_find_format("qcow2");
-        create_options = parse_option_parameters("", bdrv_qcow2->create_options,
-                                                 NULL);
+    bs->explicit_options = qdict_clone_shallow(options);
 
-        set_option_parameter_int(create_options, BLOCK_OPT_SIZE, total_size);
-        set_option_parameter(create_options, BLOCK_OPT_BACKING_FILE,
-                             backing_filename);
-        if (drv) {
-            set_option_parameter(create_options, BLOCK_OPT_BACKING_FMT,
-                drv->format_name);
-        }
+    if (child_class) {
+        bool parent_is_format;
 
-        ret = bdrv_create(bdrv_qcow2, tmp_filename, create_options);
-        free_option_parameters(create_options);
-        if (ret < 0) {
-            goto fail;
+        if (parent->drv) {
+            parent_is_format = parent->drv->is_format;
+        } else {
+            /*
+             * parent->drv is not set yet because this node is opened for
+             * (potential) format probing.  That means that @parent is going
+             * to be a format node.
+             */
+            parent_is_format = true;
         }
 
-        filename = tmp_filename;
-        drv = bdrv_qcow2;
-        bs->is_temporary = 1;
+        bs->inherits_from = parent;
+        child_class->inherit_options(child_role, parent_is_format,
+                                     &flags, options,
+                                     parent->open_flags, parent->options);
     }
 
-    /* Open image file without format layer */
-    if (flags & BDRV_O_RDWR) {
-        flags |= BDRV_O_ALLOW_RDWR;
+    ret = bdrv_fill_options(&options, filename, &flags, &local_err);
+    if (ret < 0) {
+        goto fail;
     }
 
-    extract_subqdict(options, &file_options, "file.");
+    /*
+     * Set the BDRV_O_RDWR and BDRV_O_ALLOW_RDWR flags.
+     * Caution: getting a boolean member of @options requires care.
+     * When @options come from -blockdev or blockdev_add, members are
+     * typed according to the QAPI schema, but when they come from
+     * -drive, they're all QString.
+     */
+    if (g_strcmp0(qdict_get_try_str(options, BDRV_OPT_READ_ONLY), "on") &&
+        !qdict_get_try_bool(options, BDRV_OPT_READ_ONLY, false)) {
+        flags |= (BDRV_O_RDWR | BDRV_O_ALLOW_RDWR);
+    } else {
+        flags &= ~BDRV_O_RDWR;
+    }
 
-    ret = bdrv_file_open(&file, filename, file_options,
-                         bdrv_open_flags(bs, flags | BDRV_O_UNMAP));
-    if (ret < 0) {
-        goto fail;
+    if (flags & BDRV_O_SNAPSHOT) {
+        snapshot_options = qdict_new();
+        bdrv_temp_snapshot_options(&snapshot_flags, snapshot_options,
+                                   flags, options);
+        /* Let bdrv_backing_options() override "read-only" */
+        qdict_del(options, BDRV_OPT_READ_ONLY);
+        bdrv_inherited_options(BDRV_CHILD_COW, true,
+                               &flags, options, flags, options);
     }
 
+    bs->open_flags = flags;
+    bs->options = options;
+    options = qdict_clone_shallow(options);
+
     /* Find the right image format driver */
+    /* See cautionary note on accessing @options above */
     drvname = qdict_get_try_str(options, "driver");
     if (drvname) {
-        drv = bdrv_find_whitelisted_format(drvname, !(flags & BDRV_O_RDWR));
-        qdict_del(options, "driver");
+        drv = bdrv_find_format(drvname);
+        if (!drv) {
+            error_setg(errp, "Unknown driver: '%s'", drvname);
+            goto fail;
+        }
     }
 
-    if (!drv) {
-        ret = find_image_format(file, filename, &drv);
+    assert(drvname || !(flags & BDRV_O_PROTOCOL));
+
+    /* See cautionary note on accessing @options above */
+    backing = qdict_get_try_str(options, "backing");
+    if (qobject_to(QNull, qdict_get(options, "backing")) != NULL ||
+        (backing && *backing == '\0'))
+    {
+        if (backing) {
+            warn_report("Use of \"backing\": \"\" is deprecated; "
+                        "use \"backing\": null instead");
+        }
+        flags |= BDRV_O_NO_BACKING;
+        qdict_del(bs->explicit_options, "backing");
+        qdict_del(bs->options, "backing");
+        qdict_del(options, "backing");
+    }
+
+    /* Open image file without format layer. This BlockBackend is only used for
+     * probing, the block drivers will do their own bdrv_open_child() for the
+     * same BDS, which is why we put the node name back into options. */
+    if ((flags & BDRV_O_PROTOCOL) == 0) {
+        BlockDriverState *file_bs;
+
+        file_bs = bdrv_open_child_bs(filename, options, "file", bs,
+                                     &child_of_bds, BDRV_CHILD_IMAGE,
+                                     true, &local_err);
+        if (local_err) {
+            goto fail;
+        }
+        if (file_bs != NULL) {
+            /* Not requesting BLK_PERM_CONSISTENT_READ because we're only
+             * looking at the header to guess the image format. This works even
+             * in cases where a guest would not see a consistent state. */
+            file = blk_new(bdrv_get_aio_context(file_bs), 0, BLK_PERM_ALL);
+            blk_insert_bs(file, file_bs, &local_err);
+            bdrv_unref(file_bs);
+            if (local_err) {
+                goto fail;
+            }
+
+            qdict_put_str(options, "file", bdrv_get_node_name(file_bs));
+        }
     }
 
-    if (!drv) {
-        goto unlink_and_fail;
+    /* Image format probing */
+    bs->probed = !drv;
+    if (!drv && file) {
+        ret = find_image_format(file, filename, &drv, &local_err);
+        if (ret < 0) {
+            goto fail;
+        }
+        /*
+         * This option update would logically belong in bdrv_fill_options(),
+         * but we first need to open bs->file for the probing to work, while
+         * opening bs->file already requires the (mostly) final set of options
+         * so that cache mode etc. can be inherited.
+         *
+         * Adding the driver later is somewhat ugly, but it's not an option
+         * that would ever be inherited, so it's correct. We just need to make
+         * sure to update both bs->options (which has the full effective
+         * options for bs) and options (which has file.* already removed).
+         */
+        qdict_put_str(bs->options, "driver", drv->format_name);
+        qdict_put_str(options, "driver", drv->format_name);
+    } else if (!drv) {
+        error_setg(errp, "Must specify either driver or file");
+        goto fail;
     }
 
+    /* BDRV_O_PROTOCOL must be set iff a protocol BDS is about to be created */
+    assert(!!(flags & BDRV_O_PROTOCOL) == !!drv->bdrv_file_open);
+    /* file must be NULL if a protocol BDS is about to be created
+     * (the inverse results in an error message from bdrv_open_common()) */
+    assert(!(flags & BDRV_O_PROTOCOL) || !file);
+
     /* Open the image */
-    ret = bdrv_open_common(bs, file, options, flags, drv);
+    ret = bdrv_open_common(bs, file, options, &local_err);
     if (ret < 0) {
-        goto unlink_and_fail;
+        goto fail;
     }
 
-    if (bs->file != file) {
-        bdrv_delete(file);
+    if (file) {
+        blk_unref(file);
         file = NULL;
     }
 
     /* If there is a backing file, use it */
     if ((flags & BDRV_O_NO_BACKING) == 0) {
-        QDict *backing_options;
-
-        extract_subqdict(options, &backing_options, "backing.");
-        ret = bdrv_open_backing_file(bs, backing_options);
+        ret = bdrv_open_backing_file(bs, options, "backing", &local_err);
         if (ret < 0) {
             goto close_and_fail;
         }
     }
 
+    /* Remove all children options and references
+     * from bs->options and bs->explicit_options */
+    QLIST_FOREACH(child, &bs->children, next) {
+        char *child_key_dot;
+        child_key_dot = g_strdup_printf("%s.", child->name);
+        qdict_extract_subqdict(bs->explicit_options, NULL, child_key_dot);
+        qdict_extract_subqdict(bs->options, NULL, child_key_dot);
+        qdict_del(bs->explicit_options, child->name);
+        qdict_del(bs->options, child->name);
+        g_free(child_key_dot);
+    }
+
     /* Check if any unknown options were used */
     if (qdict_size(options) != 0) {
         const QDictEntry *entry = qdict_first(options);
-        qerror_report(ERROR_CLASS_GENERIC_ERROR, "Block format '%s' used by "
-            "device '%s' doesn't support the option '%s'",
-            drv->format_name, bs->device_name, entry->key);
+        if (flags & BDRV_O_PROTOCOL) {
+            error_setg(errp, "Block protocol '%s' doesn't support the option "
+                       "'%s'", drv->format_name, entry->key);
+        } else {
+            error_setg(errp,
+                       "Block format '%s' does not support the option '%s'",
+                       drv->format_name, entry->key);
+        }
 
-        ret = -EINVAL;
         goto close_and_fail;
     }
-    QDECREF(options);
 
-    if (!bdrv_key_required(bs)) {
-        bdrv_dev_change_media_cb(bs, true);
-    }
+    bdrv_parent_cb_change_media(bs, true);
 
-    /* throttling disk I/O limits */
-    if (bs->io_limits_enabled) {
-        bdrv_io_limits_enable(bs);
+    qobject_unref(options);
+    options = NULL;
+
+    /* For snapshot=on, create a temporary qcow2 overlay. bs points to the
+     * temporary snapshot afterwards. */
+    if (snapshot_flags) {
+        BlockDriverState *snapshot_bs;
+        snapshot_bs = bdrv_append_temp_snapshot(bs, snapshot_flags,
+                                                snapshot_options, &local_err);
+        snapshot_options = NULL;
+        if (local_err) {
+            goto close_and_fail;
+        }
+        /* We are not going to return bs but the overlay on top of it
+         * (snapshot_bs); thus, we have to drop the strong reference to bs
+         * (which we obtained by calling bdrv_new()). bs will not be deleted,
+         * though, because the overlay still has a reference to it. */
+        bdrv_unref(bs);
+        bs = snapshot_bs;
     }
 
-    return 0;
+    return bs;
 
-unlink_and_fail:
-    if (file != NULL) {
-        bdrv_delete(file);
-    }
-    if (bs->is_temporary) {
-        unlink(filename);
-    }
 fail:
-    QDECREF(bs->options);
-    QDECREF(options);
+    blk_unref(file);
+    qobject_unref(snapshot_options);
+    qobject_unref(bs->explicit_options);
+    qobject_unref(bs->options);
+    qobject_unref(options);
     bs->options = NULL;
-    return ret;
+    bs->explicit_options = NULL;
+    bdrv_unref(bs);
+    error_propagate(errp, local_err);
+    return NULL;
 
 close_and_fail:
-    bdrv_close(bs);
-    QDECREF(options);
-    return ret;
+    bdrv_unref(bs);
+    qobject_unref(snapshot_options);
+    qobject_unref(options);
+    error_propagate(errp, local_err);
+    return NULL;
 }
 
-typedef struct BlockReopenQueueEntry {
-     bool prepared;
-     BDRVReopenState state;
-     QSIMPLEQ_ENTRY(BlockReopenQueueEntry) entry;
-} BlockReopenQueueEntry;
+BlockDriverState *bdrv_open(const char *filename, const char *reference,
+                            QDict *options, int flags, Error **errp)
+{
+    return bdrv_open_inherit(filename, reference, options, flags, NULL,
+                             NULL, 0, errp);
+}
+
+/* Return true if the NULL-terminated @list contains @str */
+static bool is_str_in_list(const char *str, const char *const *list)
+{
+    if (str && list) {
+        int i;
+        for (i = 0; list[i] != NULL; i++) {
+            if (!strcmp(str, list[i])) {
+                return true;
+            }
+        }
+    }
+    return false;
+}
+
+/*
+ * Check that every option set in @bs->options is also set in
+ * @new_opts.
+ *
+ * Options listed in the common_options list and in
+ * @bs->drv->mutable_opts are skipped.
+ *
+ * Return 0 on success, otherwise return -EINVAL and set @errp.
+ */
+static int bdrv_reset_options_allowed(BlockDriverState *bs,
+                                      const QDict *new_opts, Error **errp)
+{
+    const QDictEntry *e;
+    /* These options are common to all block drivers and are handled
+     * in bdrv_reopen_prepare() so they can be left out of @new_opts */
+    const char *const common_options[] = {
+        "node-name", "discard", "cache.direct", "cache.no-flush",
+        "read-only", "auto-read-only", "detect-zeroes", NULL
+    };
+
+    for (e = qdict_first(bs->options); e; e = qdict_next(bs->options, e)) {
+        if (!qdict_haskey(new_opts, e->key) &&
+            !is_str_in_list(e->key, common_options) &&
+            !is_str_in_list(e->key, bs->drv->mutable_opts)) {
+            error_setg(errp, "Option '%s' cannot be reset "
+                       "to its default value", e->key);
+            return -EINVAL;
+        }
+    }
+
+    return 0;
+}
+
+/*
+ * Returns true if @child can be reached recursively from @bs
+ */
+static bool bdrv_recurse_has_child(BlockDriverState *bs,
+                                   BlockDriverState *child)
+{
+    BdrvChild *c;
+
+    if (bs == child) {
+        return true;
+    }
+
+    QLIST_FOREACH(c, &bs->children, next) {
+        if (bdrv_recurse_has_child(c->bs, child)) {
+            return true;
+        }
+    }
+
+    return false;
+}
 
 /*
  * Adds a BlockDriverState to a simple queue for an atomic, transactional
  * reopen of multiple devices.
  *
- * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT
+ * bs_queue can either be an existing BlockReopenQueue that has had QTAILQ_INIT
  * already performed, or alternatively may be NULL a new BlockReopenQueue will
  * be created and initialized. This newly created BlockReopenQueue should be
  * passed back in for subsequent calls that are intended to be of the same
@@ -1156,36 +3601,184 @@ typedef struct BlockReopenQueueEntry {
  *
  * bs is the BlockDriverState to add to the reopen queue.
  *
+ * options contains the changed options for the associated bs
+ * (the BlockReopenQueue takes ownership)
+ *
  * flags contains the open flags for the associated bs
  *
  * returns a pointer to bs_queue, which is either the newly allocated
  * bs_queue, or the existing bs_queue being used.
  *
+ * bs must be drained between bdrv_reopen_queue() and bdrv_reopen_multiple().
  */
-BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
-                                    BlockDriverState *bs, int flags)
+static BlockReopenQueue *bdrv_reopen_queue_child(BlockReopenQueue *bs_queue,
+                                                 BlockDriverState *bs,
+                                                 QDict *options,
+                                                 const BdrvChildClass *klass,
+                                                 BdrvChildRole role,
+                                                 bool parent_is_format,
+                                                 QDict *parent_options,
+                                                 int parent_flags,
+                                                 bool keep_old_opts)
 {
     assert(bs != NULL);
 
     BlockReopenQueueEntry *bs_entry;
+    BdrvChild *child;
+    QDict *old_options, *explicit_options, *options_copy;
+    int flags;
+    QemuOpts *opts;
+
+    /* Make sure that the caller remembered to use a drained section. This is
+     * important to avoid graph changes between the recursive queuing here and
+     * bdrv_reopen_multiple(). */
+    assert(bs->quiesce_counter > 0);
+
     if (bs_queue == NULL) {
         bs_queue = g_new0(BlockReopenQueue, 1);
-        QSIMPLEQ_INIT(bs_queue);
+        QTAILQ_INIT(bs_queue);
+    }
+
+    if (!options) {
+        options = qdict_new();
     }
 
-    if (bs->file) {
-        bdrv_reopen_queue(bs_queue, bs->file, flags);
+    /* Check if this BlockDriverState is already in the queue */
+    QTAILQ_FOREACH(bs_entry, bs_queue, entry) {
+        if (bs == bs_entry->state.bs) {
+            break;
+        }
+    }
+
+    /*
+     * Precedence of options:
+     * 1. Explicitly passed in options (highest)
+     * 2. Retained from explicitly set options of bs
+     * 3. Inherited from parent node
+     * 4. Retained from effective options of bs
+     */
+
+    /* Old explicitly set values (don't overwrite by inherited value) */
+    if (bs_entry || keep_old_opts) {
+        old_options = qdict_clone_shallow(bs_entry ?
+                                          bs_entry->state.explicit_options :
+                                          bs->explicit_options);
+        bdrv_join_options(bs, options, old_options);
+        qobject_unref(old_options);
+    }
+
+    explicit_options = qdict_clone_shallow(options);
+
+    /* Inherit from parent node */
+    if (parent_options) {
+        flags = 0;
+        klass->inherit_options(role, parent_is_format, &flags, options,
+                               parent_flags, parent_options);
+    } else {
+        flags = bdrv_get_flags(bs);
     }
 
-    bs_entry = g_new0(BlockReopenQueueEntry, 1);
-    QSIMPLEQ_INSERT_TAIL(bs_queue, bs_entry, entry);
+    if (keep_old_opts) {
+        /* Old values are used for options that aren't set yet */
+        old_options = qdict_clone_shallow(bs->options);
+        bdrv_join_options(bs, options, old_options);
+        qobject_unref(old_options);
+    }
+
+    /* We have the final set of options so let's update the flags */
+    options_copy = qdict_clone_shallow(options);
+    opts = qemu_opts_create(&bdrv_runtime_opts, NULL, 0, &error_abort);
+    qemu_opts_absorb_qdict(opts, options_copy, NULL);
+    update_flags_from_options(&flags, opts);
+    qemu_opts_del(opts);
+    qobject_unref(options_copy);
+
+    /* bdrv_open_inherit() sets and clears some additional flags internally */
+    flags &= ~BDRV_O_PROTOCOL;
+    if (flags & BDRV_O_RDWR) {
+        flags |= BDRV_O_ALLOW_RDWR;
+    }
+
+    if (!bs_entry) {
+        bs_entry = g_new0(BlockReopenQueueEntry, 1);
+        QTAILQ_INSERT_TAIL(bs_queue, bs_entry, entry);
+    } else {
+        qobject_unref(bs_entry->state.options);
+        qobject_unref(bs_entry->state.explicit_options);
+    }
 
     bs_entry->state.bs = bs;
+    bs_entry->state.options = options;
+    bs_entry->state.explicit_options = explicit_options;
     bs_entry->state.flags = flags;
 
+    /* This needs to be overwritten in bdrv_reopen_prepare() */
+    bs_entry->state.perm = UINT64_MAX;
+    bs_entry->state.shared_perm = 0;
+
+    /*
+     * If keep_old_opts is false then it means that unspecified
+     * options must be reset to their original value. We don't allow
+     * resetting 'backing' but we need to know if the option is
+     * missing in order to decide if we have to return an error.
+     */
+    if (!keep_old_opts) {
+        bs_entry->state.backing_missing =
+            !qdict_haskey(options, "backing") &&
+            !qdict_haskey(options, "backing.driver");
+    }
+
+    QLIST_FOREACH(child, &bs->children, next) {
+        QDict *new_child_options = NULL;
+        bool child_keep_old = keep_old_opts;
+
+        /* reopen can only change the options of block devices that were
+         * implicitly created and inherited options. For other (referenced)
+         * block devices, a syntax like "backing.foo" results in an error. */
+        if (child->bs->inherits_from != bs) {
+            continue;
+        }
+
+        /* Check if the options contain a child reference */
+        if (qdict_haskey(options, child->name)) {
+            const char *childref = qdict_get_try_str(options, child->name);
+            /*
+             * The current child must not be reopened if the child
+             * reference is null or points to a different node.
+             */
+            if (g_strcmp0(childref, child->bs->node_name)) {
+                continue;
+            }
+            /*
+             * If the child reference points to the current child then
+             * reopen it with its existing set of options (note that
+             * it can still inherit new options from the parent).
+             */
+            child_keep_old = true;
+        } else {
+            /* Extract child options ("child-name.*") */
+            char *child_key_dot = g_strdup_printf("%s.", child->name);
+            qdict_extract_subqdict(explicit_options, NULL, child_key_dot);
+            qdict_extract_subqdict(options, &new_child_options, child_key_dot);
+            g_free(child_key_dot);
+        }
+
+        bdrv_reopen_queue_child(bs_queue, child->bs, new_child_options,
+                                child->klass, child->role, bs->drv->is_format,
+                                options, flags, child_keep_old);
+    }
+
     return bs_queue;
 }
 
+BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
+                                    BlockDriverState *bs,
+                                    QDict *options, bool keep_old_opts)
+{
+    return bdrv_reopen_queue_child(bs_queue, bs, options, NULL, 0, false,
+                                   NULL, 0, keep_old_opts);
+}
+
 /*
  * Reopen multiple BlockDriverStates atomically & transactionally.
  *
@@ -1194,66 +3787,321 @@ BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue,
  *
  * Reopens all BDS specified in the queue, with the appropriate
  * flags.  All devices are prepared for reopen, and failure of any
- * device will cause all device changes to be abandonded, and intermediate
+ * device will cause all device changes to be abandoned, and intermediate
  * data cleaned up.
  *
  * If all devices prepare successfully, then the changes are committed
  * to all devices.
  *
+ * All affected nodes must be drained between bdrv_reopen_queue() and
+ * bdrv_reopen_multiple().
  */
 int bdrv_reopen_multiple(BlockReopenQueue *bs_queue, Error **errp)
 {
     int ret = -1;
     BlockReopenQueueEntry *bs_entry, *next;
-    Error *local_err = NULL;
 
     assert(bs_queue != NULL);
 
-    bdrv_drain_all();
-
-    QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) {
-        if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, &local_err)) {
-            error_propagate(errp, local_err);
+    QTAILQ_FOREACH(bs_entry, bs_queue, entry) {
+        assert(bs_entry->state.bs->quiesce_counter > 0);
+        if (bdrv_reopen_prepare(&bs_entry->state, bs_queue, errp)) {
             goto cleanup;
         }
         bs_entry->prepared = true;
     }
 
-    /* If we reach this point, we have success and just need to apply the
-     * changes
+    QTAILQ_FOREACH(bs_entry, bs_queue, entry) {
+        BDRVReopenState *state = &bs_entry->state;
+        ret = bdrv_check_perm(state->bs, bs_queue, state->perm,
+                              state->shared_perm, NULL, NULL, errp);
+        if (ret < 0) {
+            goto cleanup_perm;
+        }
+        /* Check if new_backing_bs would accept the new permissions */
+        if (state->replace_backing_bs && state->new_backing_bs) {
+            uint64_t nperm, nshared;
+            bdrv_child_perm(state->bs, state->new_backing_bs,
+                            NULL, bdrv_backing_role(state->bs),
+                            bs_queue, state->perm, state->shared_perm,
+                            &nperm, &nshared);
+            ret = bdrv_check_update_perm(state->new_backing_bs, NULL,
+                                         nperm, nshared, NULL, NULL, errp);
+            if (ret < 0) {
+                goto cleanup_perm;
+            }
+        }
+        bs_entry->perms_checked = true;
+    }
+
+    /*
+     * If we reach this point, we have success and just need to apply the
+     * changes.
+     *
+     * Reverse order is used to comfort qcow2 driver: on commit it need to write
+     * IN_USE flag to the image, to mark bitmaps in the image as invalid. But
+     * children are usually goes after parents in reopen-queue, so go from last
+     * to first element.
      */
-    QSIMPLEQ_FOREACH(bs_entry, bs_queue, entry) {
+    QTAILQ_FOREACH_REVERSE(bs_entry, bs_queue, entry) {
         bdrv_reopen_commit(&bs_entry->state);
     }
 
     ret = 0;
+cleanup_perm:
+    QTAILQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) {
+        BDRVReopenState *state = &bs_entry->state;
+
+        if (!bs_entry->perms_checked) {
+            continue;
+        }
+
+        if (ret == 0) {
+            bdrv_set_perm(state->bs, state->perm, state->shared_perm);
+        } else {
+            bdrv_abort_perm_update(state->bs);
+            if (state->replace_backing_bs && state->new_backing_bs) {
+                bdrv_abort_perm_update(state->new_backing_bs);
+            }
+        }
+    }
+
+    if (ret == 0) {
+        QTAILQ_FOREACH_REVERSE(bs_entry, bs_queue, entry) {
+            BlockDriverState *bs = bs_entry->state.bs;
 
+            if (bs->drv->bdrv_reopen_commit_post)
+                bs->drv->bdrv_reopen_commit_post(&bs_entry->state);
+        }
+    }
 cleanup:
-    QSIMPLEQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) {
-        if (ret && bs_entry->prepared) {
-            bdrv_reopen_abort(&bs_entry->state);
+    QTAILQ_FOREACH_SAFE(bs_entry, bs_queue, entry, next) {
+        if (ret) {
+            if (bs_entry->prepared) {
+                bdrv_reopen_abort(&bs_entry->state);
+            }
+            qobject_unref(bs_entry->state.explicit_options);
+            qobject_unref(bs_entry->state.options);
+        }
+        if (bs_entry->state.new_backing_bs) {
+            bdrv_unref(bs_entry->state.new_backing_bs);
         }
         g_free(bs_entry);
     }
     g_free(bs_queue);
+
+    return ret;
+}
+
+int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only,
+                              Error **errp)
+{
+    int ret;
+    BlockReopenQueue *queue;
+    QDict *opts = qdict_new();
+
+    qdict_put_bool(opts, BDRV_OPT_READ_ONLY, read_only);
+
+    bdrv_subtree_drained_begin(bs);
+    queue = bdrv_reopen_queue(NULL, bs, opts, true);
+    ret = bdrv_reopen_multiple(queue, errp);
+    bdrv_subtree_drained_end(bs);
+
     return ret;
 }
 
+static BlockReopenQueueEntry *find_parent_in_reopen_queue(BlockReopenQueue *q,
+                                                          BdrvChild *c)
+{
+    BlockReopenQueueEntry *entry;
+
+    QTAILQ_FOREACH(entry, q, entry) {
+        BlockDriverState *bs = entry->state.bs;
+        BdrvChild *child;
+
+        QLIST_FOREACH(child, &bs->children, next) {
+            if (child == c) {
+                return entry;
+            }
+        }
+    }
 
-/* Reopen a single BlockDriverState with the specified flags. */
-int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp)
+    return NULL;
+}
+
+static void bdrv_reopen_perm(BlockReopenQueue *q, BlockDriverState *bs,
+                             uint64_t *perm, uint64_t *shared)
 {
-    int ret = -1;
-    Error *local_err = NULL;
-    BlockReopenQueue *queue = bdrv_reopen_queue(NULL, bs, bdrv_flags);
+    BdrvChild *c;
+    BlockReopenQueueEntry *parent;
+    uint64_t cumulative_perms = 0;
+    uint64_t cumulative_shared_perms = BLK_PERM_ALL;
 
-    ret = bdrv_reopen_multiple(queue, &local_err);
-    if (local_err != NULL) {
-        error_propagate(errp, local_err);
+    QLIST_FOREACH(c, &bs->parents, next_parent) {
+        parent = find_parent_in_reopen_queue(q, c);
+        if (!parent) {
+            cumulative_perms |= c->perm;
+            cumulative_shared_perms &= c->shared_perm;
+        } else {
+            uint64_t nperm, nshared;
+
+            bdrv_child_perm(parent->state.bs, bs, c, c->role, q,
+                            parent->state.perm, parent->state.shared_perm,
+                            &nperm, &nshared);
+
+            cumulative_perms |= nperm;
+            cumulative_shared_perms &= nshared;
+        }
+    }
+    *perm = cumulative_perms;
+    *shared = cumulative_shared_perms;
+}
+
+static bool bdrv_reopen_can_attach(BlockDriverState *parent,
+                                   BdrvChild *child,
+                                   BlockDriverState *new_child,
+                                   Error **errp)
+{
+    AioContext *parent_ctx = bdrv_get_aio_context(parent);
+    AioContext *child_ctx = bdrv_get_aio_context(new_child);
+    GSList *ignore;
+    bool ret;
+
+    ignore = g_slist_prepend(NULL, child);
+    ret = bdrv_can_set_aio_context(new_child, parent_ctx, &ignore, NULL);
+    g_slist_free(ignore);
+    if (ret) {
+        return ret;
     }
+
+    ignore = g_slist_prepend(NULL, child);
+    ret = bdrv_can_set_aio_context(parent, child_ctx, &ignore, errp);
+    g_slist_free(ignore);
     return ret;
 }
 
+/*
+ * Take a BDRVReopenState and check if the value of 'backing' in the
+ * reopen_state->options QDict is valid or not.
+ *
+ * If 'backing' is missing from the QDict then return 0.
+ *
+ * If 'backing' contains the node name of the backing file of
+ * reopen_state->bs then return 0.
+ *
+ * If 'backing' contains a different node name (or is null) then check
+ * whether the current backing file can be replaced with the new one.
+ * If that's the case then reopen_state->replace_backing_bs is set to
+ * true and reopen_state->new_backing_bs contains a pointer to the new
+ * backing BlockDriverState (or NULL).
+ *
+ * Return 0 on success, otherwise return < 0 and set @errp.
+ */
+static int bdrv_reopen_parse_backing(BDRVReopenState *reopen_state,
+                                     Error **errp)
+{
+    BlockDriverState *bs = reopen_state->bs;
+    BlockDriverState *overlay_bs, *below_bs, *new_backing_bs;
+    QObject *value;
+    const char *str;
+
+    value = qdict_get(reopen_state->options, "backing");
+    if (value == NULL) {
+        return 0;
+    }
+
+    switch (qobject_type(value)) {
+    case QTYPE_QNULL:
+        new_backing_bs = NULL;
+        break;
+    case QTYPE_QSTRING:
+        str = qobject_get_try_str(value);
+        new_backing_bs = bdrv_lookup_bs(NULL, str, errp);
+        if (new_backing_bs == NULL) {
+            return -EINVAL;
+        } else if (bdrv_recurse_has_child(new_backing_bs, bs)) {
+            error_setg(errp, "Making '%s' a backing file of '%s' "
+                       "would create a cycle", str, bs->node_name);
+            return -EINVAL;
+        }
+        break;
+    default:
+        /* 'backing' does not allow any other data type */
+        g_assert_not_reached();
+    }
+
+    /*
+     * Check AioContext compatibility so that the bdrv_set_backing_hd() call in
+     * bdrv_reopen_commit() won't fail.
+     */
+    if (new_backing_bs) {
+        if (!bdrv_reopen_can_attach(bs, bs->backing, new_backing_bs, errp)) {
+            return -EINVAL;
+        }
+    }
+
+    /*
+     * Ensure that @bs can really handle backing files, because we are
+     * about to give it one (or swap the existing one)
+     */
+    if (bs->drv->is_filter) {
+        /* Filters always have a file or a backing child */
+        if (!bs->backing) {
+            error_setg(errp, "'%s' is a %s filter node that does not support a "
+                       "backing child", bs->node_name, bs->drv->format_name);
+            return -EINVAL;
+        }
+    } else if (!bs->drv->supports_backing) {
+        error_setg(errp, "Driver '%s' of node '%s' does not support backing "
+                   "files", bs->drv->format_name, bs->node_name);
+        return -EINVAL;
+    }
+
+    /*
+     * Find the "actual" backing file by skipping all links that point
+     * to an implicit node, if any (e.g. a commit filter node).
+     * We cannot use any of the bdrv_skip_*() functions here because
+     * those return the first explicit node, while we are looking for
+     * its overlay here.
+     */
+    overlay_bs = bs;
+    for (below_bs = bdrv_filter_or_cow_bs(overlay_bs);
+         below_bs && below_bs->implicit;
+         below_bs = bdrv_filter_or_cow_bs(overlay_bs))
+    {
+        overlay_bs = below_bs;
+    }
+
+    /* If we want to replace the backing file we need some extra checks */
+    if (new_backing_bs != bdrv_filter_or_cow_bs(overlay_bs)) {
+        /* Check for implicit nodes between bs and its backing file */
+        if (bs != overlay_bs) {
+            error_setg(errp, "Cannot change backing link if '%s' has "
+                       "an implicit backing file", bs->node_name);
+            return -EPERM;
+        }
+        /*
+         * Check if the backing link that we want to replace is frozen.
+         * Note that
+         * bdrv_filter_or_cow_child(overlay_bs) == overlay_bs->backing,
+         * because we know that overlay_bs == bs, and that @bs
+         * either is a filter that uses ->backing or a COW format BDS
+         * with bs->drv->supports_backing == true.
+         */
+        if (bdrv_is_backing_chain_frozen(overlay_bs,
+                                         child_bs(overlay_bs->backing), errp))
+        {
+            return -EPERM;
+        }
+        reopen_state->replace_backing_bs = true;
+        if (new_backing_bs) {
+            bdrv_ref(new_backing_bs);
+            reopen_state->new_backing_bs = new_backing_bs;
+        }
+    }
+
+    return 0;
+}
 
 /*
  * Prepares a BlockDriverState for reopen. All changes are staged in the
@@ -1276,36 +4124,98 @@ int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue,
                         Error **errp)
 {
     int ret = -1;
+    int old_flags;
     Error *local_err = NULL;
     BlockDriver *drv;
+    QemuOpts *opts;
+    QDict *orig_reopen_opts;
+    char *discard = NULL;
+    bool read_only;
+    bool drv_prepared = false;
 
     assert(reopen_state != NULL);
     assert(reopen_state->bs->drv != NULL);
     drv = reopen_state->bs->drv;
 
-    /* if we are to stay read-only, do not allow permission change
-     * to r/w */
-    if (!(reopen_state->bs->open_flags & BDRV_O_ALLOW_RDWR) &&
-        reopen_state->flags & BDRV_O_RDWR) {
-        error_set(errp, QERR_DEVICE_IS_READ_ONLY,
-                  reopen_state->bs->device_name);
+    /* This function and each driver's bdrv_reopen_prepare() remove
+     * entries from reopen_state->options as they are processed, so
+     * we need to make a copy of the original QDict. */
+    orig_reopen_opts = qdict_clone_shallow(reopen_state->options);
+
+    /* Process generic block layer options */
+    opts = qemu_opts_create(&bdrv_runtime_opts, NULL, 0, &error_abort);
+    if (!qemu_opts_absorb_qdict(opts, reopen_state->options, errp)) {
+        ret = -EINVAL;
+        goto error;
+    }
+
+    /* This was already called in bdrv_reopen_queue_child() so the flags
+     * are up-to-date. This time we simply want to remove the options from
+     * QemuOpts in order to indicate that they have been processed. */
+    old_flags = reopen_state->flags;
+    update_flags_from_options(&reopen_state->flags, opts);
+    assert(old_flags == reopen_state->flags);
+
+    discard = qemu_opt_get_del(opts, BDRV_OPT_DISCARD);
+    if (discard != NULL) {
+        if (bdrv_parse_discard_flags(discard, &reopen_state->flags) != 0) {
+            error_setg(errp, "Invalid discard option");
+            ret = -EINVAL;
+            goto error;
+        }
+    }
+
+    reopen_state->detect_zeroes =
+        bdrv_parse_detect_zeroes(opts, reopen_state->flags, &local_err);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        ret = -EINVAL;
         goto error;
     }
 
+    /* All other options (including node-name and driver) must be unchanged.
+     * Put them back into the QDict, so that they are checked at the end
+     * of this function. */
+    qemu_opts_to_qdict(opts, reopen_state->options);
+
+    /* If we are to stay read-only, do not allow permission change
+     * to r/w. Attempting to set to r/w may fail if either BDRV_O_ALLOW_RDWR is
+     * not set, or if the BDS still has copy_on_read enabled */
+    read_only = !(reopen_state->flags & BDRV_O_RDWR);
+    ret = bdrv_can_set_read_only(reopen_state->bs, read_only, true, &local_err);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        goto error;
+    }
+
+    /* Calculate required permissions after reopening */
+    bdrv_reopen_perm(queue, reopen_state->bs,
+                     &reopen_state->perm, &reopen_state->shared_perm);
 
     ret = bdrv_flush(reopen_state->bs);
     if (ret) {
-        error_set(errp, ERROR_CLASS_GENERIC_ERROR, "Error (%s) flushing drive",
-                  strerror(-ret));
+        error_setg_errno(errp, -ret, "Error flushing drive");
         goto error;
     }
 
     if (drv->bdrv_reopen_prepare) {
+        /*
+         * If a driver-specific option is missing, it means that we
+         * should reset it to its default value.
+         * But not all options allow that, so we need to check it first.
+         */
+        ret = bdrv_reset_options_allowed(reopen_state->bs,
+                                         reopen_state->options, errp);
+        if (ret) {
+            goto error;
+        }
+
         ret = drv->bdrv_reopen_prepare(reopen_state, queue, &local_err);
         if (ret) {
             if (local_err != NULL) {
                 error_propagate(errp, local_err);
             } else {
+                bdrv_refresh_filename(reopen_state->bs);
                 error_setg(errp, "failed while preparing to reopen image '%s'",
                            reopen_state->bs->filename);
             }
@@ -1314,16 +4224,110 @@ int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue,
     } else {
         /* It is currently mandatory to have a bdrv_reopen_prepare()
          * handler for each supported drv. */
-        error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
-                  drv->format_name, reopen_state->bs->device_name,
-                 "reopening of file");
+        error_setg(errp, "Block format '%s' used by node '%s' "
+                   "does not support reopening files", drv->format_name,
+                   bdrv_get_device_or_node_name(reopen_state->bs));
         ret = -1;
         goto error;
     }
 
+    drv_prepared = true;
+
+    /*
+     * We must provide the 'backing' option if the BDS has a backing
+     * file or if the image file has a backing file name as part of
+     * its metadata. Otherwise the 'backing' option can be omitted.
+     */
+    if (drv->supports_backing && reopen_state->backing_missing &&
+        (reopen_state->bs->backing || reopen_state->bs->backing_file[0])) {
+        error_setg(errp, "backing is missing for '%s'",
+                   reopen_state->bs->node_name);
+        ret = -EINVAL;
+        goto error;
+    }
+
+    /*
+     * Allow changing the 'backing' option. The new value can be
+     * either a reference to an existing node (using its node name)
+     * or NULL to simply detach the current backing file.
+     */
+    ret = bdrv_reopen_parse_backing(reopen_state, errp);
+    if (ret < 0) {
+        goto error;
+    }
+    qdict_del(reopen_state->options, "backing");
+
+    /* Options that are not handled are only okay if they are unchanged
+     * compared to the old state. It is expected that some options are only
+     * used for the initial open, but not reopen (e.g. filename) */
+    if (qdict_size(reopen_state->options)) {
+        const QDictEntry *entry = qdict_first(reopen_state->options);
+
+        do {
+            QObject *new = entry->value;
+            QObject *old = qdict_get(reopen_state->bs->options, entry->key);
+
+            /* Allow child references (child_name=node_name) as long as they
+             * point to the current child (i.e. everything stays the same). */
+            if (qobject_type(new) == QTYPE_QSTRING) {
+                BdrvChild *child;
+                QLIST_FOREACH(child, &reopen_state->bs->children, next) {
+                    if (!strcmp(child->name, entry->key)) {
+                        break;
+                    }
+                }
+
+                if (child) {
+                    const char *str = qobject_get_try_str(new);
+                    if (!strcmp(child->bs->node_name, str)) {
+                        continue; /* Found child with this name, skip option */
+                    }
+                }
+            }
+
+            /*
+             * TODO: When using -drive to specify blockdev options, all values
+             * will be strings; however, when using -blockdev, blockdev-add or
+             * filenames using the json:{} pseudo-protocol, they will be
+             * correctly typed.
+             * In contrast, reopening options are (currently) always strings
+             * (because you can only specify them through qemu-io; all other
+             * callers do not specify any options).
+             * Therefore, when using anything other than -drive to create a BDS,
+             * this cannot detect non-string options as unchanged, because
+             * qobject_is_equal() always returns false for objects of different
+             * type.  In the future, this should be remedied by correctly typing
+             * all options.  For now, this is not too big of an issue because
+             * the user can simply omit options which cannot be changed anyway,
+             * so they will stay unchanged.
+             */
+            if (!qobject_is_equal(new, old)) {
+                error_setg(errp, "Cannot change the option '%s'", entry->key);
+                ret = -EINVAL;
+                goto error;
+            }
+        } while ((entry = qdict_next(reopen_state->options, entry)));
+    }
+
     ret = 0;
 
+    /* Restore the original reopen_state->options QDict */
+    qobject_unref(reopen_state->options);
+    reopen_state->options = qobject_ref(orig_reopen_opts);
+
 error:
+    if (ret < 0 && drv_prepared) {
+        /* drv->bdrv_reopen_prepare() has succeeded, so we need to
+         * call drv->bdrv_reopen_abort() before signaling an error
+         * (bdrv_reopen_multiple() will not call bdrv_reopen_abort()
+         * when the respective bdrv_reopen_prepare() has failed) */
+        if (drv->bdrv_reopen_abort) {
+            drv->bdrv_reopen_abort(reopen_state);
+        }
+    }
+    qemu_opts_del(opts);
+    qobject_unref(orig_reopen_opts);
+    g_free(discard);
     return ret;
 }
 
@@ -1335,9 +4339,12 @@ error:
 void bdrv_reopen_commit(BDRVReopenState *reopen_state)
 {
     BlockDriver *drv;
+    BlockDriverState *bs;
+    BdrvChild *child;
 
     assert(reopen_state != NULL);
-    drv = reopen_state->bs->drv;
+    bs = reopen_state->bs;
+    drv = bs->drv;
     assert(drv != NULL);
 
     /* If there are any driver level actions to take */
@@ -1346,10 +4353,43 @@ void bdrv_reopen_commit(BDRVReopenState *reopen_state)
     }
 
     /* set BDS specific flags now */
-    reopen_state->bs->open_flags         = reopen_state->flags;
-    reopen_state->bs->enable_write_cache = !!(reopen_state->flags &
-                                              BDRV_O_CACHE_WB);
-    reopen_state->bs->read_only = !(reopen_state->flags & BDRV_O_RDWR);
+    qobject_unref(bs->explicit_options);
+    qobject_unref(bs->options);
+
+    bs->explicit_options   = reopen_state->explicit_options;
+    bs->options            = reopen_state->options;
+    bs->open_flags         = reopen_state->flags;
+    bs->read_only = !(reopen_state->flags & BDRV_O_RDWR);
+    bs->detect_zeroes      = reopen_state->detect_zeroes;
+
+    if (reopen_state->replace_backing_bs) {
+        qdict_del(bs->explicit_options, "backing");
+        qdict_del(bs->options, "backing");
+    }
+
+    /* Remove child references from bs->options and bs->explicit_options.
+     * Child options were already removed in bdrv_reopen_queue_child() */
+    QLIST_FOREACH(child, &bs->children, next) {
+        qdict_del(bs->explicit_options, child->name);
+        qdict_del(bs->options, child->name);
+    }
+
+    /*
+     * Change the backing file if a new one was specified. We do this
+     * after updating bs->options, so bdrv_refresh_filename() (called
+     * from bdrv_set_backing_hd()) has the new values.
+     */
+    if (reopen_state->replace_backing_bs) {
+        BlockDriverState *old_backing_bs = child_bs(bs->backing);
+        assert(!old_backing_bs || !old_backing_bs->implicit);
+        /* Abort the permission update on the backing bs we're detaching */
+        if (old_backing_bs) {
+            bdrv_abort_perm_update(old_backing_bs);
+        }
+        bdrv_set_backing_hd(bs, reopen_state->new_backing_bs, &error_abort);
+    }
+
+    bdrv_refresh_limits(bs, NULL);
 }
 
 /*
@@ -1370,209 +4410,236 @@ void bdrv_reopen_abort(BDRVReopenState *reopen_state)
 }
 
 
-void bdrv_close(BlockDriverState *bs)
+static void bdrv_close(BlockDriverState *bs)
 {
-    if (bs->job) {
-        block_job_cancel_sync(bs->job);
-    }
-    bdrv_drain_all(); /* complete I/O */
+    BdrvAioNotifier *ban, *ban_next;
+    BdrvChild *child, *next;
+
+    assert(!bs->refcnt);
+
+    bdrv_drained_begin(bs); /* complete I/O */
     bdrv_flush(bs);
-    bdrv_drain_all(); /* in case flush left pending I/O */
-    notifier_list_notify(&bs->close_notifiers, bs);
+    bdrv_drain(bs); /* in case flush left pending I/O */
 
     if (bs->drv) {
-        if (bs->backing_hd) {
-            bdrv_delete(bs->backing_hd);
-            bs->backing_hd = NULL;
-        }
-        bs->drv->bdrv_close(bs);
-        g_free(bs->opaque);
-#ifdef _WIN32
-        if (bs->is_temporary) {
-            unlink(bs->filename);
+        if (bs->drv->bdrv_close) {
+            /* Must unfreeze all children, so bdrv_unref_child() works */
+            bs->drv->bdrv_close(bs);
         }
-#endif
-        bs->opaque = NULL;
         bs->drv = NULL;
-        bs->copy_on_read = 0;
-        bs->backing_file[0] = '\0';
-        bs->backing_format[0] = '\0';
-        bs->total_sectors = 0;
-        bs->encrypted = 0;
-        bs->valid_key = 0;
-        bs->sg = 0;
-        bs->growable = 0;
-        QDECREF(bs->options);
-        bs->options = NULL;
+    }
 
-        if (bs->file != NULL) {
-            bdrv_delete(bs->file);
-            bs->file = NULL;
-        }
+    QLIST_FOREACH_SAFE(child, &bs->children, next, next) {
+        bdrv_unref_child(bs, child);
     }
 
-    bdrv_dev_change_media_cb(bs, false);
+    bs->backing = NULL;
+    bs->file = NULL;
+    g_free(bs->opaque);
+    bs->opaque = NULL;
+    qatomic_set(&bs->copy_on_read, 0);
+    bs->backing_file[0] = '\0';
+    bs->backing_format[0] = '\0';
+    bs->total_sectors = 0;
+    bs->encrypted = false;
+    bs->sg = false;
+    qobject_unref(bs->options);
+    qobject_unref(bs->explicit_options);
+    bs->options = NULL;
+    bs->explicit_options = NULL;
+    qobject_unref(bs->full_open_options);
+    bs->full_open_options = NULL;
+
+    bdrv_release_named_dirty_bitmaps(bs);
+    assert(QLIST_EMPTY(&bs->dirty_bitmaps));
 
-    /*throttling disk I/O limits*/
-    if (bs->io_limits_enabled) {
-        bdrv_io_limits_disable(bs);
+    QLIST_FOREACH_SAFE(ban, &bs->aio_notifiers, list, ban_next) {
+        g_free(ban);
     }
-}
-
-void bdrv_close_all(void)
-{
-    BlockDriverState *bs;
+    QLIST_INIT(&bs->aio_notifiers);
+    bdrv_drained_end(bs);
 
-    QTAILQ_FOREACH(bs, &bdrv_states, list) {
-        bdrv_close(bs);
+    /*
+     * If we're still inside some bdrv_drain_all_begin()/end() sections, end
+     * them now since this BDS won't exist anymore when bdrv_drain_all_end()
+     * gets called.
+     */
+    if (bs->quiesce_counter) {
+        bdrv_drain_all_end_quiesce(bs);
     }
 }
 
-/*
- * Wait for pending requests to complete across all BlockDriverStates
- *
- * This function does not flush data to disk, use bdrv_flush_all() for that
- * after calling this function.
- *
- * Note that completion of an asynchronous I/O operation can trigger any
- * number of other I/O operations on other devices---for example a coroutine
- * can be arbitrarily complex and a constant flow of I/O can come until the
- * coroutine is complete.  Because of this, it is not possible to have a
- * function to drain a single device's I/O queue.
- */
-void bdrv_drain_all(void)
+void bdrv_close_all(void)
 {
-    BlockDriverState *bs;
-    bool busy;
+    assert(job_next(NULL) == NULL);
+    blk_exp_close_all();
 
-    do {
-        busy = qemu_aio_wait();
+    /* Drop references from requests still in flight, such as canceled block
+     * jobs whose AIO context has not been polled yet */
+    bdrv_drain_all();
 
-        /* FIXME: We do not have timer support here, so this is effectively
-         * a busy wait.
-         */
-        QTAILQ_FOREACH(bs, &bdrv_states, list) {
-            while (qemu_co_enter_next(&bs->throttled_reqs)) {
-                busy = true;
-            }
-        }
-    } while (busy);
+    blk_remove_all_bs();
+    blockdev_close_all_bdrv_states();
 
-    /* If requests are still pending there is a bug somewhere */
-    QTAILQ_FOREACH(bs, &bdrv_states, list) {
-        assert(QLIST_EMPTY(&bs->tracked_requests));
-        assert(qemu_co_queue_empty(&bs->throttled_reqs));
-    }
+    assert(QTAILQ_EMPTY(&all_bdrv_states));
 }
 
-/* make a BlockDriverState anonymous by removing from bdrv_state list.
-   Also, NULL terminate the device_name to prevent double remove */
-void bdrv_make_anon(BlockDriverState *bs)
+static bool should_update_child(BdrvChild *c, BlockDriverState *to)
 {
-    if (bs->device_name[0] != '\0') {
-        QTAILQ_REMOVE(&bdrv_states, bs, list);
-    }
-    bs->device_name[0] = '\0';
-}
+    GQueue *queue;
+    GHashTable *found;
+    bool ret;
 
-static void bdrv_rebind(BlockDriverState *bs)
-{
-    if (bs->drv && bs->drv->bdrv_rebind) {
-        bs->drv->bdrv_rebind(bs);
+    if (c->klass->stay_at_node) {
+        return false;
     }
-}
-
-static void bdrv_move_feature_fields(BlockDriverState *bs_dest,
-                                     BlockDriverState *bs_src)
-{
-    /* move some fields that need to stay attached to the device */
-    bs_dest->open_flags         = bs_src->open_flags;
 
-    /* dev info */
-    bs_dest->dev_ops            = bs_src->dev_ops;
-    bs_dest->dev_opaque         = bs_src->dev_opaque;
-    bs_dest->dev                = bs_src->dev;
-    bs_dest->buffer_alignment   = bs_src->buffer_alignment;
-    bs_dest->copy_on_read       = bs_src->copy_on_read;
+    /* If the child @c belongs to the BDS @to, replacing the current
+     * c->bs by @to would mean to create a loop.
+     *
+     * Such a case occurs when appending a BDS to a backing chain.
+     * For instance, imagine the following chain:
+     *
+     *   guest device -> node A -> further backing chain...
+     *
+     * Now we create a new BDS B which we want to put on top of this
+     * chain, so we first attach A as its backing node:
+     *
+     *                   node B
+     *                     |
+     *                     v
+     *   guest device -> node A -> further backing chain...
+     *
+     * Finally we want to replace A by B.  When doing that, we want to
+     * replace all pointers to A by pointers to B -- except for the
+     * pointer from B because (1) that would create a loop, and (2)
+     * that pointer should simply stay intact:
+     *
+     *   guest device -> node B
+     *                     |
+     *                     v
+     *                   node A -> further backing chain...
+     *
+     * In general, when replacing a node A (c->bs) by a node B (@to),
+     * if A is a child of B, that means we cannot replace A by B there
+     * because that would create a loop.  Silently detaching A from B
+     * is also not really an option.  So overall just leaving A in
+     * place there is the most sensible choice.
+     *
+     * We would also create a loop in any cases where @c is only
+     * indirectly referenced by @to. Prevent this by returning false
+     * if @c is found (by breadth-first search) anywhere in the whole
+     * subtree of @to.
+     */
 
-    bs_dest->enable_write_cache = bs_src->enable_write_cache;
+    ret = true;
+    found = g_hash_table_new(NULL, NULL);
+    g_hash_table_add(found, to);
+    queue = g_queue_new();
+    g_queue_push_tail(queue, to);
 
-    /* i/o timing parameters */
-    bs_dest->slice_start        = bs_src->slice_start;
-    bs_dest->slice_end          = bs_src->slice_end;
-    bs_dest->slice_submitted    = bs_src->slice_submitted;
-    bs_dest->io_limits          = bs_src->io_limits;
-    bs_dest->throttled_reqs     = bs_src->throttled_reqs;
-    bs_dest->block_timer        = bs_src->block_timer;
-    bs_dest->io_limits_enabled  = bs_src->io_limits_enabled;
+    while (!g_queue_is_empty(queue)) {
+        BlockDriverState *v = g_queue_pop_head(queue);
+        BdrvChild *c2;
 
-    /* r/w error */
-    bs_dest->on_read_error      = bs_src->on_read_error;
-    bs_dest->on_write_error     = bs_src->on_write_error;
+        QLIST_FOREACH(c2, &v->children, next) {
+            if (c2 == c) {
+                ret = false;
+                break;
+            }
 
-    /* i/o status */
-    bs_dest->iostatus_enabled   = bs_src->iostatus_enabled;
-    bs_dest->iostatus           = bs_src->iostatus;
+            if (g_hash_table_contains(found, c2->bs)) {
+                continue;
+            }
 
-    /* dirty bitmap */
-    bs_dest->dirty_bitmap       = bs_src->dirty_bitmap;
+            g_queue_push_tail(queue, c2->bs);
+            g_hash_table_add(found, c2->bs);
+        }
+    }
 
-    /* job */
-    bs_dest->in_use             = bs_src->in_use;
-    bs_dest->job                = bs_src->job;
+    g_queue_free(queue);
+    g_hash_table_destroy(found);
 
-    /* keep the same entry in bdrv_states */
-    pstrcpy(bs_dest->device_name, sizeof(bs_dest->device_name),
-            bs_src->device_name);
-    bs_dest->list = bs_src->list;
+    return ret;
 }
 
 /*
- * Swap bs contents for two image chains while they are live,
- * while keeping required fields on the BlockDriverState that is
- * actually attached to a device.
- *
- * This will modify the BlockDriverState fields, and swap contents
- * between bs_new and bs_old. Both bs_new and bs_old are modified.
- *
- * bs_new is required to be anonymous.
+ * With auto_skip=true bdrv_replace_node_common skips updating from parents
+ * if it creates a parent-child relation loop or if parent is block-job.
  *
- * This function does not create any image files.
+ * With auto_skip=false the error is returned if from has a parent which should
+ * not be updated.
  */
-void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old)
+static void bdrv_replace_node_common(BlockDriverState *from,
+                                     BlockDriverState *to,
+                                     bool auto_skip, Error **errp)
 {
-    BlockDriverState tmp;
+    BdrvChild *c, *next;
+    GSList *list = NULL, *p;
+    uint64_t perm = 0, shared = BLK_PERM_ALL;
+    int ret;
 
-    /* bs_new must be anonymous and shouldn't have anything fancy enabled */
-    assert(bs_new->device_name[0] == '\0');
-    assert(bs_new->dirty_bitmap == NULL);
-    assert(bs_new->job == NULL);
-    assert(bs_new->dev == NULL);
-    assert(bs_new->in_use == 0);
-    assert(bs_new->io_limits_enabled == false);
-    assert(bs_new->block_timer == NULL);
+    /* Make sure that @from doesn't go away until we have successfully attached
+     * all of its parents to @to. */
+    bdrv_ref(from);
 
-    tmp = *bs_new;
-    *bs_new = *bs_old;
-    *bs_old = tmp;
+    assert(qemu_get_current_aio_context() == qemu_get_aio_context());
+    assert(bdrv_get_aio_context(from) == bdrv_get_aio_context(to));
+    bdrv_drained_begin(from);
 
-    /* there are some fields that should not be swapped, move them back */
-    bdrv_move_feature_fields(&tmp, bs_old);
-    bdrv_move_feature_fields(bs_old, bs_new);
-    bdrv_move_feature_fields(bs_new, &tmp);
+    /* Put all parents into @list and calculate their cumulative permissions */
+    QLIST_FOREACH_SAFE(c, &from->parents, next_parent, next) {
+        assert(c->bs == from);
+        if (!should_update_child(c, to)) {
+            if (auto_skip) {
+                continue;
+            }
+            error_setg(errp, "Should not change '%s' link to '%s'",
+                       c->name, from->node_name);
+            goto out;
+        }
+        if (c->frozen) {
+            error_setg(errp, "Cannot change '%s' link to '%s'",
+                       c->name, from->node_name);
+            goto out;
+        }
+        list = g_slist_prepend(list, c);
+        perm |= c->perm;
+        shared &= c->shared_perm;
+    }
+
+    /* Check whether the required permissions can be granted on @to, ignoring
+     * all BdrvChild in @list so that they can't block themselves. */
+    ret = bdrv_check_update_perm(to, NULL, perm, shared, list, NULL, errp);
+    if (ret < 0) {
+        bdrv_abort_perm_update(to);
+        goto out;
+    }
+
+    /* Now actually perform the change. We performed the permission check for
+     * all elements of @list at once, so set the permissions all at once at the
+     * very end. */
+    for (p = list; p != NULL; p = p->next) {
+        c = p->data;
 
-    /* bs_new shouldn't be in bdrv_states even after the swap!  */
-    assert(bs_new->device_name[0] == '\0');
+        bdrv_ref(to);
+        bdrv_replace_child_noperm(c, to);
+        bdrv_unref(from);
+    }
 
-    /* Check a few fields that should remain attached to the device */
-    assert(bs_new->dev == NULL);
-    assert(bs_new->job == NULL);
-    assert(bs_new->in_use == 0);
-    assert(bs_new->io_limits_enabled == false);
-    assert(bs_new->block_timer == NULL);
+    bdrv_get_cumulative_perm(to, &perm, &shared);
+    bdrv_set_perm(to, perm, shared);
 
-    bdrv_rebind(bs_new);
-    bdrv_rebind(bs_old);
+out:
+    g_slist_free(list);
+    bdrv_drained_end(from);
+    bdrv_unref(from);
+}
+
+void bdrv_replace_node(BlockDriverState *from, BlockDriverState *to,
+                       Error **errp)
+{
+    return bdrv_replace_node_common(from, to, true, errp);
 }
 
 /*
@@ -1582,170 +4649,53 @@ void bdrv_swap(BlockDriverState *bs_new, BlockDriverState *bs_old)
  * This will modify the BlockDriverState fields, and swap contents
  * between bs_new and bs_top. Both bs_new and bs_top are modified.
  *
- * bs_new is required to be anonymous.
+ * bs_new must not be attached to a BlockBackend.
  *
  * This function does not create any image files.
+ *
+ * bdrv_append() takes ownership of a bs_new reference and unrefs it because
+ * that's what the callers commonly need. bs_new will be referenced by the old
+ * parents of bs_top after bdrv_append() returns. If the caller needs to keep a
+ * reference of its own, it must call bdrv_ref().
  */
-void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top)
-{
-    bdrv_swap(bs_new, bs_top);
-
-    /* The contents of 'tmp' will become bs_top, as we are
-     * swapping bs_new and bs_top contents. */
-    bs_top->backing_hd = bs_new;
-    bs_top->open_flags &= ~BDRV_O_NO_BACKING;
-    pstrcpy(bs_top->backing_file, sizeof(bs_top->backing_file),
-            bs_new->filename);
-    pstrcpy(bs_top->backing_format, sizeof(bs_top->backing_format),
-            bs_new->drv ? bs_new->drv->format_name : "");
-}
-
-void bdrv_delete(BlockDriverState *bs)
-{
-    assert(!bs->dev);
-    assert(!bs->job);
-    assert(!bs->in_use);
-
-    /* remove from list, if necessary */
-    bdrv_make_anon(bs);
-
-    bdrv_close(bs);
-
-    g_free(bs);
-}
-
-int bdrv_attach_dev(BlockDriverState *bs, void *dev)
-/* TODO change to DeviceState *dev when all users are qdevified */
-{
-    if (bs->dev) {
-        return -EBUSY;
-    }
-    bs->dev = dev;
-    bdrv_iostatus_reset(bs);
-    return 0;
-}
-
-/* TODO qdevified devices don't use this, remove when devices are qdevified */
-void bdrv_attach_dev_nofail(BlockDriverState *bs, void *dev)
+void bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top,
+                 Error **errp)
 {
-    if (bdrv_attach_dev(bs, dev) < 0) {
-        abort();
-    }
-}
-
-void bdrv_detach_dev(BlockDriverState *bs, void *dev)
-/* TODO change to DeviceState *dev when all users are qdevified */
-{
-    assert(bs->dev == dev);
-    bs->dev = NULL;
-    bs->dev_ops = NULL;
-    bs->dev_opaque = NULL;
-    bs->buffer_alignment = 512;
-}
-
-/* TODO change to return DeviceState * when all users are qdevified */
-void *bdrv_get_attached_dev(BlockDriverState *bs)
-{
-    return bs->dev;
-}
-
-void bdrv_set_dev_ops(BlockDriverState *bs, const BlockDevOps *ops,
-                      void *opaque)
-{
-    bs->dev_ops = ops;
-    bs->dev_opaque = opaque;
-}
-
-void bdrv_emit_qmp_error_event(const BlockDriverState *bdrv,
-                               enum MonitorEvent ev,
-                               BlockErrorAction action, bool is_read)
-{
-    QObject *data;
-    const char *action_str;
+    Error *local_err = NULL;
 
-    switch (action) {
-    case BDRV_ACTION_REPORT:
-        action_str = "report";
-        break;
-    case BDRV_ACTION_IGNORE:
-        action_str = "ignore";
-        break;
-    case BDRV_ACTION_STOP:
-        action_str = "stop";
-        break;
-    default:
-        abort();
+    bdrv_set_backing_hd(bs_new, bs_top, &local_err);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        goto out;
     }
 
-    data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
-                              bdrv->device_name,
-                              action_str,
-                              is_read ? "read" : "write");
-    monitor_protocol_event(ev, data);
-
-    qobject_decref(data);
-}
-
-static void bdrv_emit_qmp_eject_event(BlockDriverState *bs, bool ejected)
-{
-    QObject *data;
-
-    data = qobject_from_jsonf("{ 'device': %s, 'tray-open': %i }",
-                              bdrv_get_device_name(bs), ejected);
-    monitor_protocol_event(QEVENT_DEVICE_TRAY_MOVED, data);
-
-    qobject_decref(data);
-}
-
-static void bdrv_dev_change_media_cb(BlockDriverState *bs, bool load)
-{
-    if (bs->dev_ops && bs->dev_ops->change_media_cb) {
-        bool tray_was_closed = !bdrv_dev_is_tray_open(bs);
-        bs->dev_ops->change_media_cb(bs->dev_opaque, load);
-        if (tray_was_closed) {
-            /* tray open */
-            bdrv_emit_qmp_eject_event(bs, true);
-        }
-        if (load) {
-            /* tray close */
-            bdrv_emit_qmp_eject_event(bs, false);
-        }
+    bdrv_replace_node(bs_top, bs_new, &local_err);
+    if (local_err) {
+        error_propagate(errp, local_err);
+        bdrv_set_backing_hd(bs_new, NULL, &error_abort);
+        goto out;
     }
-}
 
-bool bdrv_dev_has_removable_media(BlockDriverState *bs)
-{
-    return !bs->dev || (bs->dev_ops && bs->dev_ops->change_media_cb);
+    /* bs_new is now referenced by its new parents, we don't need the
+     * additional reference any more. */
+out:
+    bdrv_unref(bs_new);
 }
 
-void bdrv_dev_eject_request(BlockDriverState *bs, bool force)
+static void bdrv_delete(BlockDriverState *bs)
 {
-    if (bs->dev_ops && bs->dev_ops->eject_request_cb) {
-        bs->dev_ops->eject_request_cb(bs->dev_opaque, force);
-    }
-}
+    assert(bdrv_op_blocker_is_empty(bs));
+    assert(!bs->refcnt);
 
-bool bdrv_dev_is_tray_open(BlockDriverState *bs)
-{
-    if (bs->dev_ops && bs->dev_ops->is_tray_open) {
-        return bs->dev_ops->is_tray_open(bs->dev_opaque);
+    /* remove from list, if necessary */
+    if (bs->node_name[0] != '\0') {
+        QTAILQ_REMOVE(&graph_bdrv_states, bs, node_list);
     }
-    return false;
-}
+    QTAILQ_REMOVE(&all_bdrv_states, bs, bs_list);
 
-static void bdrv_dev_resize_cb(BlockDriverState *bs)
-{
-    if (bs->dev_ops && bs->dev_ops->resize_cb) {
-        bs->dev_ops->resize_cb(bs->dev_opaque);
-    }
-}
+    bdrv_close(bs);
 
-bool bdrv_dev_is_medium_locked(BlockDriverState *bs)
-{
-    if (bs->dev_ops && bs->dev_ops->is_medium_locked) {
-        return bs->dev_ops->is_medium_locked(bs->dev_opaque);
-    }
-    return false;
+    g_free(bs);
 }
 
 /*
@@ -1755,281 +4705,177 @@ bool bdrv_dev_is_medium_locked(BlockDriverState *bs)
  * free of errors) or -errno when an internal error occurred. The results of the
  * check are stored in res.
  */
-int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res, BdrvCheckMode fix)
+int coroutine_fn bdrv_co_check(BlockDriverState *bs,
+                               BdrvCheckResult *res, BdrvCheckMode fix)
 {
-    if (bs->drv->bdrv_check == NULL) {
+    if (bs->drv == NULL) {
+        return -ENOMEDIUM;
+    }
+    if (bs->drv->bdrv_co_check == NULL) {
         return -ENOTSUP;
     }
 
     memset(res, 0, sizeof(*res));
-    return bs->drv->bdrv_check(bs, res, fix);
+    return bs->drv->bdrv_co_check(bs, res, fix);
 }
 
-#define COMMIT_BUF_SECTORS 2048
-
-/* commit COW file into the raw image */
-int bdrv_commit(BlockDriverState *bs)
+/*
+ * Return values:
+ * 0        - success
+ * -EINVAL  - backing format specified, but no file
+ * -ENOSPC  - can't update the backing file because no space is left in the
+ *            image file header
+ * -ENOTSUP - format driver doesn't support changing the backing file
+ */
+int bdrv_change_backing_file(BlockDriverState *bs, const char *backing_file,
+                             const char *backing_fmt, bool warn)
 {
     BlockDriver *drv = bs->drv;
-    int64_t sector, total_sectors;
-    int n, ro, open_flags;
-    int ret = 0;
-    uint8_t *buf;
-    char filename[PATH_MAX];
+    int ret;
 
-    if (!drv)
+    if (!drv) {
         return -ENOMEDIUM;
-    
-    if (!bs->backing_hd) {
-        return -ENOTSUP;
-    }
-
-    if (bdrv_in_use(bs) || bdrv_in_use(bs->backing_hd)) {
-        return -EBUSY;
     }
 
-    ro = bs->backing_hd->read_only;
-    /* Use pstrcpy (not strncpy): filename must be NUL-terminated. */
-    pstrcpy(filename, sizeof(filename), bs->backing_hd->filename);
-    open_flags =  bs->backing_hd->open_flags;
-
-    if (ro) {
-        if (bdrv_reopen(bs->backing_hd, open_flags | BDRV_O_RDWR, NULL)) {
-            return -EACCES;
-        }
+    /* Backing file format doesn't make sense without a backing file */
+    if (backing_fmt && !backing_file) {
+        return -EINVAL;
     }
 
-    total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
-    buf = g_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE);
-
-    for (sector = 0; sector < total_sectors; sector += n) {
-        if (bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) {
-
-            if (bdrv_read(bs, sector, buf, n) != 0) {
-                ret = -EIO;
-                goto ro_cleanup;
-            }
-
-            if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) {
-                ret = -EIO;
-                goto ro_cleanup;
-            }
-        }
+    if (warn && backing_file && !backing_fmt) {
+        warn_report("Deprecated use of backing file without explicit "
+                    "backing format, use of this image requires "
+                    "potentially unsafe format probing");
     }
 
-    if (drv->bdrv_make_empty) {
-        ret = drv->bdrv_make_empty(bs);
-        bdrv_flush(bs);
+    if (drv->bdrv_change_backing_file != NULL) {
+        ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
+    } else {
+        ret = -ENOTSUP;
     }
 
-    /*
-     * Make sure all data we wrote to the backing device is actually
-     * stable on disk.
-     */
-    if (bs->backing_hd)
-        bdrv_flush(bs->backing_hd);
-
-ro_cleanup:
-    g_free(buf);
-
-    if (ro) {
-        /* ignoring error return here */
-        bdrv_reopen(bs->backing_hd, open_flags & ~BDRV_O_RDWR, NULL);
+    if (ret == 0) {
+        pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: "");
+        pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: "");
+        pstrcpy(bs->auto_backing_file, sizeof(bs->auto_backing_file),
+                backing_file ?: "");
     }
-
     return ret;
 }
 
-int bdrv_commit_all(void)
+/*
+ * Finds the first non-filter node above bs in the chain between
+ * active and bs.  The returned node is either an immediate parent of
+ * bs, or there are only filter nodes between the two.
+ *
+ * Returns NULL if bs is not found in active's image chain,
+ * or if active == bs.
+ *
+ * Returns the bottommost base image if bs == NULL.
+ */
+BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
+                                    BlockDriverState *bs)
 {
-    BlockDriverState *bs;
+    bs = bdrv_skip_filters(bs);
+    active = bdrv_skip_filters(active);
 
-    QTAILQ_FOREACH(bs, &bdrv_states, list) {
-        if (bs->drv && bs->backing_hd) {
-            int ret = bdrv_commit(bs);
-            if (ret < 0) {
-                return ret;
-            }
+    while (active) {
+        BlockDriverState *next = bdrv_backing_chain_next(active);
+        if (bs == next) {
+            return active;
         }
+        active = next;
     }
-    return 0;
-}
 
-/**
- * Remove an active request from the tracked requests list
- *
- * This function should be called when a tracked request is completing.
- */
-static void tracked_request_end(BdrvTrackedRequest *req)
-{
-    QLIST_REMOVE(req, list);
-    qemu_co_queue_restart_all(&req->wait_queue);
+    return NULL;
 }
 
-/**
- * Add an active request to the tracked requests list
- */
-static void tracked_request_begin(BdrvTrackedRequest *req,
-                                  BlockDriverState *bs,
-                                  int64_t sector_num,
-                                  int nb_sectors, bool is_write)
-{
-    *req = (BdrvTrackedRequest){
-        .bs = bs,
-        .sector_num = sector_num,
-        .nb_sectors = nb_sectors,
-        .is_write = is_write,
-        .co = qemu_coroutine_self(),
-    };
-
-    qemu_co_queue_init(&req->wait_queue);
-
-    QLIST_INSERT_HEAD(&bs->tracked_requests, req, list);
+/* Given a BDS, searches for the base layer. */
+BlockDriverState *bdrv_find_base(BlockDriverState *bs)
+{
+    return bdrv_find_overlay(bs, NULL);
 }
 
-/**
- * Round a region to cluster boundaries
+/*
+ * Return true if at least one of the COW (backing) and filter links
+ * between @bs and @base is frozen. @errp is set if that's the case.
+ * @base must be reachable from @bs, or NULL.
  */
-void bdrv_round_to_clusters(BlockDriverState *bs,
-                            int64_t sector_num, int nb_sectors,
-                            int64_t *cluster_sector_num,
-                            int *cluster_nb_sectors)
+bool bdrv_is_backing_chain_frozen(BlockDriverState *bs, BlockDriverState *base,
+                                  Error **errp)
 {
-    BlockDriverInfo bdi;
+    BlockDriverState *i;
+    BdrvChild *child;
 
-    if (bdrv_get_info(bs, &bdi) < 0 || bdi.cluster_size == 0) {
-        *cluster_sector_num = sector_num;
-        *cluster_nb_sectors = nb_sectors;
-    } else {
-        int64_t c = bdi.cluster_size / BDRV_SECTOR_SIZE;
-        *cluster_sector_num = QEMU_ALIGN_DOWN(sector_num, c);
-        *cluster_nb_sectors = QEMU_ALIGN_UP(sector_num - *cluster_sector_num +
-                                            nb_sectors, c);
-    }
-}
+    for (i = bs; i != base; i = child_bs(child)) {
+        child = bdrv_filter_or_cow_child(i);
 
-static bool tracked_request_overlaps(BdrvTrackedRequest *req,
-                                     int64_t sector_num, int nb_sectors) {
-    /*        aaaa   bbbb */
-    if (sector_num >= req->sector_num + req->nb_sectors) {
-        return false;
-    }
-    /* bbbb   aaaa        */
-    if (req->sector_num >= sector_num + nb_sectors) {
-        return false;
+        if (child && child->frozen) {
+            error_setg(errp, "Cannot change '%s' link from '%s' to '%s'",
+                       child->name, i->node_name, child->bs->node_name);
+            return true;
+        }
     }
-    return true;
-}
-
-static void coroutine_fn wait_for_overlapping_requests(BlockDriverState *bs,
-        int64_t sector_num, int nb_sectors)
-{
-    BdrvTrackedRequest *req;
-    int64_t cluster_sector_num;
-    int cluster_nb_sectors;
-    bool retry;
 
-    /* If we touch the same cluster it counts as an overlap.  This guarantees
-     * that allocating writes will be serialized and not race with each other
-     * for the same cluster.  For example, in copy-on-read it ensures that the
-     * CoR read and write operations are atomic and guest writes cannot
-     * interleave between them.
-     */
-    bdrv_round_to_clusters(bs, sector_num, nb_sectors,
-                           &cluster_sector_num, &cluster_nb_sectors);
-
-    do {
-        retry = false;
-        QLIST_FOREACH(req, &bs->tracked_requests, list) {
-            if (tracked_request_overlaps(req, cluster_sector_num,
-                                         cluster_nb_sectors)) {
-                /* Hitting this means there was a reentrant request, for
-                 * example, a block driver issuing nested requests.  This must
-                 * never happen since it means deadlock.
-                 */
-                assert(qemu_coroutine_self() != req->co);
-
-                qemu_co_queue_wait(&req->wait_queue);
-                retry = true;
-                break;
-            }
-        }
-    } while (retry);
+    return false;
 }
 
 /*
- * Return values:
- * 0        - success
- * -EINVAL  - backing format specified, but no file
- * -ENOSPC  - can't update the backing file because no space is left in the
- *            image file header
- * -ENOTSUP - format driver doesn't support changing the backing file
+ * Freeze all COW (backing) and filter links between @bs and @base.
+ * If any of the links is already frozen the operation is aborted and
+ * none of the links are modified.
+ * @base must be reachable from @bs, or NULL.
+ * Returns 0 on success. On failure returns < 0 and sets @errp.
  */
-int bdrv_change_backing_file(BlockDriverState *bs,
-    const char *backing_file, const char *backing_fmt)
+int bdrv_freeze_backing_chain(BlockDriverState *bs, BlockDriverState *base,
+                              Error **errp)
 {
-    BlockDriver *drv = bs->drv;
-    int ret;
+    BlockDriverState *i;
+    BdrvChild *child;
 
-    /* Backing file format doesn't make sense without a backing file */
-    if (backing_fmt && !backing_file) {
-        return -EINVAL;
+    if (bdrv_is_backing_chain_frozen(bs, base, errp)) {
+        return -EPERM;
     }
 
-    if (drv->bdrv_change_backing_file != NULL) {
-        ret = drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
-    } else {
-        ret = -ENOTSUP;
+    for (i = bs; i != base; i = child_bs(child)) {
+        child = bdrv_filter_or_cow_child(i);
+        if (child && child->bs->never_freeze) {
+            error_setg(errp, "Cannot freeze '%s' link to '%s'",
+                       child->name, child->bs->node_name);
+            return -EPERM;
+        }
     }
 
-    if (ret == 0) {
-        pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: "");
-        pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: "");
+    for (i = bs; i != base; i = child_bs(child)) {
+        child = bdrv_filter_or_cow_child(i);
+        if (child) {
+            child->frozen = true;
+        }
     }
-    return ret;
+
+    return 0;
 }
 
 /*
- * Finds the image layer in the chain that has 'bs' as its backing file.
- *
- * active is the current topmost image.
- *
- * Returns NULL if bs is not found in active's image chain,
- * or if active == bs.
+ * Unfreeze all COW (backing) and filter links between @bs and @base.
+ * The caller must ensure that all links are frozen before using this
+ * function.
+ * @base must be reachable from @bs, or NULL.
  */
-BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
-                                    BlockDriverState *bs)
+void bdrv_unfreeze_backing_chain(BlockDriverState *bs, BlockDriverState *base)
 {
-    BlockDriverState *overlay = NULL;
-    BlockDriverState *intermediate;
-
-    assert(active != NULL);
-    assert(bs != NULL);
-
-    /* if bs is the same as active, then by definition it has no overlay
-     */
-    if (active == bs) {
-        return NULL;
-    }
+    BlockDriverState *i;
+    BdrvChild *child;
 
-    intermediate = active;
-    while (intermediate->backing_hd) {
-        if (intermediate->backing_hd == bs) {
-            overlay = intermediate;
-            break;
+    for (i = bs; i != base; i = child_bs(child)) {
+        child = bdrv_filter_or_cow_child(i);
+        if (child) {
+            assert(child->frozen);
+            child->frozen = false;
         }
-        intermediate = intermediate->backing_hd;
     }
-
-    return overlay;
 }
 
-typedef struct BlkIntermediateStates {
-    BlockDriverState *bs;
-    QSIMPLEQ_ENTRY(BlkIntermediateStates) entry;
-} BlkIntermediateStates;
-
-
 /*
  * Drops images above 'base' up to and including 'top', and sets the image
  * above 'top' to have base as its backing file.
@@ -2052,894 +4898,592 @@ typedef struct BlkIntermediateStates {
  *
  * base <- active
  *
+ * If backing_file_str is non-NULL, it will be used when modifying top's
+ * overlay image metadata.
+ *
  * Error conditions:
  *  if active == top, that is considered an error
  *
  */
-int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top,
-                           BlockDriverState *base)
+int bdrv_drop_intermediate(BlockDriverState *top, BlockDriverState *base,
+                           const char *backing_file_str)
 {
-    BlockDriverState *intermediate;
-    BlockDriverState *base_bs = NULL;
-    BlockDriverState *new_top_bs = NULL;
-    BlkIntermediateStates *intermediate_state, *next;
+    BlockDriverState *explicit_top = top;
+    bool update_inherits_from;
+    BdrvChild *c;
+    Error *local_err = NULL;
     int ret = -EIO;
+    g_autoptr(GSList) updated_children = NULL;
+    GSList *p;
 
-    QSIMPLEQ_HEAD(states_to_delete, BlkIntermediateStates) states_to_delete;
-    QSIMPLEQ_INIT(&states_to_delete);
+    bdrv_ref(top);
+    bdrv_subtree_drained_begin(top);
 
     if (!top->drv || !base->drv) {
         goto exit;
     }
 
-    new_top_bs = bdrv_find_overlay(active, top);
-
-    if (new_top_bs == NULL) {
-        /* we could not find the image above 'top', this is an error */
+    /* Make sure that base is in the backing chain of top */
+    if (!bdrv_chain_contains(top, base)) {
         goto exit;
     }
 
-    /* special case of new_top_bs->backing_hd already pointing to base - nothing
-     * to do, no intermediate images */
-    if (new_top_bs->backing_hd == base) {
-        ret = 0;
-        goto exit;
-    }
+    /* If 'base' recursively inherits from 'top' then we should set
+     * base->inherits_from to top->inherits_from after 'top' and all
+     * other intermediate nodes have been dropped.
+     * If 'top' is an implicit node (e.g. "commit_top") we should skip
+     * it because no one inherits from it. We use explicit_top for that. */
+    explicit_top = bdrv_skip_implicit_filters(explicit_top);
+    update_inherits_from = bdrv_inherits_from_recursive(base, explicit_top);
 
-    intermediate = top;
-
-    /* now we will go down through the list, and add each BDS we find
-     * into our deletion queue, until we hit the 'base'
-     */
-    while (intermediate) {
-        intermediate_state = g_malloc0(sizeof(BlkIntermediateStates));
-        intermediate_state->bs = intermediate;
-        QSIMPLEQ_INSERT_TAIL(&states_to_delete, intermediate_state, entry);
-
-        if (intermediate->backing_hd == base) {
-            base_bs = intermediate->backing_hd;
-            break;
-        }
-        intermediate = intermediate->backing_hd;
+    /* success - we can delete the intermediate states, and link top->base */
+    /* TODO Check graph modification op blockers (BLK_PERM_GRAPH_MOD) once
+     * we've figured out how they should work. */
+    if (!backing_file_str) {
+        bdrv_refresh_filename(base);
+        backing_file_str = base->filename;
     }
-    if (base_bs == NULL) {
-        /* something went wrong, we did not end at the base. safely
-         * unravel everything, and exit with error */
-        goto exit;
+
+    QLIST_FOREACH(c, &top->parents, next_parent) {
+        updated_children = g_slist_prepend(updated_children, c);
     }
 
-    /* success - we can delete the intermediate states, and link top->base */
-    ret = bdrv_change_backing_file(new_top_bs, base_bs->filename,
-                                   base_bs->drv ? base_bs->drv->format_name : "");
-    if (ret) {
+    bdrv_replace_node_common(top, base, false, &local_err);
+    if (local_err) {
+        error_report_err(local_err);
         goto exit;
     }
-    new_top_bs->backing_hd = base_bs;
 
+    for (p = updated_children; p; p = p->next) {
+        c = p->data;
 
-    QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) {
-        /* so that bdrv_close() does not recursively close the chain */
-        intermediate_state->bs->backing_hd = NULL;
-        bdrv_delete(intermediate_state->bs);
+        if (c->klass->update_filename) {
+            ret = c->klass->update_filename(c, base, backing_file_str,
+                                            &local_err);
+            if (ret < 0) {
+                /*
+                 * TODO: Actually, we want to rollback all previous iterations
+                 * of this loop, and (which is almost impossible) previous
+                 * bdrv_replace_node()...
+                 *
+                 * Note, that c->klass->update_filename may lead to permission
+                 * update, so it's a bad idea to call it inside permission
+                 * update transaction of bdrv_replace_node.
+                 */
+                error_report_err(local_err);
+                goto exit;
+            }
+        }
     }
-    ret = 0;
 
-exit:
-    QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) {
-        g_free(intermediate_state);
+    if (update_inherits_from) {
+        base->inherits_from = explicit_top->inherits_from;
     }
-    return ret;
-}
-
-
-static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
-                                   size_t size)
-{
-    int64_t len;
-
-    if (!bdrv_is_inserted(bs))
-        return -ENOMEDIUM;
 
-    if (bs->growable)
-        return 0;
-
-    len = bdrv_getlength(bs);
-
-    if (offset < 0)
-        return -EIO;
-
-    if ((offset > len) || (len - offset < size))
-        return -EIO;
-
-    return 0;
-}
-
-static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
-                              int nb_sectors)
-{
-    return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE,
-                                   nb_sectors * BDRV_SECTOR_SIZE);
+    ret = 0;
+exit:
+    bdrv_subtree_drained_end(top);
+    bdrv_unref(top);
+    return ret;
 }
 
-typedef struct RwCo {
-    BlockDriverState *bs;
-    int64_t sector_num;
-    int nb_sectors;
-    QEMUIOVector *qiov;
-    bool is_write;
-    int ret;
-    BdrvRequestFlags flags;
-} RwCo;
-
-static void coroutine_fn bdrv_rw_co_entry(void *opaque)
+/**
+ * Implementation of BlockDriver.bdrv_get_allocated_file_size() that
+ * sums the size of all data-bearing children.  (This excludes backing
+ * children.)
+ */
+static int64_t bdrv_sum_allocated_file_size(BlockDriverState *bs)
 {
-    RwCo *rwco = opaque;
+    BdrvChild *child;
+    int64_t child_size, sum = 0;
 
-    if (!rwco->is_write) {
-        rwco->ret = bdrv_co_do_readv(rwco->bs, rwco->sector_num,
-                                     rwco->nb_sectors, rwco->qiov,
-                                     rwco->flags);
-    } else {
-        rwco->ret = bdrv_co_do_writev(rwco->bs, rwco->sector_num,
-                                      rwco->nb_sectors, rwco->qiov,
-                                      rwco->flags);
+    QLIST_FOREACH(child, &bs->children, next) {
+        if (child->role & (BDRV_CHILD_DATA | BDRV_CHILD_METADATA |
+                           BDRV_CHILD_FILTERED))
+        {
+            child_size = bdrv_get_allocated_file_size(child->bs);
+            if (child_size < 0) {
+                return child_size;
+            }
+            sum += child_size;
+        }
     }
+
+    return sum;
 }
 
-/*
- * Process a vectored synchronous request using coroutines
+/**
+ * Length of a allocated file in bytes. Sparse files are counted by actual
+ * allocated space. Return < 0 if error or unknown.
  */
-static int bdrv_rwv_co(BlockDriverState *bs, int64_t sector_num,
-                       QEMUIOVector *qiov, bool is_write,
-                       BdrvRequestFlags flags)
+int64_t bdrv_get_allocated_file_size(BlockDriverState *bs)
 {
-    Coroutine *co;
-    RwCo rwco = {
-        .bs = bs,
-        .sector_num = sector_num,
-        .nb_sectors = qiov->size >> BDRV_SECTOR_BITS,
-        .qiov = qiov,
-        .is_write = is_write,
-        .ret = NOT_DONE,
-        .flags = flags,
-    };
-    assert((qiov->size & (BDRV_SECTOR_SIZE - 1)) == 0);
-
-    /**
-     * In sync call context, when the vcpu is blocked, this throttling timer
-     * will not fire; so the I/O throttling function has to be disabled here
-     * if it has been enabled.
-     */
-    if (bs->io_limits_enabled) {
-        fprintf(stderr, "Disabling I/O throttling on '%s' due "
-                        "to synchronous I/O.\n", bdrv_get_device_name(bs));
-        bdrv_io_limits_disable(bs);
+    BlockDriver *drv = bs->drv;
+    if (!drv) {
+        return -ENOMEDIUM;
+    }
+    if (drv->bdrv_get_allocated_file_size) {
+        return drv->bdrv_get_allocated_file_size(bs);
     }
 
-    if (qemu_in_coroutine()) {
-        /* Fast-path if already in coroutine context */
-        bdrv_rw_co_entry(&rwco);
+    if (drv->bdrv_file_open) {
+        /*
+         * Protocol drivers default to -ENOTSUP (most of their data is
+         * not stored in any of their children (if they even have any),
+         * so there is no generic way to figure it out).
+         */
+        return -ENOTSUP;
+    } else if (drv->is_filter) {
+        /* Filter drivers default to the size of their filtered child */
+        return bdrv_get_allocated_file_size(bdrv_filter_bs(bs));
     } else {
-        co = qemu_coroutine_create(bdrv_rw_co_entry);
-        qemu_coroutine_enter(co, &rwco);
-        while (rwco.ret == NOT_DONE) {
-            qemu_aio_wait();
-        }
+        /* Other drivers default to summing their children's sizes */
+        return bdrv_sum_allocated_file_size(bs);
     }
-    return rwco.ret;
 }
 
 /*
- * Process a synchronous request using coroutines
+ * bdrv_measure:
+ * @drv: Format driver
+ * @opts: Creation options for new image
+ * @in_bs: Existing image containing data for new image (may be NULL)
+ * @errp: Error object
+ * Returns: A #BlockMeasureInfo (free using qapi_free_BlockMeasureInfo())
+ *          or NULL on error
+ *
+ * Calculate file size required to create a new image.
+ *
+ * If @in_bs is given then space for allocated clusters and zero clusters
+ * from that image are included in the calculation.  If @opts contains a
+ * backing file that is shared by @in_bs then backing clusters may be omitted
+ * from the calculation.
+ *
+ * If @in_bs is NULL then the calculation includes no allocated clusters
+ * unless a preallocation option is given in @opts.
+ *
+ * Note that @in_bs may use a different BlockDriver from @drv.
+ *
+ * If an error occurs the @errp pointer is set.
  */
-static int bdrv_rw_co(BlockDriverState *bs, int64_t sector_num, uint8_t *buf,
-                      int nb_sectors, bool is_write, BdrvRequestFlags flags)
-{
-    QEMUIOVector qiov;
-    struct iovec iov = {
-        .iov_base = (void *)buf,
-        .iov_len = nb_sectors * BDRV_SECTOR_SIZE,
-    };
-
-    qemu_iovec_init_external(&qiov, &iov, 1);
-    return bdrv_rwv_co(bs, sector_num, &qiov, is_write, flags);
-}
-
-/* return < 0 if error. See bdrv_write() for the return codes */
-int bdrv_read(BlockDriverState *bs, int64_t sector_num,
-              uint8_t *buf, int nb_sectors)
-{
-    return bdrv_rw_co(bs, sector_num, buf, nb_sectors, false, 0);
-}
-
-/* Just like bdrv_read(), but with I/O throttling temporarily disabled */
-int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num,
-                          uint8_t *buf, int nb_sectors)
-{
-    bool enabled;
-    int ret;
-
-    enabled = bs->io_limits_enabled;
-    bs->io_limits_enabled = false;
-    ret = bdrv_read(bs, sector_num, buf, nb_sectors);
-    bs->io_limits_enabled = enabled;
-    return ret;
-}
-
-/* Return < 0 if error. Important errors are:
-  -EIO         generic I/O error (may happen for all errors)
-  -ENOMEDIUM   No media inserted.
-  -EINVAL      Invalid sector number or nb_sectors
-  -EACCES      Trying to write a read-only device
-*/
-int bdrv_write(BlockDriverState *bs, int64_t sector_num,
-               const uint8_t *buf, int nb_sectors)
-{
-    return bdrv_rw_co(bs, sector_num, (uint8_t *)buf, nb_sectors, true, 0);
-}
-
-int bdrv_writev(BlockDriverState *bs, int64_t sector_num, QEMUIOVector *qiov)
+BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts,
+                               BlockDriverState *in_bs, Error **errp)
 {
-    return bdrv_rwv_co(bs, sector_num, qiov, true, 0);
-}
-
-int bdrv_write_zeroes(BlockDriverState *bs, int64_t sector_num, int nb_sectors)
-{
-    return bdrv_rw_co(bs, sector_num, NULL, nb_sectors, true,
-                      BDRV_REQ_ZERO_WRITE);
-}
-
-int bdrv_pread(BlockDriverState *bs, int64_t offset,
-               void *buf, int count1)
-{
-    uint8_t tmp_buf[BDRV_SECTOR_SIZE];
-    int len, nb_sectors, count;
-    int64_t sector_num;
-    int ret;
-
-    count = count1;
-    /* first read to align to sector start */
-    len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
-    if (len > count)
-        len = count;
-    sector_num = offset >> BDRV_SECTOR_BITS;
-    if (len > 0) {
-        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
-            return ret;
-        memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
-        count -= len;
-        if (count == 0)
-            return count1;
-        sector_num++;
-        buf += len;
-    }
-
-    /* read the sectors "in place" */
-    nb_sectors = count >> BDRV_SECTOR_BITS;
-    if (nb_sectors > 0) {
-        if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0)
-            return ret;
-        sector_num += nb_sectors;
-        len = nb_sectors << BDRV_SECTOR_BITS;
-        buf += len;
-        count -= len;
+    if (!drv->bdrv_measure) {
+        error_setg(errp, "Block driver '%s' does not support size measurement",
+                   drv->format_name);
+        return NULL;
     }
 
-    /* add data from the last sector */
-    if (count > 0) {
-        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
-            return ret;
-        memcpy(buf, tmp_buf, count);
-    }
-    return count1;
+    return drv->bdrv_measure(opts, in_bs, errp);
 }
 
-int bdrv_pwritev(BlockDriverState *bs, int64_t offset, QEMUIOVector *qiov)
+/**
+ * Return number of sectors on success, -errno on error.
+ */
+int64_t bdrv_nb_sectors(BlockDriverState *bs)
 {
-    uint8_t tmp_buf[BDRV_SECTOR_SIZE];
-    int len, nb_sectors, count;
-    int64_t sector_num;
-    int ret;
+    BlockDriver *drv = bs->drv;
 
-    count = qiov->size;
+    if (!drv)
+        return -ENOMEDIUM;
 
-    /* first write to align to sector start */
-    len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
-    if (len > count)
-        len = count;
-    sector_num = offset >> BDRV_SECTOR_BITS;
-    if (len > 0) {
-        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
-            return ret;
-        qemu_iovec_to_buf(qiov, 0, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)),
-                          len);
-        if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
-            return ret;
-        count -= len;
-        if (count == 0)
-            return qiov->size;
-        sector_num++;
-    }
-
-    /* write the sectors "in place" */
-    nb_sectors = count >> BDRV_SECTOR_BITS;
-    if (nb_sectors > 0) {
-        QEMUIOVector qiov_inplace;
-
-        qemu_iovec_init(&qiov_inplace, qiov->niov);
-        qemu_iovec_concat(&qiov_inplace, qiov, len,
-                          nb_sectors << BDRV_SECTOR_BITS);
-        ret = bdrv_writev(bs, sector_num, &qiov_inplace);
-        qemu_iovec_destroy(&qiov_inplace);
+    if (drv->has_variable_length) {
+        int ret = refresh_total_sectors(bs, bs->total_sectors);
         if (ret < 0) {
             return ret;
         }
-
-        sector_num += nb_sectors;
-        len = nb_sectors << BDRV_SECTOR_BITS;
-        count -= len;
     }
-
-    /* add data from the last sector */
-    if (count > 0) {
-        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
-            return ret;
-        qemu_iovec_to_buf(qiov, qiov->size - count, tmp_buf, count);
-        if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
-            return ret;
-    }
-    return qiov->size;
+    return bs->total_sectors;
 }
 
-int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
-                const void *buf, int count1)
-{
-    QEMUIOVector qiov;
-    struct iovec iov = {
-        .iov_base   = (void *) buf,
-        .iov_len    = count1,
-    };
-
-    qemu_iovec_init_external(&qiov, &iov, 1);
-    return bdrv_pwritev(bs, offset, &qiov);
-}
-
-/*
- * Writes to the file and ensures that no writes are reordered across this
- * request (acts as a barrier)
- *
- * Returns 0 on success, -errno in error cases.
+/**
+ * Return length in bytes on success, -errno on error.
+ * The length is always a multiple of BDRV_SECTOR_SIZE.
  */
-int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset,
-    const void *buf, int count)
+int64_t bdrv_getlength(BlockDriverState *bs)
 {
-    int ret;
+    int64_t ret = bdrv_nb_sectors(bs);
 
-    ret = bdrv_pwrite(bs, offset, buf, count);
     if (ret < 0) {
         return ret;
     }
-
-    /* No flush needed for cache modes that already do it */
-    if (bs->enable_write_cache) {
-        bdrv_flush(bs);
+    if (ret > INT64_MAX / BDRV_SECTOR_SIZE) {
+        return -EFBIG;
     }
-
-    return 0;
+    return ret * BDRV_SECTOR_SIZE;
 }
 
-static int coroutine_fn bdrv_co_do_copy_on_readv(BlockDriverState *bs,
-        int64_t sector_num, int nb_sectors, QEMUIOVector *qiov)
+/* return 0 as number of sectors if no device present or error */
+void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
 {
-    /* Perform I/O through a temporary buffer so that users who scribble over
-     * their read buffer while the operation is in progress do not end up
-     * modifying the image file.  This is critical for zero-copy guest I/O
-     * where anything might happen inside guest memory.
-     */
-    void *bounce_buffer;
-
-    BlockDriver *drv = bs->drv;
-    struct iovec iov;
-    QEMUIOVector bounce_qiov;
-    int64_t cluster_sector_num;
-    int cluster_nb_sectors;
-    size_t skip_bytes;
-    int ret;
+    int64_t nb_sectors = bdrv_nb_sectors(bs);
 
-    /* Cover entire cluster so no additional backing file I/O is required when
-     * allocating cluster in the image file.
-     */
-    bdrv_round_to_clusters(bs, sector_num, nb_sectors,
-                           &cluster_sector_num, &cluster_nb_sectors);
+    *nb_sectors_ptr = nb_sectors < 0 ? 0 : nb_sectors;
+}
 
-    trace_bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors,
-                                   cluster_sector_num, cluster_nb_sectors);
+bool bdrv_is_sg(BlockDriverState *bs)
+{
+    return bs->sg;
+}
 
-    iov.iov_len = cluster_nb_sectors * BDRV_SECTOR_SIZE;
-    iov.iov_base = bounce_buffer = qemu_blockalign(bs, iov.iov_len);
-    qemu_iovec_init_external(&bounce_qiov, &iov, 1);
+/**
+ * Return whether the given node supports compressed writes.
+ */
+bool bdrv_supports_compressed_writes(BlockDriverState *bs)
+{
+    BlockDriverState *filtered;
 
-    ret = drv->bdrv_co_readv(bs, cluster_sector_num, cluster_nb_sectors,
-                             &bounce_qiov);
-    if (ret < 0) {
-        goto err;
+    if (!bs->drv || !block_driver_can_compress(bs->drv)) {
+        return false;
     }
 
-    if (drv->bdrv_co_write_zeroes &&
-        buffer_is_zero(bounce_buffer, iov.iov_len)) {
-        ret = bdrv_co_do_write_zeroes(bs, cluster_sector_num,
-                                      cluster_nb_sectors);
-    } else {
-        /* This does not change the data on the disk, it is not necessary
-         * to flush even in cache=writethrough mode.
+    filtered = bdrv_filter_bs(bs);
+    if (filtered) {
+        /*
+         * Filters can only forward compressed writes, so we have to
+         * check the child.
          */
-        ret = drv->bdrv_co_writev(bs, cluster_sector_num, cluster_nb_sectors,
-                                  &bounce_qiov);
+        return bdrv_supports_compressed_writes(filtered);
     }
 
-    if (ret < 0) {
-        /* It might be okay to ignore write errors for guest requests.  If this
-         * is a deliberate copy-on-read then we don't want to ignore the error.
-         * Simply report it in all cases.
-         */
-        goto err;
-    }
+    return true;
+}
 
-    skip_bytes = (sector_num - cluster_sector_num) * BDRV_SECTOR_SIZE;
-    qemu_iovec_from_buf(qiov, 0, bounce_buffer + skip_bytes,
-                        nb_sectors * BDRV_SECTOR_SIZE);
+const char *bdrv_get_format_name(BlockDriverState *bs)
+{
+    return bs->drv ? bs->drv->format_name : NULL;
+}
 
-err:
-    qemu_vfree(bounce_buffer);
-    return ret;
+static int qsort_strcmp(const void *a, const void *b)
+{
+    return strcmp(*(char *const *)a, *(char *const *)b);
 }
 
-/*
- * Handle a read request in coroutine context
- */
-static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs,
-    int64_t sector_num, int nb_sectors, QEMUIOVector *qiov,
-    BdrvRequestFlags flags)
+void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
+                         void *opaque, bool read_only)
 {
-    BlockDriver *drv = bs->drv;
-    BdrvTrackedRequest req;
-    int ret;
+    BlockDriver *drv;
+    int count = 0;
+    int i;
+    const char **formats = NULL;
 
-    if (!drv) {
-        return -ENOMEDIUM;
-    }
-    if (bdrv_check_request(bs, sector_num, nb_sectors)) {
-        return -EIO;
-    }
+    QLIST_FOREACH(drv, &bdrv_drivers, list) {
+        if (drv->format_name) {
+            bool found = false;
+            int i = count;
 
-    /* throttling disk read I/O */
-    if (bs->io_limits_enabled) {
-        bdrv_io_limits_intercept(bs, false, nb_sectors);
-    }
+            if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv, read_only)) {
+                continue;
+            }
 
-    if (bs->copy_on_read) {
-        flags |= BDRV_REQ_COPY_ON_READ;
-    }
-    if (flags & BDRV_REQ_COPY_ON_READ) {
-        bs->copy_on_read_in_flight++;
-    }
+            while (formats && i && !found) {
+                found = !strcmp(formats[--i], drv->format_name);
+            }
 
-    if (bs->copy_on_read_in_flight) {
-        wait_for_overlapping_requests(bs, sector_num, nb_sectors);
+            if (!found) {
+                formats = g_renew(const char *, formats, count + 1);
+                formats[count++] = drv->format_name;
+            }
+        }
     }
 
-    tracked_request_begin(&req, bs, sector_num, nb_sectors, false);
+    for (i = 0; i < (int)ARRAY_SIZE(block_driver_modules); i++) {
+        const char *format_name = block_driver_modules[i].format_name;
 
-    if (flags & BDRV_REQ_COPY_ON_READ) {
-        int pnum;
+        if (format_name) {
+            bool found = false;
+            int j = count;
 
-        ret = bdrv_co_is_allocated(bs, sector_num, nb_sectors, &pnum);
-        if (ret < 0) {
-            goto out;
-        }
+            if (use_bdrv_whitelist &&
+                !bdrv_format_is_whitelisted(format_name, read_only)) {
+                continue;
+            }
 
-        if (!ret || pnum != nb_sectors) {
-            ret = bdrv_co_do_copy_on_readv(bs, sector_num, nb_sectors, qiov);
-            goto out;
+            while (formats && j && !found) {
+                found = !strcmp(formats[--j], format_name);
+            }
+
+            if (!found) {
+                formats = g_renew(const char *, formats, count + 1);
+                formats[count++] = format_name;
+            }
         }
     }
 
-    ret = drv->bdrv_co_readv(bs, sector_num, nb_sectors, qiov);
-
-out:
-    tracked_request_end(&req);
+    qsort(formats, count, sizeof(formats[0]), qsort_strcmp);
 
-    if (flags & BDRV_REQ_COPY_ON_READ) {
-        bs->copy_on_read_in_flight--;
+    for (i = 0; i < count; i++) {
+        it(opaque, formats[i]);
     }
 
-    return ret;
+    g_free(formats);
 }
 
-int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num,
-    int nb_sectors, QEMUIOVector *qiov)
+/* This function is to find a node in the bs graph */
+BlockDriverState *bdrv_find_node(const char *node_name)
 {
-    trace_bdrv_co_readv(bs, sector_num, nb_sectors);
-
-    return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov, 0);
-}
+    BlockDriverState *bs;
 
-int coroutine_fn bdrv_co_copy_on_readv(BlockDriverState *bs,
-    int64_t sector_num, int nb_sectors, QEMUIOVector *qiov)
-{
-    trace_bdrv_co_copy_on_readv(bs, sector_num, nb_sectors);
+    assert(node_name);
 
-    return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov,
-                            BDRV_REQ_COPY_ON_READ);
+    QTAILQ_FOREACH(bs, &graph_bdrv_states, node_list) {
+        if (!strcmp(node_name, bs->node_name)) {
+            return bs;
+        }
+    }
+    return NULL;
 }
 
-static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs,
-    int64_t sector_num, int nb_sectors)
+/* Put this QMP function here so it can access the static graph_bdrv_states. */
+BlockDeviceInfoList *bdrv_named_nodes_list(bool flat,
+                                           Error **errp)
 {
-    BlockDriver *drv = bs->drv;
-    QEMUIOVector qiov;
-    struct iovec iov;
-    int ret;
-
-    /* TODO Emulate only part of misaligned requests instead of letting block
-     * drivers return -ENOTSUP and emulate everything */
+    BlockDeviceInfoList *list;
+    BlockDriverState *bs;
 
-    /* First try the efficient write zeroes operation */
-    if (drv->bdrv_co_write_zeroes) {
-        ret = drv->bdrv_co_write_zeroes(bs, sector_num, nb_sectors);
-        if (ret != -ENOTSUP) {
-            return ret;
+    list = NULL;
+    QTAILQ_FOREACH(bs, &graph_bdrv_states, node_list) {
+        BlockDeviceInfo *info = bdrv_block_device_info(NULL, bs, flat, errp);
+        if (!info) {
+            qapi_free_BlockDeviceInfoList(list);
+            return NULL;
         }
+        QAPI_LIST_PREPEND(list, info);
     }
 
-    /* Fall back to bounce buffer if write zeroes is unsupported */
-    iov.iov_len  = nb_sectors * BDRV_SECTOR_SIZE;
-    iov.iov_base = qemu_blockalign(bs, iov.iov_len);
-    memset(iov.iov_base, 0, iov.iov_len);
-    qemu_iovec_init_external(&qiov, &iov, 1);
-
-    ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, &qiov);
-
-    qemu_vfree(iov.iov_base);
-    return ret;
+    return list;
 }
 
-/*
- * Handle a write request in coroutine context
- */
-static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs,
-    int64_t sector_num, int nb_sectors, QEMUIOVector *qiov,
-    BdrvRequestFlags flags)
-{
-    BlockDriver *drv = bs->drv;
-    BdrvTrackedRequest req;
-    int ret;
-
-    if (!bs->drv) {
-        return -ENOMEDIUM;
-    }
-    if (bs->read_only) {
-        return -EACCES;
-    }
-    if (bdrv_check_request(bs, sector_num, nb_sectors)) {
-        return -EIO;
-    }
+typedef struct XDbgBlockGraphConstructor {
+    XDbgBlockGraph *graph;
+    GHashTable *graph_nodes;
+} XDbgBlockGraphConstructor;
 
-    /* throttling disk write I/O */
-    if (bs->io_limits_enabled) {
-        bdrv_io_limits_intercept(bs, true, nb_sectors);
-    }
+static XDbgBlockGraphConstructor *xdbg_graph_new(void)
+{
+    XDbgBlockGraphConstructor *gr = g_new(XDbgBlockGraphConstructor, 1);
 
-    if (bs->copy_on_read_in_flight) {
-        wait_for_overlapping_requests(bs, sector_num, nb_sectors);
-    }
+    gr->graph = g_new0(XDbgBlockGraph, 1);
+    gr->graph_nodes = g_hash_table_new(NULL, NULL);
 
-    tracked_request_begin(&req, bs, sector_num, nb_sectors, true);
+    return gr;
+}
 
-    ret = notifier_with_return_list_notify(&bs->before_write_notifiers, &req);
+static XDbgBlockGraph *xdbg_graph_finalize(XDbgBlockGraphConstructor *gr)
+{
+    XDbgBlockGraph *graph = gr->graph;
 
-    if (ret < 0) {
-        /* Do nothing, write notifier decided to fail this request */
-    } else if (flags & BDRV_REQ_ZERO_WRITE) {
-        ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors);
-    } else {
-        ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov);
-    }
+    g_hash_table_destroy(gr->graph_nodes);
+    g_free(gr);
 
-    if (ret == 0 && !bs->enable_write_cache) {
-        ret = bdrv_co_flush(bs);
-    }
+    return graph;
+}
 
-    if (bs->dirty_bitmap) {
-        bdrv_set_dirty(bs, sector_num, nb_sectors);
-    }
+static uintptr_t xdbg_graph_node_num(XDbgBlockGraphConstructor *gr, void *node)
+{
+    uintptr_t ret = (uintptr_t)g_hash_table_lookup(gr->graph_nodes, node);
 
-    if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
-        bs->wr_highest_sector = sector_num + nb_sectors - 1;
+    if (ret != 0) {
+        return ret;
     }
 
-    tracked_request_end(&req);
+    /*
+     * Start counting from 1, not 0, because 0 interferes with not-found (NULL)
+     * answer of g_hash_table_lookup.
+     */
+    ret = g_hash_table_size(gr->graph_nodes) + 1;
+    g_hash_table_insert(gr->graph_nodes, node, (void *)ret);
 
     return ret;
 }
 
-int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num,
-    int nb_sectors, QEMUIOVector *qiov)
+static void xdbg_graph_add_node(XDbgBlockGraphConstructor *gr, void *node,
+                                XDbgBlockGraphNodeType type, const char *name)
 {
-    trace_bdrv_co_writev(bs, sector_num, nb_sectors);
+    XDbgBlockGraphNode *n;
 
-    return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, 0);
-}
+    n = g_new0(XDbgBlockGraphNode, 1);
 
-int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs,
-                                      int64_t sector_num, int nb_sectors)
-{
-    trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors);
+    n->id = xdbg_graph_node_num(gr, node);
+    n->type = type;
+    n->name = g_strdup(name);
 
-    return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL,
-                             BDRV_REQ_ZERO_WRITE);
+    QAPI_LIST_PREPEND(gr->graph->nodes, n);
 }
 
-/**
- * Truncate file to 'offset' bytes (needed only for file protocols)
- */
-int bdrv_truncate(BlockDriverState *bs, int64_t offset)
+static void xdbg_graph_add_edge(XDbgBlockGraphConstructor *gr, void *parent,
+                                const BdrvChild *child)
 {
-    BlockDriver *drv = bs->drv;
-    int ret;
-    if (!drv)
-        return -ENOMEDIUM;
-    if (!drv->bdrv_truncate)
-        return -ENOTSUP;
-    if (bs->read_only)
-        return -EACCES;
-    if (bdrv_in_use(bs))
-        return -EBUSY;
-    ret = drv->bdrv_truncate(bs, offset);
-    if (ret == 0) {
-        ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
-        bdrv_dev_resize_cb(bs);
-    }
-    return ret;
-}
+    BlockPermission qapi_perm;
+    XDbgBlockGraphEdge *edge;
 
-/**
- * Length of a allocated file in bytes. Sparse files are counted by actual
- * allocated space. Return < 0 if error or unknown.
- */
-int64_t bdrv_get_allocated_file_size(BlockDriverState *bs)
-{
-    BlockDriver *drv = bs->drv;
-    if (!drv) {
-        return -ENOMEDIUM;
-    }
-    if (drv->bdrv_get_allocated_file_size) {
-        return drv->bdrv_get_allocated_file_size(bs);
-    }
-    if (bs->file) {
-        return bdrv_get_allocated_file_size(bs->file);
-    }
-    return -ENOTSUP;
-}
+    edge = g_new0(XDbgBlockGraphEdge, 1);
 
-/**
- * Length of a file in bytes. Return < 0 if error or unknown.
- */
-int64_t bdrv_getlength(BlockDriverState *bs)
-{
-    BlockDriver *drv = bs->drv;
-    if (!drv)
-        return -ENOMEDIUM;
+    edge->parent = xdbg_graph_node_num(gr, parent);
+    edge->child = xdbg_graph_node_num(gr, child->bs);
+    edge->name = g_strdup(child->name);
 
-    if (bs->growable || bdrv_dev_has_removable_media(bs)) {
-        if (drv->bdrv_getlength) {
-            return drv->bdrv_getlength(bs);
+    for (qapi_perm = 0; qapi_perm < BLOCK_PERMISSION__MAX; qapi_perm++) {
+        uint64_t flag = bdrv_qapi_perm_to_blk_perm(qapi_perm);
+
+        if (flag & child->perm) {
+            QAPI_LIST_PREPEND(edge->perm, qapi_perm);
+        }
+        if (flag & child->shared_perm) {
+            QAPI_LIST_PREPEND(edge->shared_perm, qapi_perm);
         }
     }
-    return bs->total_sectors * BDRV_SECTOR_SIZE;
-}
 
-/* return 0 as number of sectors if no device present or error */
-void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
-{
-    int64_t length;
-    length = bdrv_getlength(bs);
-    if (length < 0)
-        length = 0;
-    else
-        length = length >> BDRV_SECTOR_BITS;
-    *nb_sectors_ptr = length;
+    QAPI_LIST_PREPEND(gr->graph->edges, edge);
 }
 
-/* throttling disk io limits */
-void bdrv_set_io_limits(BlockDriverState *bs,
-                        BlockIOLimit *io_limits)
-{
-    bs->io_limits = *io_limits;
-    bs->io_limits_enabled = bdrv_io_limits_enabled(bs);
-}
 
-void bdrv_set_on_error(BlockDriverState *bs, BlockdevOnError on_read_error,
-                       BlockdevOnError on_write_error)
+XDbgBlockGraph *bdrv_get_xdbg_block_graph(Error **errp)
 {
-    bs->on_read_error = on_read_error;
-    bs->on_write_error = on_write_error;
-}
+    BlockBackend *blk;
+    BlockJob *job;
+    BlockDriverState *bs;
+    BdrvChild *child;
+    XDbgBlockGraphConstructor *gr = xdbg_graph_new();
 
-BlockdevOnError bdrv_get_on_error(BlockDriverState *bs, bool is_read)
-{
-    return is_read ? bs->on_read_error : bs->on_write_error;
-}
+    for (blk = blk_all_next(NULL); blk; blk = blk_all_next(blk)) {
+        char *allocated_name = NULL;
+        const char *name = blk_name(blk);
 
-BlockErrorAction bdrv_get_error_action(BlockDriverState *bs, bool is_read, int error)
-{
-    BlockdevOnError on_err = is_read ? bs->on_read_error : bs->on_write_error;
+        if (!*name) {
+            name = allocated_name = blk_get_attached_dev_id(blk);
+        }
+        xdbg_graph_add_node(gr, blk, X_DBG_BLOCK_GRAPH_NODE_TYPE_BLOCK_BACKEND,
+                           name);
+        g_free(allocated_name);
+        if (blk_root(blk)) {
+            xdbg_graph_add_edge(gr, blk, blk_root(blk));
+        }
+    }
 
-    switch (on_err) {
-    case BLOCKDEV_ON_ERROR_ENOSPC:
-        return (error == ENOSPC) ? BDRV_ACTION_STOP : BDRV_ACTION_REPORT;
-    case BLOCKDEV_ON_ERROR_STOP:
-        return BDRV_ACTION_STOP;
-    case BLOCKDEV_ON_ERROR_REPORT:
-        return BDRV_ACTION_REPORT;
-    case BLOCKDEV_ON_ERROR_IGNORE:
-        return BDRV_ACTION_IGNORE;
-    default:
-        abort();
+    for (job = block_job_next(NULL); job; job = block_job_next(job)) {
+        GSList *el;
+
+        xdbg_graph_add_node(gr, job, X_DBG_BLOCK_GRAPH_NODE_TYPE_BLOCK_JOB,
+                           job->job.id);
+        for (el = job->nodes; el; el = el->next) {
+            xdbg_graph_add_edge(gr, job, (BdrvChild *)el->data);
+        }
     }
-}
 
-/* This is done by device models because, while the block layer knows
- * about the error, it does not know whether an operation comes from
- * the device or the block layer (from a job, for example).
- */
-void bdrv_error_action(BlockDriverState *bs, BlockErrorAction action,
-                       bool is_read, int error)
-{
-    assert(error >= 0);
-    bdrv_emit_qmp_error_event(bs, QEVENT_BLOCK_IO_ERROR, action, is_read);
-    if (action == BDRV_ACTION_STOP) {
-        vm_stop(RUN_STATE_IO_ERROR);
-        bdrv_iostatus_set_err(bs, error);
+    QTAILQ_FOREACH(bs, &graph_bdrv_states, node_list) {
+        xdbg_graph_add_node(gr, bs, X_DBG_BLOCK_GRAPH_NODE_TYPE_BLOCK_DRIVER,
+                           bs->node_name);
+        QLIST_FOREACH(child, &bs->children, next) {
+            xdbg_graph_add_edge(gr, bs, child);
+        }
     }
-}
 
-int bdrv_is_read_only(BlockDriverState *bs)
-{
-    return bs->read_only;
+    return xdbg_graph_finalize(gr);
 }
 
-int bdrv_is_sg(BlockDriverState *bs)
+BlockDriverState *bdrv_lookup_bs(const char *device,
+                                 const char *node_name,
+                                 Error **errp)
 {
-    return bs->sg;
-}
+    BlockBackend *blk;
+    BlockDriverState *bs;
 
-int bdrv_enable_write_cache(BlockDriverState *bs)
-{
-    return bs->enable_write_cache;
-}
+    if (device) {
+        blk = blk_by_name(device);
 
-void bdrv_set_enable_write_cache(BlockDriverState *bs, bool wce)
-{
-    bs->enable_write_cache = wce;
+        if (blk) {
+            bs = blk_bs(blk);
+            if (!bs) {
+                error_setg(errp, "Device '%s' has no medium", device);
+            }
 
-    /* so a reopen() will preserve wce */
-    if (wce) {
-        bs->open_flags |= BDRV_O_CACHE_WB;
-    } else {
-        bs->open_flags &= ~BDRV_O_CACHE_WB;
+            return bs;
+        }
     }
-}
 
-int bdrv_is_encrypted(BlockDriverState *bs)
-{
-    if (bs->backing_hd && bs->backing_hd->encrypted)
-        return 1;
-    return bs->encrypted;
-}
+    if (node_name) {
+        bs = bdrv_find_node(node_name);
 
-int bdrv_key_required(BlockDriverState *bs)
-{
-    BlockDriverState *backing_hd = bs->backing_hd;
+        if (bs) {
+            return bs;
+        }
+    }
 
-    if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
-        return 1;
-    return (bs->encrypted && !bs->valid_key);
+    error_setg(errp, "Cannot find device=%s nor node_name=%s",
+                     device ? device : "",
+                     node_name ? node_name : "");
+    return NULL;
 }
 
-int bdrv_set_key(BlockDriverState *bs, const char *key)
+/* If 'base' is in the same chain as 'top', return true. Otherwise,
+ * return false.  If either argument is NULL, return false. */
+bool bdrv_chain_contains(BlockDriverState *top, BlockDriverState *base)
 {
-    int ret;
-    if (bs->backing_hd && bs->backing_hd->encrypted) {
-        ret = bdrv_set_key(bs->backing_hd, key);
-        if (ret < 0)
-            return ret;
-        if (!bs->encrypted)
-            return 0;
-    }
-    if (!bs->encrypted) {
-        return -EINVAL;
-    } else if (!bs->drv || !bs->drv->bdrv_set_key) {
-        return -ENOMEDIUM;
+    while (top && top != base) {
+        top = bdrv_filter_or_cow_bs(top);
     }
-    ret = bs->drv->bdrv_set_key(bs, key);
-    if (ret < 0) {
-        bs->valid_key = 0;
-    } else if (!bs->valid_key) {
-        bs->valid_key = 1;
-        /* call the change callback now, we skipped it on open */
-        bdrv_dev_change_media_cb(bs, true);
-    }
-    return ret;
-}
 
-const char *bdrv_get_format_name(BlockDriverState *bs)
-{
-    return bs->drv ? bs->drv->format_name : NULL;
+    return top != NULL;
 }
 
-void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
-                         void *opaque)
+BlockDriverState *bdrv_next_node(BlockDriverState *bs)
 {
-    BlockDriver *drv;
-
-    QLIST_FOREACH(drv, &bdrv_drivers, list) {
-        it(opaque, drv->format_name);
+    if (!bs) {
+        return QTAILQ_FIRST(&graph_bdrv_states);
     }
+    return QTAILQ_NEXT(bs, node_list);
 }
 
-BlockDriverState *bdrv_find(const char *name)
+BlockDriverState *bdrv_next_all_states(BlockDriverState *bs)
 {
-    BlockDriverState *bs;
-
-    QTAILQ_FOREACH(bs, &bdrv_states, list) {
-        if (!strcmp(name, bs->device_name)) {
-            return bs;
-        }
+    if (!bs) {
+        return QTAILQ_FIRST(&all_bdrv_states);
     }
-    return NULL;
+    return QTAILQ_NEXT(bs, bs_list);
 }
 
-BlockDriverState *bdrv_next(BlockDriverState *bs)
+const char *bdrv_get_node_name(const BlockDriverState *bs)
 {
-    if (!bs) {
-        return QTAILQ_FIRST(&bdrv_states);
-    }
-    return QTAILQ_NEXT(bs, list);
+    return bs->node_name;
 }
 
-void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
+const char *bdrv_get_parent_name(const BlockDriverState *bs)
 {
-    BlockDriverState *bs;
+    BdrvChild *c;
+    const char *name;
 
-    QTAILQ_FOREACH(bs, &bdrv_states, list) {
-        it(opaque, bs);
+    /* If multiple parents have a name, just pick the first one. */
+    QLIST_FOREACH(c, &bs->parents, next_parent) {
+        if (c->klass->get_name) {
+            name = c->klass->get_name(c);
+            if (name && *name) {
+                return name;
+            }
+        }
     }
+
+    return NULL;
 }
 
-const char *bdrv_get_device_name(BlockDriverState *bs)
+/* TODO check what callers really want: bs->node_name or blk_name() */
+const char *bdrv_get_device_name(const BlockDriverState *bs)
 {
-    return bs->device_name;
+    return bdrv_get_parent_name(bs) ?: "";
 }
 
-int bdrv_get_flags(BlockDriverState *bs)
+/* This can be used to identify nodes that might not have a device
+ * name associated. Since node and device names live in the same
+ * namespace, the result is unambiguous. The exception is if both are
+ * absent, then this returns an empty (non-null) string. */
+const char *bdrv_get_device_or_node_name(const BlockDriverState *bs)
 {
-    return bs->open_flags;
+    return bdrv_get_parent_name(bs) ?: bs->node_name;
 }
 
-int bdrv_flush_all(void)
+int bdrv_get_flags(BlockDriverState *bs)
 {
-    BlockDriverState *bs;
-    int result = 0;
-
-    QTAILQ_FOREACH(bs, &bdrv_states, list) {
-        int ret = bdrv_flush(bs);
-        if (ret < 0 && !result) {
-            result = ret;
-        }
-    }
-
-    return result;
+    return bs->open_flags;
 }
 
 int bdrv_has_zero_init_1(BlockDriverState *bs)
@@ -2949,197 +5493,37 @@ int bdrv_has_zero_init_1(BlockDriverState *bs)
 
 int bdrv_has_zero_init(BlockDriverState *bs)
 {
-    assert(bs->drv);
-
-    if (bs->drv->bdrv_has_zero_init) {
-        return bs->drv->bdrv_has_zero_init(bs);
-    }
-
-    /* safe default */
-    return 0;
-}
-
-typedef struct BdrvCoIsAllocatedData {
-    BlockDriverState *bs;
-    BlockDriverState *base;
-    int64_t sector_num;
-    int nb_sectors;
-    int *pnum;
-    int ret;
-    bool done;
-} BdrvCoIsAllocatedData;
+    BlockDriverState *filtered;
 
-/*
- * Returns true iff the specified sector is present in the disk image. Drivers
- * not implementing the functionality are assumed to not support backing files,
- * hence all their sectors are reported as allocated.
- *
- * If 'sector_num' is beyond the end of the disk image the return value is 0
- * and 'pnum' is set to 0.
- *
- * 'pnum' is set to the number of sectors (including and immediately following
- * the specified sector) that are known to be in the same
- * allocated/unallocated state.
- *
- * 'nb_sectors' is the max value 'pnum' should be set to.  If nb_sectors goes
- * beyond the end of the disk image it will be clamped.
- */
-int coroutine_fn bdrv_co_is_allocated(BlockDriverState *bs, int64_t sector_num,
-                                      int nb_sectors, int *pnum)
-{
-    int64_t n;
-
-    if (sector_num >= bs->total_sectors) {
-        *pnum = 0;
+    if (!bs->drv) {
         return 0;
     }
 
-    n = bs->total_sectors - sector_num;
-    if (n < nb_sectors) {
-        nb_sectors = n;
-    }
-
-    if (!bs->drv->bdrv_co_is_allocated) {
-        *pnum = nb_sectors;
-        return 1;
+    /* If BS is a copy on write image, it is initialized to
+       the contents of the base image, which may not be zeroes.  */
+    if (bdrv_cow_child(bs)) {
+        return 0;
     }
-
-    return bs->drv->bdrv_co_is_allocated(bs, sector_num, nb_sectors, pnum);
-}
-
-/* Coroutine wrapper for bdrv_is_allocated() */
-static void coroutine_fn bdrv_is_allocated_co_entry(void *opaque)
-{
-    BdrvCoIsAllocatedData *data = opaque;
-    BlockDriverState *bs = data->bs;
-
-    data->ret = bdrv_co_is_allocated(bs, data->sector_num, data->nb_sectors,
-                                     data->pnum);
-    data->done = true;
-}
-
-/*
- * Synchronous wrapper around bdrv_co_is_allocated().
- *
- * See bdrv_co_is_allocated() for details.
- */
-int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
-                      int *pnum)
-{
-    Coroutine *co;
-    BdrvCoIsAllocatedData data = {
-        .bs = bs,
-        .sector_num = sector_num,
-        .nb_sectors = nb_sectors,
-        .pnum = pnum,
-        .done = false,
-    };
-
-    co = qemu_coroutine_create(bdrv_is_allocated_co_entry);
-    qemu_coroutine_enter(co, &data);
-    while (!data.done) {
-        qemu_aio_wait();
+    if (bs->drv->bdrv_has_zero_init) {
+        return bs->drv->bdrv_has_zero_init(bs);
     }
-    return data.ret;
-}
 
-/*
- * Given an image chain: ... -> [BASE] -> [INTER1] -> [INTER2] -> [TOP]
- *
- * Return true if the given sector is allocated in any image between
- * BASE and TOP (inclusive).  BASE can be NULL to check if the given
- * sector is allocated in any image of the chain.  Return false otherwise.
- *
- * 'pnum' is set to the number of sectors (including and immediately following
- *  the specified sector) that are known to be in the same
- *  allocated/unallocated state.
- *
- */
-int coroutine_fn bdrv_co_is_allocated_above(BlockDriverState *top,
-                                            BlockDriverState *base,
-                                            int64_t sector_num,
-                                            int nb_sectors, int *pnum)
-{
-    BlockDriverState *intermediate;
-    int ret, n = nb_sectors;
-
-    intermediate = top;
-    while (intermediate && intermediate != base) {
-        int pnum_inter;
-        ret = bdrv_co_is_allocated(intermediate, sector_num, nb_sectors,
-                                   &pnum_inter);
-        if (ret < 0) {
-            return ret;
-        } else if (ret) {
-            *pnum = pnum_inter;
-            return 1;
-        }
-
-        /*
-         * [sector_num, nb_sectors] is unallocated on top but intermediate
-         * might have
-         *
-         * [sector_num+x, nr_sectors] allocated.
-         */
-        if (n > pnum_inter &&
-            (intermediate == top ||
-             sector_num + pnum_inter < intermediate->total_sectors)) {
-            n = pnum_inter;
-        }
-
-        intermediate = intermediate->backing_hd;
+    filtered = bdrv_filter_bs(bs);
+    if (filtered) {
+        return bdrv_has_zero_init(filtered);
     }
 
-    *pnum = n;
+    /* safe default */
     return 0;
 }
 
-/* Coroutine wrapper for bdrv_is_allocated_above() */
-static void coroutine_fn bdrv_is_allocated_above_co_entry(void *opaque)
+bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs)
 {
-    BdrvCoIsAllocatedData *data = opaque;
-    BlockDriverState *top = data->bs;
-    BlockDriverState *base = data->base;
-
-    data->ret = bdrv_co_is_allocated_above(top, base, data->sector_num,
-                                           data->nb_sectors, data->pnum);
-    data->done = true;
-}
-
-/*
- * Synchronous wrapper around bdrv_co_is_allocated_above().
- *
- * See bdrv_co_is_allocated_above() for details.
- */
-int bdrv_is_allocated_above(BlockDriverState *top, BlockDriverState *base,
-                            int64_t sector_num, int nb_sectors, int *pnum)
-{
-    Coroutine *co;
-    BdrvCoIsAllocatedData data = {
-        .bs = top,
-        .base = base,
-        .sector_num = sector_num,
-        .nb_sectors = nb_sectors,
-        .pnum = pnum,
-        .done = false,
-    };
-
-    co = qemu_coroutine_create(bdrv_is_allocated_above_co_entry);
-    qemu_coroutine_enter(co, &data);
-    while (!data.done) {
-        qemu_aio_wait();
+    if (!(bs->open_flags & BDRV_O_UNMAP)) {
+        return false;
     }
-    return data.ret;
-}
 
-const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
-{
-    if (bs->backing_hd && bs->backing_hd->encrypted)
-        return bs->backing_file;
-    else if (bs->encrypted)
-        return bs->filename;
-    else
-        return NULL;
+    return bs->supported_zero_flags & BDRV_REQ_MAY_UNMAP;
 }
 
 void bdrv_get_backing_filename(BlockDriverState *bs,
@@ -3148,75 +5532,44 @@ void bdrv_get_backing_filename(BlockDriverState *bs,
     pstrcpy(filename, filename_size, bs->backing_file);
 }
 
-int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
-                          const uint8_t *buf, int nb_sectors)
-{
-    BlockDriver *drv = bs->drv;
-    if (!drv)
-        return -ENOMEDIUM;
-    if (!drv->bdrv_write_compressed)
-        return -ENOTSUP;
-    if (bdrv_check_request(bs, sector_num, nb_sectors))
-        return -EIO;
-
-    assert(!bs->dirty_bitmap);
-
-    return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
-}
-
 int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
 {
     BlockDriver *drv = bs->drv;
-    if (!drv)
+    /* if bs->drv == NULL, bs is closed, so there's nothing to do here */
+    if (!drv) {
         return -ENOMEDIUM;
-    if (!drv->bdrv_get_info)
+    }
+    if (!drv->bdrv_get_info) {
+        BlockDriverState *filtered = bdrv_filter_bs(bs);
+        if (filtered) {
+            return bdrv_get_info(filtered, bdi);
+        }
         return -ENOTSUP;
+    }
     memset(bdi, 0, sizeof(*bdi));
     return drv->bdrv_get_info(bs, bdi);
 }
 
-int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
-                      int64_t pos, int size)
-{
-    QEMUIOVector qiov;
-    struct iovec iov = {
-        .iov_base   = (void *) buf,
-        .iov_len    = size,
-    };
-
-    qemu_iovec_init_external(&qiov, &iov, 1);
-    return bdrv_writev_vmstate(bs, &qiov, pos);
-}
-
-int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos)
+ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs,
+                                          Error **errp)
 {
     BlockDriver *drv = bs->drv;
-
-    if (!drv) {
-        return -ENOMEDIUM;
-    } else if (drv->bdrv_save_vmstate) {
-        return drv->bdrv_save_vmstate(bs, qiov, pos);
-    } else if (bs->file) {
-        return bdrv_writev_vmstate(bs->file, qiov, pos);
+    if (drv && drv->bdrv_get_specific_info) {
+        return drv->bdrv_get_specific_info(bs, errp);
     }
-
-    return -ENOTSUP;
+    return NULL;
 }
 
-int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
-                      int64_t pos, int size)
+BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs)
 {
     BlockDriver *drv = bs->drv;
-    if (!drv)
-        return -ENOMEDIUM;
-    if (drv->bdrv_load_vmstate)
-        return drv->bdrv_load_vmstate(bs, buf, pos, size);
-    if (bs->file)
-        return bdrv_load_vmstate(bs->file, buf, pos, size);
-    return -ENOTSUP;
+    if (!drv || !drv->bdrv_get_specific_stats) {
+        return NULL;
+    }
+    return drv->bdrv_get_specific_stats(bs);
 }
 
-void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event)
+void bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event)
 {
     if (!bs || !bs->drv || !bs->drv->bdrv_debug_event) {
         return;
@@ -3225,24 +5578,45 @@ void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event)
     bs->drv->bdrv_debug_event(bs, event);
 }
 
-int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
-                          const char *tag)
+static BlockDriverState *bdrv_find_debug_node(BlockDriverState *bs)
 {
     while (bs && bs->drv && !bs->drv->bdrv_debug_breakpoint) {
-        bs = bs->file;
+        bs = bdrv_primary_bs(bs);
     }
 
     if (bs && bs->drv && bs->drv->bdrv_debug_breakpoint) {
+        assert(bs->drv->bdrv_debug_remove_breakpoint);
+        return bs;
+    }
+
+    return NULL;
+}
+
+int bdrv_debug_breakpoint(BlockDriverState *bs, const char *event,
+                          const char *tag)
+{
+    bs = bdrv_find_debug_node(bs);
+    if (bs) {
         return bs->drv->bdrv_debug_breakpoint(bs, event, tag);
     }
 
     return -ENOTSUP;
 }
 
+int bdrv_debug_remove_breakpoint(BlockDriverState *bs, const char *tag)
+{
+    bs = bdrv_find_debug_node(bs);
+    if (bs) {
+        return bs->drv->bdrv_debug_remove_breakpoint(bs, tag);
+    }
+
+    return -ENOTSUP;
+}
+
 int bdrv_debug_resume(BlockDriverState *bs, const char *tag)
 {
-    while (bs && bs->drv && !bs->drv->bdrv_debug_resume) {
-        bs = bs->file;
+    while (bs && (!bs->drv || !bs->drv->bdrv_debug_resume)) {
+        bs = bdrv_primary_bs(bs);
     }
 
     if (bs && bs->drv && bs->drv->bdrv_debug_resume) {
@@ -3255,7 +5629,7 @@ int bdrv_debug_resume(BlockDriverState *bs, const char *tag)
 bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag)
 {
     while (bs && bs->drv && !bs->drv->bdrv_debug_is_suspended) {
-        bs = bs->file;
+        bs = bdrv_primary_bs(bs);
     }
 
     if (bs && bs->drv && bs->drv->bdrv_debug_is_suspended) {
@@ -3265,11 +5639,6 @@ bool bdrv_debug_is_suspended(BlockDriverState *bs, const char *tag)
     return false;
 }
 
-int bdrv_is_snapshot(BlockDriverState *bs)
-{
-    return !!(bs->open_flags & BDRV_O_SNAPSHOT);
-}
-
 /* backing_file can either be relative, or absolute, or a protocol.  If it is
  * relative, it must be relative to the chain.  So, passing in bs->filename
  * from a BDS as backing_file should not be done, as that may be relative to
@@ -3281,8 +5650,10 @@ BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
     char *backing_file_full = NULL;
     char *filename_tmp = NULL;
     int is_protocol = 0;
+    bool filenames_refreshed = false;
     BlockDriverState *curr_bs = NULL;
     BlockDriverState *retval = NULL;
+    BlockDriverState *bs_below;
 
     if (!bs || !bs->drv || !backing_file) {
         return NULL;
@@ -3290,41 +5661,86 @@ BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
 
     filename_full     = g_malloc(PATH_MAX);
     backing_file_full = g_malloc(PATH_MAX);
-    filename_tmp      = g_malloc(PATH_MAX);
 
     is_protocol = path_has_protocol(backing_file);
 
-    for (curr_bs = bs; curr_bs->backing_hd; curr_bs = curr_bs->backing_hd) {
+    /*
+     * Being largely a legacy function, skip any filters here
+     * (because filters do not have normal filenames, so they cannot
+     * match anyway; and allowing json:{} filenames is a bit out of
+     * scope).
+     */
+    for (curr_bs = bdrv_skip_filters(bs);
+         bdrv_cow_child(curr_bs) != NULL;
+         curr_bs = bs_below)
+    {
+        bs_below = bdrv_backing_chain_next(curr_bs);
+
+        if (bdrv_backing_overridden(curr_bs)) {
+            /*
+             * If the backing file was overridden, we can only compare
+             * directly against the backing node's filename.
+             */
+
+            if (!filenames_refreshed) {
+                /*
+                 * This will automatically refresh all of the
+                 * filenames in the rest of the backing chain, so we
+                 * only need to do this once.
+                 */
+                bdrv_refresh_filename(bs_below);
+                filenames_refreshed = true;
+            }
+
+            if (strcmp(backing_file, bs_below->filename) == 0) {
+                retval = bs_below;
+                break;
+            }
+        } else if (is_protocol || path_has_protocol(curr_bs->backing_file)) {
+            /*
+             * If either of the filename paths is actually a protocol, then
+             * compare unmodified paths; otherwise make paths relative.
+             */
+            char *backing_file_full_ret;
 
-        /* If either of the filename paths is actually a protocol, then
-         * compare unmodified paths; otherwise make paths relative */
-        if (is_protocol || path_has_protocol(curr_bs->backing_file)) {
             if (strcmp(backing_file, curr_bs->backing_file) == 0) {
-                retval = curr_bs->backing_hd;
+                retval = bs_below;
                 break;
             }
+            /* Also check against the full backing filename for the image */
+            backing_file_full_ret = bdrv_get_full_backing_filename(curr_bs,
+                                                                   NULL);
+            if (backing_file_full_ret) {
+                bool equal = strcmp(backing_file, backing_file_full_ret) == 0;
+                g_free(backing_file_full_ret);
+                if (equal) {
+                    retval = bs_below;
+                    break;
+                }
+            }
         } else {
             /* If not an absolute filename path, make it relative to the current
              * image's filename path */
-            path_combine(filename_tmp, PATH_MAX, curr_bs->filename,
-                         backing_file);
-
-            /* We are going to compare absolute pathnames */
-            if (!realpath(filename_tmp, filename_full)) {
+            filename_tmp = bdrv_make_absolute_filename(curr_bs, backing_file,
+                                                       NULL);
+            /* We are going to compare canonicalized absolute pathnames */
+            if (!filename_tmp || !realpath(filename_tmp, filename_full)) {
+                g_free(filename_tmp);
                 continue;
             }
+            g_free(filename_tmp);
 
             /* We need to make sure the backing filename we are comparing against
              * is relative to the current image filename (or absolute) */
-            path_combine(filename_tmp, PATH_MAX, curr_bs->filename,
-                         curr_bs->backing_file);
-
-            if (!realpath(filename_tmp, backing_file_full)) {
+            filename_tmp = bdrv_get_full_backing_filename(curr_bs, NULL);
+            if (!filename_tmp || !realpath(filename_tmp, backing_file_full)) {
+                g_free(filename_tmp);
                 continue;
             }
+            g_free(filename_tmp);
 
             if (strcmp(backing_file_full, filename_full) == 0) {
-                retval = curr_bs->backing_hd;
+                retval = bs_below;
                 break;
             }
         }
@@ -3332,1290 +5748,1497 @@ BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs,
 
     g_free(filename_full);
     g_free(backing_file_full);
-    g_free(filename_tmp);
     return retval;
 }
 
-int bdrv_get_backing_file_depth(BlockDriverState *bs)
+void bdrv_init(void)
 {
-    if (!bs->drv) {
-        return 0;
-    }
-
-    if (!bs->backing_hd) {
-        return 0;
-    }
+    module_call_init(MODULE_INIT_BLOCK);
+}
 
-    return 1 + bdrv_get_backing_file_depth(bs->backing_hd);
+void bdrv_init_with_whitelist(void)
+{
+    use_bdrv_whitelist = 1;
+    bdrv_init();
 }
 
-BlockDriverState *bdrv_find_base(BlockDriverState *bs)
+int coroutine_fn bdrv_co_invalidate_cache(BlockDriverState *bs, Error **errp)
 {
-    BlockDriverState *curr_bs = NULL;
+    BdrvChild *child, *parent;
+    uint64_t perm, shared_perm;
+    Error *local_err = NULL;
+    int ret;
+    BdrvDirtyBitmap *bm;
 
-    if (!bs) {
-        return NULL;
+    if (!bs->drv)  {
+        return -ENOMEDIUM;
     }
 
-    curr_bs = bs;
-
-    while (curr_bs->backing_hd) {
-        curr_bs = curr_bs->backing_hd;
+    QLIST_FOREACH(child, &bs->children, next) {
+        bdrv_co_invalidate_cache(child->bs, &local_err);
+        if (local_err) {
+            error_propagate(errp, local_err);
+            return -EINVAL;
+        }
     }
-    return curr_bs;
-}
 
-/**************************************************************/
-/* async I/Os */
-
-BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
-                                 QEMUIOVector *qiov, int nb_sectors,
-                                 BlockDriverCompletionFunc *cb, void *opaque)
-{
-    trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque);
+    /*
+     * Update permissions, they may differ for inactive nodes.
+     *
+     * Note that the required permissions of inactive images are always a
+     * subset of the permissions required after activating the image. This
+     * allows us to just get the permissions upfront without restricting
+     * drv->bdrv_invalidate_cache().
+     *
+     * It also means that in error cases, we don't have to try and revert to
+     * the old permissions (which is an operation that could fail, too). We can
+     * just keep the extended permissions for the next time that an activation
+     * of the image is tried.
+     */
+    if (bs->open_flags & BDRV_O_INACTIVE) {
+        bs->open_flags &= ~BDRV_O_INACTIVE;
+        bdrv_get_cumulative_perm(bs, &perm, &shared_perm);
+        ret = bdrv_check_perm(bs, NULL, perm, shared_perm, NULL, NULL, errp);
+        if (ret < 0) {
+            bdrv_abort_perm_update(bs);
+            bs->open_flags |= BDRV_O_INACTIVE;
+            return ret;
+        }
+        bdrv_set_perm(bs, perm, shared_perm);
 
-    return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors,
-                                 cb, opaque, false);
-}
+        if (bs->drv->bdrv_co_invalidate_cache) {
+            bs->drv->bdrv_co_invalidate_cache(bs, &local_err);
+            if (local_err) {
+                bs->open_flags |= BDRV_O_INACTIVE;
+                error_propagate(errp, local_err);
+                return -EINVAL;
+            }
+        }
 
-BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
-                                  QEMUIOVector *qiov, int nb_sectors,
-                                  BlockDriverCompletionFunc *cb, void *opaque)
-{
-    trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque);
+        FOR_EACH_DIRTY_BITMAP(bs, bm) {
+            bdrv_dirty_bitmap_skip_store(bm, false);
+        }
 
-    return bdrv_co_aio_rw_vector(bs, sector_num, qiov, nb_sectors,
-                                 cb, opaque, true);
-}
+        ret = refresh_total_sectors(bs, bs->total_sectors);
+        if (ret < 0) {
+            bs->open_flags |= BDRV_O_INACTIVE;
+            error_setg_errno(errp, -ret, "Could not refresh total sector count");
+            return ret;
+        }
+    }
 
+    QLIST_FOREACH(parent, &bs->parents, next_parent) {
+        if (parent->klass->activate) {
+            parent->klass->activate(parent, &local_err);
+            if (local_err) {
+                bs->open_flags |= BDRV_O_INACTIVE;
+                error_propagate(errp, local_err);
+                return -EINVAL;
+            }
+        }
+    }
 
-typedef struct MultiwriteCB {
-    int error;
-    int num_requests;
-    int num_callbacks;
-    struct {
-        BlockDriverCompletionFunc *cb;
-        void *opaque;
-        QEMUIOVector *free_qiov;
-    } callbacks[];
-} MultiwriteCB;
+    return 0;
+}
 
-static void multiwrite_user_cb(MultiwriteCB *mcb)
+void bdrv_invalidate_cache_all(Error **errp)
 {
-    int i;
+    BlockDriverState *bs;
+    BdrvNextIterator it;
+
+    for (bs = bdrv_first(&it); bs; bs = bdrv_next(&it)) {
+        AioContext *aio_context = bdrv_get_aio_context(bs);
+        int ret;
 
-    for (i = 0; i < mcb->num_callbacks; i++) {
-        mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
-        if (mcb->callbacks[i].free_qiov) {
-            qemu_iovec_destroy(mcb->callbacks[i].free_qiov);
+        aio_context_acquire(aio_context);
+        ret = bdrv_invalidate_cache(bs, errp);
+        aio_context_release(aio_context);
+        if (ret < 0) {
+            bdrv_next_cleanup(&it);
+            return;
         }
-        g_free(mcb->callbacks[i].free_qiov);
     }
 }
 
-static void multiwrite_cb(void *opaque, int ret)
+static bool bdrv_has_bds_parent(BlockDriverState *bs, bool only_active)
 {
-    MultiwriteCB *mcb = opaque;
+    BdrvChild *parent;
 
-    trace_multiwrite_cb(mcb, ret);
-
-    if (ret < 0 && !mcb->error) {
-        mcb->error = ret;
+    QLIST_FOREACH(parent, &bs->parents, next_parent) {
+        if (parent->klass->parent_is_bds) {
+            BlockDriverState *parent_bs = parent->opaque;
+            if (!only_active || !(parent_bs->open_flags & BDRV_O_INACTIVE)) {
+                return true;
+            }
+        }
     }
 
-    mcb->num_requests--;
-    if (mcb->num_requests == 0) {
-        multiwrite_user_cb(mcb);
-        g_free(mcb);
-    }
+    return false;
 }
 
-static int multiwrite_req_compare(const void *a, const void *b)
+static int bdrv_inactivate_recurse(BlockDriverState *bs)
 {
-    const BlockRequest *req1 = a, *req2 = b;
+    BdrvChild *child, *parent;
+    bool tighten_restrictions;
+    uint64_t perm, shared_perm;
+    int ret;
 
-    /*
-     * Note that we can't simply subtract req2->sector from req1->sector
-     * here as that could overflow the return value.
-     */
-    if (req1->sector > req2->sector) {
-        return 1;
-    } else if (req1->sector < req2->sector) {
-        return -1;
-    } else {
-        return 0;
+    if (!bs->drv) {
+        return -ENOMEDIUM;
     }
-}
 
-/*
- * Takes a bunch of requests and tries to merge them. Returns the number of
- * requests that remain after merging.
- */
-static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
-    int num_reqs, MultiwriteCB *mcb)
-{
-    int i, outidx;
-
-    // Sort requests by start sector
-    qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
+    /* Make sure that we don't inactivate a child before its parent.
+     * It will be covered by recursion from the yet active parent. */
+    if (bdrv_has_bds_parent(bs, true)) {
+        return 0;
+    }
 
-    // Check if adjacent requests touch the same clusters. If so, combine them,
-    // filling up gaps with zero sectors.
-    outidx = 0;
-    for (i = 1; i < num_reqs; i++) {
-        int merge = 0;
-        int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
+    assert(!(bs->open_flags & BDRV_O_INACTIVE));
 
-        // Handle exactly sequential writes and overlapping writes.
-        if (reqs[i].sector <= oldreq_last) {
-            merge = 1;
+    /* Inactivate this node */
+    if (bs->drv->bdrv_inactivate) {
+        ret = bs->drv->bdrv_inactivate(bs);
+        if (ret < 0) {
+            return ret;
         }
+    }
 
-        if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
-            merge = 0;
+    QLIST_FOREACH(parent, &bs->parents, next_parent) {
+        if (parent->klass->inactivate) {
+            ret = parent->klass->inactivate(parent);
+            if (ret < 0) {
+                return ret;
+            }
         }
+    }
 
-        if (merge) {
-            size_t size;
-            QEMUIOVector *qiov = g_malloc0(sizeof(*qiov));
-            qemu_iovec_init(qiov,
-                reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
+    bs->open_flags |= BDRV_O_INACTIVE;
 
-            // Add the first request to the merged one. If the requests are
-            // overlapping, drop the last sectors of the first request.
-            size = (reqs[i].sector - reqs[outidx].sector) << 9;
-            qemu_iovec_concat(qiov, reqs[outidx].qiov, 0, size);
+    /* Update permissions, they may differ for inactive nodes */
+    bdrv_get_cumulative_perm(bs, &perm, &shared_perm);
+    ret = bdrv_check_perm(bs, NULL, perm, shared_perm, NULL,
+                          &tighten_restrictions, NULL);
+    assert(tighten_restrictions == false);
+    if (ret < 0) {
+        /* We only tried to loosen restrictions, so errors are not fatal */
+        bdrv_abort_perm_update(bs);
+    } else {
+        bdrv_set_perm(bs, perm, shared_perm);
+    }
 
-            // We should need to add any zeros between the two requests
-            assert (reqs[i].sector <= oldreq_last);
 
-            // Add the second request
-            qemu_iovec_concat(qiov, reqs[i].qiov, 0, reqs[i].qiov->size);
+    /* Recursively inactivate children */
+    QLIST_FOREACH(child, &bs->children, next) {
+        ret = bdrv_inactivate_recurse(child->bs);
+        if (ret < 0) {
+            return ret;
+        }
+    }
 
-            reqs[outidx].nb_sectors = qiov->size >> 9;
-            reqs[outidx].qiov = qiov;
+    return 0;
+}
 
-            mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
-        } else {
-            outidx++;
-            reqs[outidx].sector     = reqs[i].sector;
-            reqs[outidx].nb_sectors = reqs[i].nb_sectors;
-            reqs[outidx].qiov       = reqs[i].qiov;
+int bdrv_inactivate_all(void)
+{
+    BlockDriverState *bs = NULL;
+    BdrvNextIterator it;
+    int ret = 0;
+    GSList *aio_ctxs = NULL, *ctx;
+
+    for (bs = bdrv_first(&it); bs; bs = bdrv_next(&it)) {
+        AioContext *aio_context = bdrv_get_aio_context(bs);
+
+        if (!g_slist_find(aio_ctxs, aio_context)) {
+            aio_ctxs = g_slist_prepend(aio_ctxs, aio_context);
+            aio_context_acquire(aio_context);
         }
     }
 
-    return outidx + 1;
+    for (bs = bdrv_first(&it); bs; bs = bdrv_next(&it)) {
+        /* Nodes with BDS parents are covered by recursion from the last
+         * parent that gets inactivated. Don't inactivate them a second
+         * time if that has already happened. */
+        if (bdrv_has_bds_parent(bs, false)) {
+            continue;
+        }
+        ret = bdrv_inactivate_recurse(bs);
+        if (ret < 0) {
+            bdrv_next_cleanup(&it);
+            goto out;
+        }
+    }
+
+out:
+    for (ctx = aio_ctxs; ctx != NULL; ctx = ctx->next) {
+        AioContext *aio_context = ctx->data;
+        aio_context_release(aio_context);
+    }
+    g_slist_free(aio_ctxs);
+
+    return ret;
 }
 
-/*
- * Submit multiple AIO write requests at once.
- *
- * On success, the function returns 0 and all requests in the reqs array have
- * been submitted. In error case this function returns -1, and any of the
- * requests may or may not be submitted yet. In particular, this means that the
- * callback will be called for some of the requests, for others it won't. The
- * caller must check the error field of the BlockRequest to wait for the right
- * callbacks (if error != 0, no callback will be called).
- *
- * The implementation may modify the contents of the reqs array, e.g. to merge
- * requests. However, the fields opaque and error are left unmodified as they
- * are used to signal failure for a single request to the caller.
+/**************************************************************/
+/* removable device support */
+
+/**
+ * Return TRUE if the media is present
  */
-int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
+bool bdrv_is_inserted(BlockDriverState *bs)
 {
-    MultiwriteCB *mcb;
-    int i;
+    BlockDriver *drv = bs->drv;
+    BdrvChild *child;
 
-    /* don't submit writes if we don't have a medium */
-    if (bs->drv == NULL) {
-        for (i = 0; i < num_reqs; i++) {
-            reqs[i].error = -ENOMEDIUM;
-        }
-        return -1;
+    if (!drv) {
+        return false;
     }
-
-    if (num_reqs == 0) {
-        return 0;
+    if (drv->bdrv_is_inserted) {
+        return drv->bdrv_is_inserted(bs);
+    }
+    QLIST_FOREACH(child, &bs->children, next) {
+        if (!bdrv_is_inserted(child->bs)) {
+            return false;
+        }
     }
+    return true;
+}
 
-    // Create MultiwriteCB structure
-    mcb = g_malloc0(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
-    mcb->num_requests = 0;
-    mcb->num_callbacks = num_reqs;
+/**
+ * If eject_flag is TRUE, eject the media. Otherwise, close the tray
+ */
+void bdrv_eject(BlockDriverState *bs, bool eject_flag)
+{
+    BlockDriver *drv = bs->drv;
 
-    for (i = 0; i < num_reqs; i++) {
-        mcb->callbacks[i].cb = reqs[i].cb;
-        mcb->callbacks[i].opaque = reqs[i].opaque;
+    if (drv && drv->bdrv_eject) {
+        drv->bdrv_eject(bs, eject_flag);
     }
+}
 
-    // Check for mergable requests
-    num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
+/**
+ * Lock or unlock the media (if it is locked, the user won't be able
+ * to eject it manually).
+ */
+void bdrv_lock_medium(BlockDriverState *bs, bool locked)
+{
+    BlockDriver *drv = bs->drv;
 
-    trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs);
+    trace_bdrv_lock_medium(bs, locked);
 
-    /* Run the aio requests. */
-    mcb->num_requests = num_reqs;
-    for (i = 0; i < num_reqs; i++) {
-        bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
-            reqs[i].nb_sectors, multiwrite_cb, mcb);
+    if (drv && drv->bdrv_lock_medium) {
+        drv->bdrv_lock_medium(bs, locked);
     }
-
-    return 0;
 }
 
-void bdrv_aio_cancel(BlockDriverAIOCB *acb)
+/* Get a reference to bs */
+void bdrv_ref(BlockDriverState *bs)
 {
-    acb->aiocb_info->cancel(acb);
+    bs->refcnt++;
 }
 
-/* block I/O throttling */
-static bool bdrv_exceed_bps_limits(BlockDriverState *bs, int nb_sectors,
-                 bool is_write, double elapsed_time, uint64_t *wait)
+/* Release a previously grabbed reference to bs.
+ * If after releasing, reference count is zero, the BlockDriverState is
+ * deleted. */
+void bdrv_unref(BlockDriverState *bs)
 {
-    uint64_t bps_limit = 0;
-    uint64_t extension;
-    double   bytes_limit, bytes_base, bytes_res;
-    double   slice_time, wait_time;
+    if (!bs) {
+        return;
+    }
+    assert(bs->refcnt > 0);
+    if (--bs->refcnt == 0) {
+        bdrv_delete(bs);
+    }
+}
 
-    if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) {
-        bps_limit = bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL];
-    } else if (bs->io_limits.bps[is_write]) {
-        bps_limit = bs->io_limits.bps[is_write];
-    } else {
-        if (wait) {
-            *wait = 0;
-        }
+struct BdrvOpBlocker {
+    Error *reason;
+    QLIST_ENTRY(BdrvOpBlocker) list;
+};
 
-        return false;
+bool bdrv_op_is_blocked(BlockDriverState *bs, BlockOpType op, Error **errp)
+{
+    BdrvOpBlocker *blocker;
+    assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX);
+    if (!QLIST_EMPTY(&bs->op_blockers[op])) {
+        blocker = QLIST_FIRST(&bs->op_blockers[op]);
+        error_propagate_prepend(errp, error_copy(blocker->reason),
+                                "Node '%s' is busy: ",
+                                bdrv_get_device_or_node_name(bs));
+        return true;
     }
+    return false;
+}
 
-    slice_time = bs->slice_end - bs->slice_start;
-    slice_time /= (NANOSECONDS_PER_SECOND);
-    bytes_limit = bps_limit * slice_time;
-    bytes_base  = bs->slice_submitted.bytes[is_write];
-    if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) {
-        bytes_base += bs->slice_submitted.bytes[!is_write];
-    }
+void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason)
+{
+    BdrvOpBlocker *blocker;
+    assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX);
 
-    /* bytes_base: the bytes of data which have been read/written; and
-     *             it is obtained from the history statistic info.
-     * bytes_res: the remaining bytes of data which need to be read/written.
-     * (bytes_base + bytes_res) / bps_limit: used to calcuate
-     *             the total time for completing reading/writting all data.
-     */
-    bytes_res   = (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
+    blocker = g_new0(BdrvOpBlocker, 1);
+    blocker->reason = reason;
+    QLIST_INSERT_HEAD(&bs->op_blockers[op], blocker, list);
+}
 
-    if (bytes_base + bytes_res <= bytes_limit) {
-        if (wait) {
-            *wait = 0;
+void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason)
+{
+    BdrvOpBlocker *blocker, *next;
+    assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX);
+    QLIST_FOREACH_SAFE(blocker, &bs->op_blockers[op], list, next) {
+        if (blocker->reason == reason) {
+            QLIST_REMOVE(blocker, list);
+            g_free(blocker);
         }
-
-        return false;
     }
+}
 
-    /* Calc approx time to dispatch */
-    wait_time = (bytes_base + bytes_res) / bps_limit - elapsed_time;
-
-    /* When the I/O rate at runtime exceeds the limits,
-     * bs->slice_end need to be extended in order that the current statistic
-     * info can be kept until the timer fire, so it is increased and tuned
-     * based on the result of experiment.
-     */
-    extension = wait_time * NANOSECONDS_PER_SECOND;
-    extension = DIV_ROUND_UP(extension, BLOCK_IO_SLICE_TIME) *
-                BLOCK_IO_SLICE_TIME;
-    bs->slice_end += extension;
-    if (wait) {
-        *wait = wait_time * NANOSECONDS_PER_SECOND;
+void bdrv_op_block_all(BlockDriverState *bs, Error *reason)
+{
+    int i;
+    for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) {
+        bdrv_op_block(bs, i, reason);
     }
+}
 
-    return true;
+void bdrv_op_unblock_all(BlockDriverState *bs, Error *reason)
+{
+    int i;
+    for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) {
+        bdrv_op_unblock(bs, i, reason);
+    }
 }
 
-static bool bdrv_exceed_iops_limits(BlockDriverState *bs, bool is_write,
-                             double elapsed_time, uint64_t *wait)
+bool bdrv_op_blocker_is_empty(BlockDriverState *bs)
 {
-    uint64_t iops_limit = 0;
-    double   ios_limit, ios_base;
-    double   slice_time, wait_time;
+    int i;
 
-    if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) {
-        iops_limit = bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL];
-    } else if (bs->io_limits.iops[is_write]) {
-        iops_limit = bs->io_limits.iops[is_write];
-    } else {
-        if (wait) {
-            *wait = 0;
+    for (i = 0; i < BLOCK_OP_TYPE_MAX; i++) {
+        if (!QLIST_EMPTY(&bs->op_blockers[i])) {
+            return false;
         }
-
-        return false;
     }
+    return true;
+}
 
-    slice_time = bs->slice_end - bs->slice_start;
-    slice_time /= (NANOSECONDS_PER_SECOND);
-    ios_limit  = iops_limit * slice_time;
-    ios_base   = bs->slice_submitted.ios[is_write];
-    if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) {
-        ios_base += bs->slice_submitted.ios[!is_write];
-    }
+void bdrv_img_create(const char *filename, const char *fmt,
+                     const char *base_filename, const char *base_fmt,
+                     char *options, uint64_t img_size, int flags, bool quiet,
+                     Error **errp)
+{
+    QemuOptsList *create_opts = NULL;
+    QemuOpts *opts = NULL;
+    const char *backing_fmt, *backing_file;
+    int64_t size;
+    BlockDriver *drv, *proto_drv;
+    Error *local_err = NULL;
+    int ret = 0;
 
-    if (ios_base + 1 <= ios_limit) {
-        if (wait) {
-            *wait = 0;
-        }
+    /* Find driver and parse its options */
+    drv = bdrv_find_format(fmt);
+    if (!drv) {
+        error_setg(errp, "Unknown file format '%s'", fmt);
+        return;
+    }
 
-        return false;
+    proto_drv = bdrv_find_protocol(filename, true, errp);
+    if (!proto_drv) {
+        return;
     }
 
-    /* Calc approx time to dispatch, in seconds */
-    wait_time = (ios_base + 1) / iops_limit;
-    if (wait_time > elapsed_time) {
-        wait_time = wait_time - elapsed_time;
-    } else {
-        wait_time = 0;
+    if (!drv->create_opts) {
+        error_setg(errp, "Format driver '%s' does not support image creation",
+                   drv->format_name);
+        return;
     }
 
-    /* Exceeded current slice, extend it by another slice time */
-    bs->slice_end += BLOCK_IO_SLICE_TIME;
-    if (wait) {
-        *wait = wait_time * NANOSECONDS_PER_SECOND;
+    if (!proto_drv->create_opts) {
+        error_setg(errp, "Protocol driver '%s' does not support image creation",
+                   proto_drv->format_name);
+        return;
     }
 
-    return true;
-}
+    /* Create parameter list */
+    create_opts = qemu_opts_append(create_opts, drv->create_opts);
+    create_opts = qemu_opts_append(create_opts, proto_drv->create_opts);
 
-static bool bdrv_exceed_io_limits(BlockDriverState *bs, int nb_sectors,
-                           bool is_write, int64_t *wait)
-{
-    int64_t  now, max_wait;
-    uint64_t bps_wait = 0, iops_wait = 0;
-    double   elapsed_time;
-    int      bps_ret, iops_ret;
+    opts = qemu_opts_create(create_opts, NULL, 0, &error_abort);
 
-    now = qemu_get_clock_ns(vm_clock);
-    if (now > bs->slice_end) {
-        bs->slice_start = now;
-        bs->slice_end   = now + BLOCK_IO_SLICE_TIME;
-        memset(&bs->slice_submitted, 0, sizeof(bs->slice_submitted));
+    /* Parse -o options */
+    if (options) {
+        if (!qemu_opts_do_parse(opts, options, NULL, errp)) {
+            goto out;
+        }
     }
 
-    elapsed_time  = now - bs->slice_start;
-    elapsed_time  /= (NANOSECONDS_PER_SECOND);
+    if (!qemu_opt_get(opts, BLOCK_OPT_SIZE)) {
+        qemu_opt_set_number(opts, BLOCK_OPT_SIZE, img_size, &error_abort);
+    } else if (img_size != UINT64_C(-1)) {
+        error_setg(errp, "The image size must be specified only once");
+        goto out;
+    }
 
-    bps_ret  = bdrv_exceed_bps_limits(bs, nb_sectors,
-                                      is_write, elapsed_time, &bps_wait);
-    iops_ret = bdrv_exceed_iops_limits(bs, is_write,
-                                      elapsed_time, &iops_wait);
-    if (bps_ret || iops_ret) {
-        max_wait = bps_wait > iops_wait ? bps_wait : iops_wait;
-        if (wait) {
-            *wait = max_wait;
+    if (base_filename) {
+        if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FILE, base_filename,
+                          NULL)) {
+            error_setg(errp, "Backing file not supported for file format '%s'",
+                       fmt);
+            goto out;
         }
+    }
 
-        now = qemu_get_clock_ns(vm_clock);
-        if (bs->slice_end < now + max_wait) {
-            bs->slice_end = now + max_wait;
+    if (base_fmt) {
+        if (!qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, base_fmt, NULL)) {
+            error_setg(errp, "Backing file format not supported for file "
+                             "format '%s'", fmt);
+            goto out;
         }
-
-        return true;
     }
 
-    if (wait) {
-        *wait = 0;
+    backing_file = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE);
+    if (backing_file) {
+        if (!strcmp(filename, backing_file)) {
+            error_setg(errp, "Error: Trying to create an image with the "
+                             "same filename as the backing file");
+            goto out;
+        }
+        if (backing_file[0] == '\0') {
+            error_setg(errp, "Expected backing file name, got empty string");
+            goto out;
+        }
     }
 
-    bs->slice_submitted.bytes[is_write] += (int64_t)nb_sectors *
-                                           BDRV_SECTOR_SIZE;
-    bs->slice_submitted.ios[is_write]++;
-
-    return false;
-}
-
-/**************************************************************/
-/* async block device emulation */
+    backing_fmt = qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT);
 
-typedef struct BlockDriverAIOCBSync {
-    BlockDriverAIOCB common;
-    QEMUBH *bh;
-    int ret;
-    /* vector translation state */
-    QEMUIOVector *qiov;
-    uint8_t *bounce;
-    int is_write;
-} BlockDriverAIOCBSync;
+    /* The size for the image must always be specified, unless we have a backing
+     * file and we have not been forbidden from opening it. */
+    size = qemu_opt_get_size(opts, BLOCK_OPT_SIZE, img_size);
+    if (backing_file && !(flags & BDRV_O_NO_BACKING)) {
+        BlockDriverState *bs;
+        char *full_backing;
+        int back_flags;
+        QDict *backing_options = NULL;
 
-static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
-{
-    BlockDriverAIOCBSync *acb =
-        container_of(blockacb, BlockDriverAIOCBSync, common);
-    qemu_bh_delete(acb->bh);
-    acb->bh = NULL;
-    qemu_aio_release(acb);
-}
+        full_backing =
+            bdrv_get_full_backing_filename_from_filename(filename, backing_file,
+                                                         &local_err);
+        if (local_err) {
+            goto out;
+        }
+        assert(full_backing);
 
-static const AIOCBInfo bdrv_em_aiocb_info = {
-    .aiocb_size         = sizeof(BlockDriverAIOCBSync),
-    .cancel             = bdrv_aio_cancel_em,
-};
+        /* backing files always opened read-only */
+        back_flags = flags;
+        back_flags &= ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
 
-static void bdrv_aio_bh_cb(void *opaque)
-{
-    BlockDriverAIOCBSync *acb = opaque;
+        backing_options = qdict_new();
+        if (backing_fmt) {
+            qdict_put_str(backing_options, "driver", backing_fmt);
+        }
+        qdict_put_bool(backing_options, BDRV_OPT_FORCE_SHARE, true);
 
-    if (!acb->is_write)
-        qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size);
-    qemu_vfree(acb->bounce);
-    acb->common.cb(acb->common.opaque, acb->ret);
-    qemu_bh_delete(acb->bh);
-    acb->bh = NULL;
-    qemu_aio_release(acb);
-}
+        bs = bdrv_open(full_backing, NULL, backing_options, back_flags,
+                       &local_err);
+        g_free(full_backing);
+        if (!bs) {
+            error_append_hint(&local_err, "Could not open backing image.\n");
+            goto out;
+        } else {
+            if (!backing_fmt) {
+                warn_report("Deprecated use of backing file without explicit "
+                            "backing format (detected format of %s)",
+                            bs->drv->format_name);
+                if (bs->drv != &bdrv_raw) {
+                    /*
+                     * A probe of raw deserves the most attention:
+                     * leaving the backing format out of the image
+                     * will ensure bs->probed is set (ensuring we
+                     * don't accidentally commit into the backing
+                     * file), and allow more spots to warn the users
+                     * to fix their toolchain when opening this image
+                     * later.  For other images, we can safely record
+                     * the format that we probed.
+                     */
+                    backing_fmt = bs->drv->format_name;
+                    qemu_opt_set(opts, BLOCK_OPT_BACKING_FMT, backing_fmt,
+                                 NULL);
+                }
+            }
+            if (size == -1) {
+                /* Opened BS, have no size */
+                size = bdrv_getlength(bs);
+                if (size < 0) {
+                    error_setg_errno(errp, -size, "Could not get size of '%s'",
+                                     backing_file);
+                    bdrv_unref(bs);
+                    goto out;
+                }
+                qemu_opt_set_number(opts, BLOCK_OPT_SIZE, size, &error_abort);
+            }
+            bdrv_unref(bs);
+        }
+        /* (backing_file && !(flags & BDRV_O_NO_BACKING)) */
+    } else if (backing_file && !backing_fmt) {
+        warn_report("Deprecated use of unopened backing file without "
+                    "explicit backing format, use of this image requires "
+                    "potentially unsafe format probing");
+    }
 
-static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
-                                            int64_t sector_num,
-                                            QEMUIOVector *qiov,
-                                            int nb_sectors,
-                                            BlockDriverCompletionFunc *cb,
-                                            void *opaque,
-                                            int is_write)
+    if (size == -1) {
+        error_setg(errp, "Image creation needs a size parameter");
+        goto out;
+    }
 
-{
-    BlockDriverAIOCBSync *acb;
+    if (!quiet) {
+        printf("Formatting '%s', fmt=%s ", filename, fmt);
+        qemu_opts_print(opts, " ");
+        puts("");
+        fflush(stdout);
+    }
 
-    acb = qemu_aio_get(&bdrv_em_aiocb_info, bs, cb, opaque);
-    acb->is_write = is_write;
-    acb->qiov = qiov;
-    acb->bounce = qemu_blockalign(bs, qiov->size);
-    acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
+    ret = bdrv_create(drv, filename, opts, &local_err);
 
-    if (is_write) {
-        qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size);
-        acb->ret = bs->drv->bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
-    } else {
-        acb->ret = bs->drv->bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
+    if (ret == -EFBIG) {
+        /* This is generally a better message than whatever the driver would
+         * deliver (especially because of the cluster_size_hint), since that
+         * is most probably not much different from "image too large". */
+        const char *cluster_size_hint = "";
+        if (qemu_opt_get_size(opts, BLOCK_OPT_CLUSTER_SIZE, 0)) {
+            cluster_size_hint = " (try using a larger cluster size)";
+        }
+        error_setg(errp, "The image size is too large for file format '%s'"
+                   "%s", fmt, cluster_size_hint);
+        error_free(local_err);
+        local_err = NULL;
     }
 
-    qemu_bh_schedule(acb->bh);
-
-    return &acb->common;
+out:
+    qemu_opts_del(opts);
+    qemu_opts_free(create_opts);
+    error_propagate(errp, local_err);
 }
 
-static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
-        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
-        BlockDriverCompletionFunc *cb, void *opaque)
+AioContext *bdrv_get_aio_context(BlockDriverState *bs)
 {
-    return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
+    return bs ? bs->aio_context : qemu_get_aio_context();
 }
 
-static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
-        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
-        BlockDriverCompletionFunc *cb, void *opaque)
+AioContext *coroutine_fn bdrv_co_enter(BlockDriverState *bs)
 {
-    return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
-}
+    Coroutine *self = qemu_coroutine_self();
+    AioContext *old_ctx = qemu_coroutine_get_aio_context(self);
+    AioContext *new_ctx;
 
+    /*
+     * Increase bs->in_flight to ensure that this operation is completed before
+     * moving the node to a different AioContext. Read new_ctx only afterwards.
+     */
+    bdrv_inc_in_flight(bs);
 
-typedef struct BlockDriverAIOCBCoroutine {
-    BlockDriverAIOCB common;
-    BlockRequest req;
-    bool is_write;
-    bool *done;
-    QEMUBH* bh;
-} BlockDriverAIOCBCoroutine;
+    new_ctx = bdrv_get_aio_context(bs);
+    aio_co_reschedule_self(new_ctx);
+    return old_ctx;
+}
 
-static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb)
+void coroutine_fn bdrv_co_leave(BlockDriverState *bs, AioContext *old_ctx)
 {
-    BlockDriverAIOCBCoroutine *acb =
-        container_of(blockacb, BlockDriverAIOCBCoroutine, common);
-    bool done = false;
-
-    acb->done = &done;
-    while (!done) {
-        qemu_aio_wait();
-    }
+    aio_co_reschedule_self(old_ctx);
+    bdrv_dec_in_flight(bs);
 }
 
-static const AIOCBInfo bdrv_em_co_aiocb_info = {
-    .aiocb_size         = sizeof(BlockDriverAIOCBCoroutine),
-    .cancel             = bdrv_aio_co_cancel_em,
-};
-
-static void bdrv_co_em_bh(void *opaque)
+void coroutine_fn bdrv_co_lock(BlockDriverState *bs)
 {
-    BlockDriverAIOCBCoroutine *acb = opaque;
+    AioContext *ctx = bdrv_get_aio_context(bs);
 
-    acb->common.cb(acb->common.opaque, acb->req.error);
+    /* In the main thread, bs->aio_context won't change concurrently */
+    assert(qemu_get_current_aio_context() == qemu_get_aio_context());
 
-    if (acb->done) {
-        *acb->done = true;
+    /*
+     * We're in coroutine context, so we already hold the lock of the main
+     * loop AioContext. Don't lock it twice to avoid deadlocks.
+     */
+    assert(qemu_in_coroutine());
+    if (ctx != qemu_get_aio_context()) {
+        aio_context_acquire(ctx);
     }
-
-    qemu_bh_delete(acb->bh);
-    qemu_aio_release(acb);
 }
 
-/* Invoke bdrv_co_do_readv/bdrv_co_do_writev */
-static void coroutine_fn bdrv_co_do_rw(void *opaque)
+void coroutine_fn bdrv_co_unlock(BlockDriverState *bs)
 {
-    BlockDriverAIOCBCoroutine *acb = opaque;
-    BlockDriverState *bs = acb->common.bs;
+    AioContext *ctx = bdrv_get_aio_context(bs);
 
-    if (!acb->is_write) {
-        acb->req.error = bdrv_co_do_readv(bs, acb->req.sector,
-            acb->req.nb_sectors, acb->req.qiov, 0);
-    } else {
-        acb->req.error = bdrv_co_do_writev(bs, acb->req.sector,
-            acb->req.nb_sectors, acb->req.qiov, 0);
+    assert(qemu_in_coroutine());
+    if (ctx != qemu_get_aio_context()) {
+        aio_context_release(ctx);
     }
-
-    acb->bh = qemu_bh_new(bdrv_co_em_bh, acb);
-    qemu_bh_schedule(acb->bh);
 }
 
-static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs,
-                                               int64_t sector_num,
-                                               QEMUIOVector *qiov,
-                                               int nb_sectors,
-                                               BlockDriverCompletionFunc *cb,
-                                               void *opaque,
-                                               bool is_write)
+void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co)
 {
-    Coroutine *co;
-    BlockDriverAIOCBCoroutine *acb;
-
-    acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque);
-    acb->req.sector = sector_num;
-    acb->req.nb_sectors = nb_sectors;
-    acb->req.qiov = qiov;
-    acb->is_write = is_write;
-    acb->done = NULL;
-
-    co = qemu_coroutine_create(bdrv_co_do_rw);
-    qemu_coroutine_enter(co, acb);
-
-    return &acb->common;
+    aio_co_enter(bdrv_get_aio_context(bs), co);
 }
 
-static void coroutine_fn bdrv_aio_flush_co_entry(void *opaque)
+static void bdrv_do_remove_aio_context_notifier(BdrvAioNotifier *ban)
 {
-    BlockDriverAIOCBCoroutine *acb = opaque;
-    BlockDriverState *bs = acb->common.bs;
-
-    acb->req.error = bdrv_co_flush(bs);
-    acb->bh = qemu_bh_new(bdrv_co_em_bh, acb);
-    qemu_bh_schedule(acb->bh);
+    QLIST_REMOVE(ban, list);
+    g_free(ban);
 }
 
-BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
-        BlockDriverCompletionFunc *cb, void *opaque)
+static void bdrv_detach_aio_context(BlockDriverState *bs)
 {
-    trace_bdrv_aio_flush(bs, opaque);
+    BdrvAioNotifier *baf, *baf_tmp;
 
-    Coroutine *co;
-    BlockDriverAIOCBCoroutine *acb;
-
-    acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque);
-    acb->done = NULL;
+    assert(!bs->walking_aio_notifiers);
+    bs->walking_aio_notifiers = true;
+    QLIST_FOREACH_SAFE(baf, &bs->aio_notifiers, list, baf_tmp) {
+        if (baf->deleted) {
+            bdrv_do_remove_aio_context_notifier(baf);
+        } else {
+            baf->detach_aio_context(baf->opaque);
+        }
+    }
+    /* Never mind iterating again to check for ->deleted.  bdrv_close() will
+     * remove remaining aio notifiers if we aren't called again.
+     */
+    bs->walking_aio_notifiers = false;
 
-    co = qemu_coroutine_create(bdrv_aio_flush_co_entry);
-    qemu_coroutine_enter(co, acb);
+    if (bs->drv && bs->drv->bdrv_detach_aio_context) {
+        bs->drv->bdrv_detach_aio_context(bs);
+    }
 
-    return &acb->common;
+    if (bs->quiesce_counter) {
+        aio_enable_external(bs->aio_context);
+    }
+    bs->aio_context = NULL;
 }
 
-static void coroutine_fn bdrv_aio_discard_co_entry(void *opaque)
+static void bdrv_attach_aio_context(BlockDriverState *bs,
+                                    AioContext *new_context)
 {
-    BlockDriverAIOCBCoroutine *acb = opaque;
-    BlockDriverState *bs = acb->common.bs;
-
-    acb->req.error = bdrv_co_discard(bs, acb->req.sector, acb->req.nb_sectors);
-    acb->bh = qemu_bh_new(bdrv_co_em_bh, acb);
-    qemu_bh_schedule(acb->bh);
-}
+    BdrvAioNotifier *ban, *ban_tmp;
 
-BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs,
-        int64_t sector_num, int nb_sectors,
-        BlockDriverCompletionFunc *cb, void *opaque)
-{
-    Coroutine *co;
-    BlockDriverAIOCBCoroutine *acb;
+    if (bs->quiesce_counter) {
+        aio_disable_external(new_context);
+    }
 
-    trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque);
+    bs->aio_context = new_context;
 
-    acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque);
-    acb->req.sector = sector_num;
-    acb->req.nb_sectors = nb_sectors;
-    acb->done = NULL;
-    co = qemu_coroutine_create(bdrv_aio_discard_co_entry);
-    qemu_coroutine_enter(co, acb);
+    if (bs->drv && bs->drv->bdrv_attach_aio_context) {
+        bs->drv->bdrv_attach_aio_context(bs, new_context);
+    }
 
-    return &acb->common;
+    assert(!bs->walking_aio_notifiers);
+    bs->walking_aio_notifiers = true;
+    QLIST_FOREACH_SAFE(ban, &bs->aio_notifiers, list, ban_tmp) {
+        if (ban->deleted) {
+            bdrv_do_remove_aio_context_notifier(ban);
+        } else {
+            ban->attached_aio_context(new_context, ban->opaque);
+        }
+    }
+    bs->walking_aio_notifiers = false;
 }
 
-void bdrv_init(void)
+/*
+ * Changes the AioContext used for fd handlers, timers, and BHs by this
+ * BlockDriverState and all its children and parents.
+ *
+ * Must be called from the main AioContext.
+ *
+ * The caller must own the AioContext lock for the old AioContext of bs, but it
+ * must not own the AioContext lock for new_context (unless new_context is the
+ * same as the current context of bs).
+ *
+ * @ignore will accumulate all visited BdrvChild object. The caller is
+ * responsible for freeing the list afterwards.
+ */
+void bdrv_set_aio_context_ignore(BlockDriverState *bs,
+                                 AioContext *new_context, GSList **ignore)
 {
-    module_call_init(MODULE_INIT_BLOCK);
-}
+    AioContext *old_context = bdrv_get_aio_context(bs);
+    BdrvChild *child;
 
-void bdrv_init_with_whitelist(void)
-{
-    use_bdrv_whitelist = 1;
-    bdrv_init();
-}
+    g_assert(qemu_get_current_aio_context() == qemu_get_aio_context());
 
-void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs,
-                   BlockDriverCompletionFunc *cb, void *opaque)
-{
-    BlockDriverAIOCB *acb;
+    if (old_context == new_context) {
+        return;
+    }
 
-    acb = g_slice_alloc(aiocb_info->aiocb_size);
-    acb->aiocb_info = aiocb_info;
-    acb->bs = bs;
-    acb->cb = cb;
-    acb->opaque = opaque;
-    return acb;
-}
+    bdrv_drained_begin(bs);
 
-void qemu_aio_release(void *p)
-{
-    BlockDriverAIOCB *acb = p;
-    g_slice_free1(acb->aiocb_info->aiocb_size, acb);
-}
+    QLIST_FOREACH(child, &bs->children, next) {
+        if (g_slist_find(*ignore, child)) {
+            continue;
+        }
+        *ignore = g_slist_prepend(*ignore, child);
+        bdrv_set_aio_context_ignore(child->bs, new_context, ignore);
+    }
+    QLIST_FOREACH(child, &bs->parents, next_parent) {
+        if (g_slist_find(*ignore, child)) {
+            continue;
+        }
+        assert(child->klass->set_aio_ctx);
+        *ignore = g_slist_prepend(*ignore, child);
+        child->klass->set_aio_ctx(child, new_context, ignore);
+    }
 
-/**************************************************************/
-/* Coroutine block device emulation */
+    bdrv_detach_aio_context(bs);
 
-typedef struct CoroutineIOCompletion {
-    Coroutine *coroutine;
-    int ret;
-} CoroutineIOCompletion;
+    /* Acquire the new context, if necessary */
+    if (qemu_get_aio_context() != new_context) {
+        aio_context_acquire(new_context);
+    }
 
-static void bdrv_co_io_em_complete(void *opaque, int ret)
-{
-    CoroutineIOCompletion *co = opaque;
+    bdrv_attach_aio_context(bs, new_context);
 
-    co->ret = ret;
-    qemu_coroutine_enter(co->coroutine, NULL);
-}
+    /*
+     * If this function was recursively called from
+     * bdrv_set_aio_context_ignore(), there may be nodes in the
+     * subtree that have not yet been moved to the new AioContext.
+     * Release the old one so bdrv_drained_end() can poll them.
+     */
+    if (qemu_get_aio_context() != old_context) {
+        aio_context_release(old_context);
+    }
 
-static int coroutine_fn bdrv_co_io_em(BlockDriverState *bs, int64_t sector_num,
-                                      int nb_sectors, QEMUIOVector *iov,
-                                      bool is_write)
-{
-    CoroutineIOCompletion co = {
-        .coroutine = qemu_coroutine_self(),
-    };
-    BlockDriverAIOCB *acb;
+    bdrv_drained_end(bs);
 
-    if (is_write) {
-        acb = bs->drv->bdrv_aio_writev(bs, sector_num, iov, nb_sectors,
-                                       bdrv_co_io_em_complete, &co);
-    } else {
-        acb = bs->drv->bdrv_aio_readv(bs, sector_num, iov, nb_sectors,
-                                      bdrv_co_io_em_complete, &co);
+    if (qemu_get_aio_context() != old_context) {
+        aio_context_acquire(old_context);
     }
-
-    trace_bdrv_co_io_em(bs, sector_num, nb_sectors, is_write, acb);
-    if (!acb) {
-        return -EIO;
+    if (qemu_get_aio_context() != new_context) {
+        aio_context_release(new_context);
     }
-    qemu_coroutine_yield();
-
-    return co.ret;
 }
 
-static int coroutine_fn bdrv_co_readv_em(BlockDriverState *bs,
-                                         int64_t sector_num, int nb_sectors,
-                                         QEMUIOVector *iov)
+static bool bdrv_parent_can_set_aio_context(BdrvChild *c, AioContext *ctx,
+                                            GSList **ignore, Error **errp)
 {
-    return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, false);
-}
+    if (g_slist_find(*ignore, c)) {
+        return true;
+    }
+    *ignore = g_slist_prepend(*ignore, c);
 
-static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs,
-                                         int64_t sector_num, int nb_sectors,
-                                         QEMUIOVector *iov)
-{
-    return bdrv_co_io_em(bs, sector_num, nb_sectors, iov, true);
+    /*
+     * A BdrvChildClass that doesn't handle AioContext changes cannot
+     * tolerate any AioContext changes
+     */
+    if (!c->klass->can_set_aio_ctx) {
+        char *user = bdrv_child_user_desc(c);
+        error_setg(errp, "Changing iothreads is not supported by %s", user);
+        g_free(user);
+        return false;
+    }
+    if (!c->klass->can_set_aio_ctx(c, ctx, ignore, errp)) {
+        assert(!errp || *errp);
+        return false;
+    }
+    return true;
 }
 
-static void coroutine_fn bdrv_flush_co_entry(void *opaque)
+bool bdrv_child_can_set_aio_context(BdrvChild *c, AioContext *ctx,
+                                    GSList **ignore, Error **errp)
 {
-    RwCo *rwco = opaque;
-
-    rwco->ret = bdrv_co_flush(rwco->bs);
+    if (g_slist_find(*ignore, c)) {
+        return true;
+    }
+    *ignore = g_slist_prepend(*ignore, c);
+    return bdrv_can_set_aio_context(c->bs, ctx, ignore, errp);
 }
 
-int coroutine_fn bdrv_co_flush(BlockDriverState *bs)
+/* @ignore will accumulate all visited BdrvChild object. The caller is
+ * responsible for freeing the list afterwards. */
+bool bdrv_can_set_aio_context(BlockDriverState *bs, AioContext *ctx,
+                              GSList **ignore, Error **errp)
 {
-    int ret;
+    BdrvChild *c;
 
-    if (!bs || !bdrv_is_inserted(bs) || bdrv_is_read_only(bs)) {
-        return 0;
+    if (bdrv_get_aio_context(bs) == ctx) {
+        return true;
     }
 
-    /* Write back cached data to the OS even with cache=unsafe */
-    BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_OS);
-    if (bs->drv->bdrv_co_flush_to_os) {
-        ret = bs->drv->bdrv_co_flush_to_os(bs);
-        if (ret < 0) {
-            return ret;
+    QLIST_FOREACH(c, &bs->parents, next_parent) {
+        if (!bdrv_parent_can_set_aio_context(c, ctx, ignore, errp)) {
+            return false;
         }
     }
-
-    /* But don't actually force it to the disk with cache=unsafe */
-    if (bs->open_flags & BDRV_O_NO_FLUSH) {
-        goto flush_parent;
-    }
-
-    BLKDBG_EVENT(bs->file, BLKDBG_FLUSH_TO_DISK);
-    if (bs->drv->bdrv_co_flush_to_disk) {
-        ret = bs->drv->bdrv_co_flush_to_disk(bs);
-    } else if (bs->drv->bdrv_aio_flush) {
-        BlockDriverAIOCB *acb;
-        CoroutineIOCompletion co = {
-            .coroutine = qemu_coroutine_self(),
-        };
-
-        acb = bs->drv->bdrv_aio_flush(bs, bdrv_co_io_em_complete, &co);
-        if (acb == NULL) {
-            ret = -EIO;
-        } else {
-            qemu_coroutine_yield();
-            ret = co.ret;
+    QLIST_FOREACH(c, &bs->children, next) {
+        if (!bdrv_child_can_set_aio_context(c, ctx, ignore, errp)) {
+            return false;
         }
-    } else {
-        /*
-         * Some block drivers always operate in either writethrough or unsafe
-         * mode and don't support bdrv_flush therefore. Usually qemu doesn't
-         * know how the server works (because the behaviour is hardcoded or
-         * depends on server-side configuration), so we can't ensure that
-         * everything is safe on disk. Returning an error doesn't work because
-         * that would break guests even if the server operates in writethrough
-         * mode.
-         *
-         * Let's hope the user knows what he's doing.
-         */
-        ret = 0;
-    }
-    if (ret < 0) {
-        return ret;
     }
 
-    /* Now flush the underlying protocol.  It will also have BDRV_O_NO_FLUSH
-     * in the case of cache=unsafe, so there are no useless flushes.
-     */
-flush_parent:
-    return bdrv_co_flush(bs->file);
+    return true;
 }
 
-void bdrv_invalidate_cache(BlockDriverState *bs)
+int bdrv_child_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
+                                   BdrvChild *ignore_child, Error **errp)
 {
-    if (bs->drv && bs->drv->bdrv_invalidate_cache) {
-        bs->drv->bdrv_invalidate_cache(bs);
-    }
-}
+    GSList *ignore;
+    bool ret;
 
-void bdrv_invalidate_cache_all(void)
-{
-    BlockDriverState *bs;
+    ignore = ignore_child ? g_slist_prepend(NULL, ignore_child) : NULL;
+    ret = bdrv_can_set_aio_context(bs, ctx, &ignore, errp);
+    g_slist_free(ignore);
 
-    QTAILQ_FOREACH(bs, &bdrv_states, list) {
-        bdrv_invalidate_cache(bs);
+    if (!ret) {
+        return -EPERM;
     }
+
+    ignore = ignore_child ? g_slist_prepend(NULL, ignore_child) : NULL;
+    bdrv_set_aio_context_ignore(bs, ctx, &ignore);
+    g_slist_free(ignore);
+
+    return 0;
 }
 
-void bdrv_clear_incoming_migration_all(void)
+int bdrv_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
+                             Error **errp)
 {
-    BlockDriverState *bs;
-
-    QTAILQ_FOREACH(bs, &bdrv_states, list) {
-        bs->open_flags = bs->open_flags & ~(BDRV_O_INCOMING);
-    }
+    return bdrv_child_try_set_aio_context(bs, ctx, NULL, errp);
 }
 
-int bdrv_flush(BlockDriverState *bs)
+void bdrv_add_aio_context_notifier(BlockDriverState *bs,
+        void (*attached_aio_context)(AioContext *new_context, void *opaque),
+        void (*detach_aio_context)(void *opaque), void *opaque)
 {
-    Coroutine *co;
-    RwCo rwco = {
-        .bs = bs,
-        .ret = NOT_DONE,
+    BdrvAioNotifier *ban = g_new(BdrvAioNotifier, 1);
+    *ban = (BdrvAioNotifier){
+        .attached_aio_context = attached_aio_context,
+        .detach_aio_context   = detach_aio_context,
+        .opaque               = opaque
     };
 
-    if (qemu_in_coroutine()) {
-        /* Fast-path if already in coroutine context */
-        bdrv_flush_co_entry(&rwco);
-    } else {
-        co = qemu_coroutine_create(bdrv_flush_co_entry);
-        qemu_coroutine_enter(co, &rwco);
-        while (rwco.ret == NOT_DONE) {
-            qemu_aio_wait();
-        }
-    }
-
-    return rwco.ret;
+    QLIST_INSERT_HEAD(&bs->aio_notifiers, ban, list);
 }
 
-static void coroutine_fn bdrv_discard_co_entry(void *opaque)
+void bdrv_remove_aio_context_notifier(BlockDriverState *bs,
+                                      void (*attached_aio_context)(AioContext *,
+                                                                   void *),
+                                      void (*detach_aio_context)(void *),
+                                      void *opaque)
 {
-    RwCo *rwco = opaque;
+    BdrvAioNotifier *ban, *ban_next;
 
-    rwco->ret = bdrv_co_discard(rwco->bs, rwco->sector_num, rwco->nb_sectors);
+    QLIST_FOREACH_SAFE(ban, &bs->aio_notifiers, list, ban_next) {
+        if (ban->attached_aio_context == attached_aio_context &&
+            ban->detach_aio_context   == detach_aio_context   &&
+            ban->opaque               == opaque               &&
+            ban->deleted              == false)
+        {
+            if (bs->walking_aio_notifiers) {
+                ban->deleted = true;
+            } else {
+                bdrv_do_remove_aio_context_notifier(ban);
+            }
+            return;
+        }
+    }
+
+    abort();
 }
 
-int coroutine_fn bdrv_co_discard(BlockDriverState *bs, int64_t sector_num,
-                                 int nb_sectors)
+int bdrv_amend_options(BlockDriverState *bs, QemuOpts *opts,
+                       BlockDriverAmendStatusCB *status_cb, void *cb_opaque,
+                       bool force,
+                       Error **errp)
 {
     if (!bs->drv) {
+        error_setg(errp, "Node is ejected");
         return -ENOMEDIUM;
-    } else if (bdrv_check_request(bs, sector_num, nb_sectors)) {
-        return -EIO;
-    } else if (bs->read_only) {
-        return -EROFS;
     }
+    if (!bs->drv->bdrv_amend_options) {
+        error_setg(errp, "Block driver '%s' does not support option amendment",
+                   bs->drv->format_name);
+        return -ENOTSUP;
+    }
+    return bs->drv->bdrv_amend_options(bs, opts, status_cb,
+                                       cb_opaque, force, errp);
+}
+
+/*
+ * This function checks whether the given @to_replace is allowed to be
+ * replaced by a node that always shows the same data as @bs.  This is
+ * used for example to verify whether the mirror job can replace
+ * @to_replace by the target mirrored from @bs.
+ * To be replaceable, @bs and @to_replace may either be guaranteed to
+ * always show the same data (because they are only connected through
+ * filters), or some driver may allow replacing one of its children
+ * because it can guarantee that this child's data is not visible at
+ * all (for example, for dissenting quorum children that have no other
+ * parents).
+ */
+bool bdrv_recurse_can_replace(BlockDriverState *bs,
+                              BlockDriverState *to_replace)
+{
+    BlockDriverState *filtered;
 
-    if (bs->dirty_bitmap) {
-        bdrv_reset_dirty(bs, sector_num, nb_sectors);
+    if (!bs || !bs->drv) {
+        return false;
     }
 
-    /* Do nothing if disabled.  */
-    if (!(bs->open_flags & BDRV_O_UNMAP)) {
-        return 0;
+    if (bs == to_replace) {
+        return true;
     }
 
-    if (bs->drv->bdrv_co_discard) {
-        return bs->drv->bdrv_co_discard(bs, sector_num, nb_sectors);
-    } else if (bs->drv->bdrv_aio_discard) {
-        BlockDriverAIOCB *acb;
-        CoroutineIOCompletion co = {
-            .coroutine = qemu_coroutine_self(),
-        };
+    /* See what the driver can do */
+    if (bs->drv->bdrv_recurse_can_replace) {
+        return bs->drv->bdrv_recurse_can_replace(bs, to_replace);
+    }
 
-        acb = bs->drv->bdrv_aio_discard(bs, sector_num, nb_sectors,
-                                        bdrv_co_io_em_complete, &co);
-        if (acb == NULL) {
-            return -EIO;
-        } else {
-            qemu_coroutine_yield();
-            return co.ret;
-        }
-    } else {
-        return 0;
+    /* For filters without an own implementation, we can recurse on our own */
+    filtered = bdrv_filter_bs(bs);
+    if (filtered) {
+        return bdrv_recurse_can_replace(filtered, to_replace);
     }
+
+    /* Safe default */
+    return false;
 }
 
-int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors)
+/*
+ * Check whether the given @node_name can be replaced by a node that
+ * has the same data as @parent_bs.  If so, return @node_name's BDS;
+ * NULL otherwise.
+ *
+ * @node_name must be a (recursive) *child of @parent_bs (or this
+ * function will return NULL).
+ *
+ * The result (whether the node can be replaced or not) is only valid
+ * for as long as no graph or permission changes occur.
+ */
+BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs,
+                                        const char *node_name, Error **errp)
 {
-    Coroutine *co;
-    RwCo rwco = {
-        .bs = bs,
-        .sector_num = sector_num,
-        .nb_sectors = nb_sectors,
-        .ret = NOT_DONE,
-    };
+    BlockDriverState *to_replace_bs = bdrv_find_node(node_name);
+    AioContext *aio_context;
 
-    if (qemu_in_coroutine()) {
-        /* Fast-path if already in coroutine context */
-        bdrv_discard_co_entry(&rwco);
-    } else {
-        co = qemu_coroutine_create(bdrv_discard_co_entry);
-        qemu_coroutine_enter(co, &rwco);
-        while (rwco.ret == NOT_DONE) {
-            qemu_aio_wait();
-        }
+    if (!to_replace_bs) {
+        error_setg(errp, "Node name '%s' not found", node_name);
+        return NULL;
     }
 
-    return rwco.ret;
-}
+    aio_context = bdrv_get_aio_context(to_replace_bs);
+    aio_context_acquire(aio_context);
 
-/**************************************************************/
-/* removable device support */
+    if (bdrv_op_is_blocked(to_replace_bs, BLOCK_OP_TYPE_REPLACE, errp)) {
+        to_replace_bs = NULL;
+        goto out;
+    }
 
-/**
- * Return TRUE if the media is present
- */
-int bdrv_is_inserted(BlockDriverState *bs)
-{
-    BlockDriver *drv = bs->drv;
+    /* We don't want arbitrary node of the BDS chain to be replaced only the top
+     * most non filter in order to prevent data corruption.
+     * Another benefit is that this tests exclude backing files which are
+     * blocked by the backing blockers.
+     */
+    if (!bdrv_recurse_can_replace(parent_bs, to_replace_bs)) {
+        error_setg(errp, "Cannot replace '%s' by a node mirrored from '%s', "
+                   "because it cannot be guaranteed that doing so would not "
+                   "lead to an abrupt change of visible data",
+                   node_name, parent_bs->node_name);
+        to_replace_bs = NULL;
+        goto out;
+    }
 
-    if (!drv)
-        return 0;
-    if (!drv->bdrv_is_inserted)
-        return 1;
-    return drv->bdrv_is_inserted(bs);
+out:
+    aio_context_release(aio_context);
+    return to_replace_bs;
 }
 
 /**
- * Return whether the media changed since the last call to this
- * function, or -ENOTSUP if we don't know.  Most drivers don't know.
+ * Iterates through the list of runtime option keys that are said to
+ * be "strong" for a BDS.  An option is called "strong" if it changes
+ * a BDS's data.  For example, the null block driver's "size" and
+ * "read-zeroes" options are strong, but its "latency-ns" option is
+ * not.
+ *
+ * If a key returned by this function ends with a dot, all options
+ * starting with that prefix are strong.
  */
-int bdrv_media_changed(BlockDriverState *bs)
+static const char *const *strong_options(BlockDriverState *bs,
+                                         const char *const *curopt)
 {
-    BlockDriver *drv = bs->drv;
+    static const char *const global_options[] = {
+        "driver", "filename", NULL
+    };
 
-    if (drv && drv->bdrv_media_changed) {
-        return drv->bdrv_media_changed(bs);
+    if (!curopt) {
+        return &global_options[0];
     }
-    return -ENOTSUP;
+
+    curopt++;
+    if (curopt == &global_options[ARRAY_SIZE(global_options) - 1] && bs->drv) {
+        curopt = bs->drv->strong_runtime_opts;
+    }
+
+    return (curopt && *curopt) ? curopt : NULL;
 }
 
 /**
- * If eject_flag is TRUE, eject the media. Otherwise, close the tray
+ * Copies all strong runtime options from bs->options to the given
+ * QDict.  The set of strong option keys is determined by invoking
+ * strong_options().
+ *
+ * Returns true iff any strong option was present in bs->options (and
+ * thus copied to the target QDict) with the exception of "filename"
+ * and "driver".  The caller is expected to use this value to decide
+ * whether the existence of strong options prevents the generation of
+ * a plain filename.
  */
-void bdrv_eject(BlockDriverState *bs, bool eject_flag)
+static bool append_strong_runtime_options(QDict *d, BlockDriverState *bs)
 {
-    BlockDriver *drv = bs->drv;
+    bool found_any = false;
+    const char *const *option_name = NULL;
 
-    if (drv && drv->bdrv_eject) {
-        drv->bdrv_eject(bs, eject_flag);
+    if (!bs->drv) {
+        return false;
     }
 
-    if (bs->device_name[0] != '\0') {
-        bdrv_emit_qmp_eject_event(bs, eject_flag);
+    while ((option_name = strong_options(bs, option_name))) {
+        bool option_given = false;
+
+        assert(strlen(*option_name) > 0);
+        if ((*option_name)[strlen(*option_name) - 1] != '.') {
+            QObject *entry = qdict_get(bs->options, *option_name);
+            if (!entry) {
+                continue;
+            }
+
+            qdict_put_obj(d, *option_name, qobject_ref(entry));
+            option_given = true;
+        } else {
+            const QDictEntry *entry;
+            for (entry = qdict_first(bs->options); entry;
+                 entry = qdict_next(bs->options, entry))
+            {
+                if (strstart(qdict_entry_key(entry), *option_name, NULL)) {
+                    qdict_put_obj(d, qdict_entry_key(entry),
+                                  qobject_ref(qdict_entry_value(entry)));
+                    option_given = true;
+                }
+            }
+        }
+
+        /* While "driver" and "filename" need to be included in a JSON filename,
+         * their existence does not prohibit generation of a plain filename. */
+        if (!found_any && option_given &&
+            strcmp(*option_name, "driver") && strcmp(*option_name, "filename"))
+        {
+            found_any = true;
+        }
+    }
+
+    if (!qdict_haskey(d, "driver")) {
+        /* Drivers created with bdrv_new_open_driver() may not have a
+         * @driver option.  Add it here. */
+        qdict_put_str(d, "driver", bs->drv->format_name);
     }
+
+    return found_any;
 }
 
-/**
- * Lock or unlock the media (if it is locked, the user won't be able
- * to eject it manually).
+/* Note: This function may return false positives; it may return true
+ * even if opening the backing file specified by bs's image header
+ * would result in exactly bs->backing. */
+bool bdrv_backing_overridden(BlockDriverState *bs)
+{
+    if (bs->backing) {
+        return strcmp(bs->auto_backing_file,
+                      bs->backing->bs->filename);
+    } else {
+        /* No backing BDS, so if the image header reports any backing
+         * file, it must have been suppressed */
+        return bs->auto_backing_file[0] != '\0';
+    }
+}
+
+/* Updates the following BDS fields:
+ *  - exact_filename: A filename which may be used for opening a block device
+ *                    which (mostly) equals the given BDS (even without any
+ *                    other options; so reading and writing must return the same
+ *                    results, but caching etc. may be different)
+ *  - full_open_options: Options which, when given when opening a block device
+ *                       (without a filename), result in a BDS (mostly)
+ *                       equalling the given one
+ *  - filename: If exact_filename is set, it is copied here. Otherwise,
+ *              full_open_options is converted to a JSON object, prefixed with
+ *              "json:" (for use through the JSON pseudo protocol) and put here.
  */
-void bdrv_lock_medium(BlockDriverState *bs, bool locked)
+void bdrv_refresh_filename(BlockDriverState *bs)
 {
     BlockDriver *drv = bs->drv;
+    BdrvChild *child;
+    BlockDriverState *primary_child_bs;
+    QDict *opts;
+    bool backing_overridden;
+    bool generate_json_filename; /* Whether our default implementation should
+                                    fill exact_filename (false) or not (true) */
 
-    trace_bdrv_lock_medium(bs, locked);
+    if (!drv) {
+        return;
+    }
 
-    if (drv && drv->bdrv_lock_medium) {
-        drv->bdrv_lock_medium(bs, locked);
+    /* This BDS's file name may depend on any of its children's file names, so
+     * refresh those first */
+    QLIST_FOREACH(child, &bs->children, next) {
+        bdrv_refresh_filename(child->bs);
     }
-}
 
-/* needed for generic scsi interface */
+    if (bs->implicit) {
+        /* For implicit nodes, just copy everything from the single child */
+        child = QLIST_FIRST(&bs->children);
+        assert(QLIST_NEXT(child, next) == NULL);
 
-int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
-{
-    BlockDriver *drv = bs->drv;
+        pstrcpy(bs->exact_filename, sizeof(bs->exact_filename),
+                child->bs->exact_filename);
+        pstrcpy(bs->filename, sizeof(bs->filename), child->bs->filename);
 
-    if (drv && drv->bdrv_ioctl)
-        return drv->bdrv_ioctl(bs, req, buf);
-    return -ENOTSUP;
-}
+        qobject_unref(bs->full_open_options);
+        bs->full_open_options = qobject_ref(child->bs->full_open_options);
 
-BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
-        unsigned long int req, void *buf,
-        BlockDriverCompletionFunc *cb, void *opaque)
-{
-    BlockDriver *drv = bs->drv;
+        return;
+    }
 
-    if (drv && drv->bdrv_aio_ioctl)
-        return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
-    return NULL;
-}
+    backing_overridden = bdrv_backing_overridden(bs);
 
-void bdrv_set_buffer_alignment(BlockDriverState *bs, int align)
-{
-    bs->buffer_alignment = align;
-}
+    if (bs->open_flags & BDRV_O_NO_IO) {
+        /* Without I/O, the backing file does not change anything.
+         * Therefore, in such a case (primarily qemu-img), we can
+         * pretend the backing file has not been overridden even if
+         * it technically has been. */
+        backing_overridden = false;
+    }
 
-void *qemu_blockalign(BlockDriverState *bs, size_t size)
-{
-    return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
-}
+    /* Gather the options QDict */
+    opts = qdict_new();
+    generate_json_filename = append_strong_runtime_options(opts, bs);
+    generate_json_filename |= backing_overridden;
 
-/*
- * Check if all memory in this vector is sector aligned.
- */
-bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov)
-{
-    int i;
+    if (drv->bdrv_gather_child_options) {
+        /* Some block drivers may not want to present all of their children's
+         * options, or name them differently from BdrvChild.name */
+        drv->bdrv_gather_child_options(bs, opts, backing_overridden);
+    } else {
+        QLIST_FOREACH(child, &bs->children, next) {
+            if (child == bs->backing && !backing_overridden) {
+                /* We can skip the backing BDS if it has not been overridden */
+                continue;
+            }
 
-    for (i = 0; i < qiov->niov; i++) {
-        if ((uintptr_t) qiov->iov[i].iov_base % bs->buffer_alignment) {
-            return false;
+            qdict_put(opts, child->name,
+                      qobject_ref(child->bs->full_open_options));
+        }
+
+        if (backing_overridden && !bs->backing) {
+            /* Force no backing file */
+            qdict_put_null(opts, "backing");
         }
     }
 
-    return true;
-}
+    qobject_unref(bs->full_open_options);
+    bs->full_open_options = opts;
 
-void bdrv_set_dirty_tracking(BlockDriverState *bs, int granularity)
-{
-    int64_t bitmap_size;
+    primary_child_bs = bdrv_primary_bs(bs);
 
-    assert((granularity & (granularity - 1)) == 0);
+    if (drv->bdrv_refresh_filename) {
+        /* Obsolete information is of no use here, so drop the old file name
+         * information before refreshing it */
+        bs->exact_filename[0] = '\0';
 
-    if (granularity) {
-        granularity >>= BDRV_SECTOR_BITS;
-        assert(!bs->dirty_bitmap);
-        bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS);
-        bs->dirty_bitmap = hbitmap_alloc(bitmap_size, ffs(granularity) - 1);
-    } else {
-        if (bs->dirty_bitmap) {
-            hbitmap_free(bs->dirty_bitmap);
-            bs->dirty_bitmap = NULL;
+        drv->bdrv_refresh_filename(bs);
+    } else if (primary_child_bs) {
+        /*
+         * Try to reconstruct valid information from the underlying
+         * file -- this only works for format nodes (filter nodes
+         * cannot be probed and as such must be selected by the user
+         * either through an options dict, or through a special
+         * filename which the filter driver must construct in its
+         * .bdrv_refresh_filename() implementation).
+         */
+
+        bs->exact_filename[0] = '\0';
+
+        /*
+         * We can use the underlying file's filename if:
+         * - it has a filename,
+         * - the current BDS is not a filter,
+         * - the file is a protocol BDS, and
+         * - opening that file (as this BDS's format) will automatically create
+         *   the BDS tree we have right now, that is:
+         *   - the user did not significantly change this BDS's behavior with
+         *     some explicit (strong) options
+         *   - no non-file child of this BDS has been overridden by the user
+         *   Both of these conditions are represented by generate_json_filename.
+         */
+        if (primary_child_bs->exact_filename[0] &&
+            primary_child_bs->drv->bdrv_file_open &&
+            !drv->is_filter && !generate_json_filename)
+        {
+            strcpy(bs->exact_filename, primary_child_bs->exact_filename);
         }
     }
-}
 
-int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
-{
-    if (bs->dirty_bitmap) {
-        return hbitmap_get(bs->dirty_bitmap, sector);
+    if (bs->exact_filename[0]) {
+        pstrcpy(bs->filename, sizeof(bs->filename), bs->exact_filename);
     } else {
-        return 0;
+        QString *json = qobject_to_json(QOBJECT(bs->full_open_options));
+        if (snprintf(bs->filename, sizeof(bs->filename), "json:%s",
+                     qstring_get_str(json)) >= sizeof(bs->filename)) {
+            /* Give user a hint if we truncated things. */
+            strcpy(bs->filename + sizeof(bs->filename) - 4, "...");
+        }
+        qobject_unref(json);
     }
 }
 
-void bdrv_dirty_iter_init(BlockDriverState *bs, HBitmapIter *hbi)
+char *bdrv_dirname(BlockDriverState *bs, Error **errp)
 {
-    hbitmap_iter_init(hbi, bs->dirty_bitmap, 0);
-}
+    BlockDriver *drv = bs->drv;
+    BlockDriverState *child_bs;
 
-void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector,
-                    int nr_sectors)
-{
-    hbitmap_set(bs->dirty_bitmap, cur_sector, nr_sectors);
-}
+    if (!drv) {
+        error_setg(errp, "Node '%s' is ejected", bs->node_name);
+        return NULL;
+    }
 
-void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
-                      int nr_sectors)
-{
-    hbitmap_reset(bs->dirty_bitmap, cur_sector, nr_sectors);
-}
+    if (drv->bdrv_dirname) {
+        return drv->bdrv_dirname(bs, errp);
+    }
 
-int64_t bdrv_get_dirty_count(BlockDriverState *bs)
-{
-    if (bs->dirty_bitmap) {
-        return hbitmap_count(bs->dirty_bitmap);
-    } else {
-        return 0;
+    child_bs = bdrv_primary_bs(bs);
+    if (child_bs) {
+        return bdrv_dirname(child_bs, errp);
     }
-}
 
-void bdrv_set_in_use(BlockDriverState *bs, int in_use)
-{
-    assert(bs->in_use != in_use);
-    bs->in_use = in_use;
-}
+    bdrv_refresh_filename(bs);
+    if (bs->exact_filename[0] != '\0') {
+        return path_combine(bs->exact_filename, "");
+    }
 
-int bdrv_in_use(BlockDriverState *bs)
-{
-    return bs->in_use;
+    error_setg(errp, "Cannot generate a base directory for %s nodes",
+               drv->format_name);
+    return NULL;
 }
 
-void bdrv_iostatus_enable(BlockDriverState *bs)
+/*
+ * Hot add/remove a BDS's child. So the user can take a child offline when
+ * it is broken and take a new child online
+ */
+void bdrv_add_child(BlockDriverState *parent_bs, BlockDriverState *child_bs,
+                    Error **errp)
 {
-    bs->iostatus_enabled = true;
-    bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK;
-}
 
-/* The I/O status is only enabled if the drive explicitly
- * enables it _and_ the VM is configured to stop on errors */
-bool bdrv_iostatus_is_enabled(const BlockDriverState *bs)
-{
-    return (bs->iostatus_enabled &&
-           (bs->on_write_error == BLOCKDEV_ON_ERROR_ENOSPC ||
-            bs->on_write_error == BLOCKDEV_ON_ERROR_STOP   ||
-            bs->on_read_error == BLOCKDEV_ON_ERROR_STOP));
-}
+    if (!parent_bs->drv || !parent_bs->drv->bdrv_add_child) {
+        error_setg(errp, "The node %s does not support adding a child",
+                   bdrv_get_device_or_node_name(parent_bs));
+        return;
+    }
 
-void bdrv_iostatus_disable(BlockDriverState *bs)
-{
-    bs->iostatus_enabled = false;
+    if (!QLIST_EMPTY(&child_bs->parents)) {
+        error_setg(errp, "The node %s already has a parent",
+                   child_bs->node_name);
+        return;
+    }
+
+    parent_bs->drv->bdrv_add_child(parent_bs, child_bs, errp);
 }
 
-void bdrv_iostatus_reset(BlockDriverState *bs)
+void bdrv_del_child(BlockDriverState *parent_bs, BdrvChild *child, Error **errp)
 {
-    if (bdrv_iostatus_is_enabled(bs)) {
-        bs->iostatus = BLOCK_DEVICE_IO_STATUS_OK;
-        if (bs->job) {
-            block_job_iostatus_reset(bs->job);
+    BdrvChild *tmp;
+
+    if (!parent_bs->drv || !parent_bs->drv->bdrv_del_child) {
+        error_setg(errp, "The node %s does not support removing a child",
+                   bdrv_get_device_or_node_name(parent_bs));
+        return;
+    }
+
+    QLIST_FOREACH(tmp, &parent_bs->children, next) {
+        if (tmp == child) {
+            break;
         }
     }
-}
 
-void bdrv_iostatus_set_err(BlockDriverState *bs, int error)
-{
-    assert(bdrv_iostatus_is_enabled(bs));
-    if (bs->iostatus == BLOCK_DEVICE_IO_STATUS_OK) {
-        bs->iostatus = error == ENOSPC ? BLOCK_DEVICE_IO_STATUS_NOSPACE :
-                                         BLOCK_DEVICE_IO_STATUS_FAILED;
+    if (!tmp) {
+        error_setg(errp, "The node %s does not have a child named %s",
+                   bdrv_get_device_or_node_name(parent_bs),
+                   bdrv_get_device_or_node_name(child->bs));
+        return;
     }
+
+    parent_bs->drv->bdrv_del_child(parent_bs, child, errp);
 }
 
-void
-bdrv_acct_start(BlockDriverState *bs, BlockAcctCookie *cookie, int64_t bytes,
-        enum BlockAcctType type)
+int bdrv_make_empty(BdrvChild *c, Error **errp)
 {
-    assert(type < BDRV_MAX_IOTYPE);
+    BlockDriver *drv = c->bs->drv;
+    int ret;
 
-    cookie->bytes = bytes;
-    cookie->start_time_ns = get_clock();
-    cookie->type = type;
-}
+    assert(c->perm & (BLK_PERM_WRITE | BLK_PERM_WRITE_UNCHANGED));
 
-void
-bdrv_acct_done(BlockDriverState *bs, BlockAcctCookie *cookie)
-{
-    assert(cookie->type < BDRV_MAX_IOTYPE);
+    if (!drv->bdrv_make_empty) {
+        error_setg(errp, "%s does not support emptying nodes",
+                   drv->format_name);
+        return -ENOTSUP;
+    }
 
-    bs->nr_bytes[cookie->type] += cookie->bytes;
-    bs->nr_ops[cookie->type]++;
-    bs->total_time_ns[cookie->type] += get_clock() - cookie->start_time_ns;
+    ret = drv->bdrv_make_empty(c->bs);
+    if (ret < 0) {
+        error_setg_errno(errp, -ret, "Failed to empty %s",
+                         c->bs->filename);
+        return ret;
+    }
+
+    return 0;
 }
 
-void bdrv_img_create(const char *filename, const char *fmt,
-                     const char *base_filename, const char *base_fmt,
-                     char *options, uint64_t img_size, int flags,
-                     Error **errp, bool quiet)
+/*
+ * Return the child that @bs acts as an overlay for, and from which data may be
+ * copied in COW or COR operations.  Usually this is the backing file.
+ */
+BdrvChild *bdrv_cow_child(BlockDriverState *bs)
 {
-    QEMUOptionParameter *param = NULL, *create_options = NULL;
-    QEMUOptionParameter *backing_fmt, *backing_file, *size;
-    BlockDriverState *bs = NULL;
-    BlockDriver *drv, *proto_drv;
-    BlockDriver *backing_drv = NULL;
-    int ret = 0;
-
-    /* Find driver and parse its options */
-    drv = bdrv_find_format(fmt);
-    if (!drv) {
-        error_setg(errp, "Unknown file format '%s'", fmt);
-        return;
+    if (!bs || !bs->drv) {
+        return NULL;
     }
 
-    proto_drv = bdrv_find_protocol(filename, true);
-    if (!proto_drv) {
-        error_setg(errp, "Unknown protocol '%s'", filename);
-        return;
+    if (bs->drv->is_filter) {
+        return NULL;
     }
 
-    create_options = append_option_parameters(create_options,
-                                              drv->create_options);
-    create_options = append_option_parameters(create_options,
-                                              proto_drv->create_options);
+    if (!bs->backing) {
+        return NULL;
+    }
 
-    /* Create parameter list with default values */
-    param = parse_option_parameters("", create_options, param);
+    assert(bs->backing->role & BDRV_CHILD_COW);
+    return bs->backing;
+}
 
-    set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size);
+/*
+ * If @bs acts as a filter for exactly one of its children, return
+ * that child.
+ */
+BdrvChild *bdrv_filter_child(BlockDriverState *bs)
+{
+    BdrvChild *c;
 
-    /* Parse -o options */
-    if (options) {
-        param = parse_option_parameters(options, create_options, param);
-        if (param == NULL) {
-            error_setg(errp, "Invalid options for file format '%s'.", fmt);
-            goto out;
-        }
+    if (!bs || !bs->drv) {
+        return NULL;
     }
 
-    if (base_filename) {
-        if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE,
-                                 base_filename)) {
-            error_setg(errp, "Backing file not supported for file format '%s'",
-                       fmt);
-            goto out;
-        }
+    if (!bs->drv->is_filter) {
+        return NULL;
     }
 
-    if (base_fmt) {
-        if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) {
-            error_setg(errp, "Backing file format not supported for file "
-                             "format '%s'", fmt);
-            goto out;
-        }
-    }
+    /* Only one of @backing or @file may be used */
+    assert(!(bs->backing && bs->file));
 
-    backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE);
-    if (backing_file && backing_file->value.s) {
-        if (!strcmp(filename, backing_file->value.s)) {
-            error_setg(errp, "Error: Trying to create an image with the "
-                             "same filename as the backing file");
-            goto out;
-        }
+    c = bs->backing ?: bs->file;
+    if (!c) {
+        return NULL;
     }
 
-    backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT);
-    if (backing_fmt && backing_fmt->value.s) {
-        backing_drv = bdrv_find_format(backing_fmt->value.s);
-        if (!backing_drv) {
-            error_setg(errp, "Unknown backing file format '%s'",
-                       backing_fmt->value.s);
-            goto out;
-        }
-    }
+    assert(c->role & BDRV_CHILD_FILTERED);
+    return c;
+}
 
-    // The size for the image must always be specified, with one exception:
-    // If we are using a backing file, we can obtain the size from there
-    size = get_option_parameter(param, BLOCK_OPT_SIZE);
-    if (size && size->value.n == -1) {
-        if (backing_file && backing_file->value.s) {
-            uint64_t size;
-            char buf[32];
-            int back_flags;
+/*
+ * Return either the result of bdrv_cow_child() or bdrv_filter_child(),
+ * whichever is non-NULL.
+ *
+ * Return NULL if both are NULL.
+ */
+BdrvChild *bdrv_filter_or_cow_child(BlockDriverState *bs)
+{
+    BdrvChild *cow_child = bdrv_cow_child(bs);
+    BdrvChild *filter_child = bdrv_filter_child(bs);
 
-            /* backing files always opened read-only */
-            back_flags =
-                flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
+    /* Filter nodes cannot have COW backing files */
+    assert(!(cow_child && filter_child));
 
-            bs = bdrv_new("");
+    return cow_child ?: filter_child;
+}
 
-            ret = bdrv_open(bs, backing_file->value.s, NULL, back_flags,
-                            backing_drv);
-            if (ret < 0) {
-                error_setg_errno(errp, -ret, "Could not open '%s'",
-                                 backing_file->value.s);
-                goto out;
-            }
-            bdrv_get_geometry(bs, &size);
-            size *= 512;
+/*
+ * Return the primary child of this node: For filters, that is the
+ * filtered child.  For other nodes, that is usually the child storing
+ * metadata.
+ * (A generally more helpful description is that this is (usually) the
+ * child that has the same filename as @bs.)
+ *
+ * Drivers do not necessarily have a primary child; for example quorum
+ * does not.
+ */
+BdrvChild *bdrv_primary_child(BlockDriverState *bs)
+{
+    BdrvChild *c, *found = NULL;
 
-            snprintf(buf, sizeof(buf), "%" PRId64, size);
-            set_option_parameter(param, BLOCK_OPT_SIZE, buf);
-        } else {
-            error_setg(errp, "Image creation needs a size parameter");
-            goto out;
+    QLIST_FOREACH(c, &bs->children, next) {
+        if (c->role & BDRV_CHILD_PRIMARY) {
+            assert(!found);
+            found = c;
         }
     }
 
-    if (!quiet) {
-        printf("Formatting '%s', fmt=%s ", filename, fmt);
-        print_option_parameters(param);
-        puts("");
+    return found;
+}
+
+static BlockDriverState *bdrv_do_skip_filters(BlockDriverState *bs,
+                                              bool stop_on_explicit_filter)
+{
+    BdrvChild *c;
+
+    if (!bs) {
+        return NULL;
     }
-    ret = bdrv_create(drv, filename, param);
-    if (ret < 0) {
-        if (ret == -ENOTSUP) {
-            error_setg(errp,"Formatting or formatting option not supported for "
-                            "file format '%s'", fmt);
-        } else if (ret == -EFBIG) {
-            const char *cluster_size_hint = "";
-            if (get_option_parameter(create_options, BLOCK_OPT_CLUSTER_SIZE)) {
-                cluster_size_hint = " (try using a larger cluster size)";
-            }
-            error_setg(errp, "The image size is too large for file format '%s'%s",
-                       fmt, cluster_size_hint);
-        } else {
-            error_setg(errp, "%s: error while creating %s: %s", filename, fmt,
-                       strerror(-ret));
+
+    while (!(stop_on_explicit_filter && !bs->implicit)) {
+        c = bdrv_filter_child(bs);
+        if (!c) {
+            /*
+             * A filter that is embedded in a working block graph must
+             * have a child.  Assert this here so this function does
+             * not return a filter node that is not expected by the
+             * caller.
+             */
+            assert(!bs->drv || !bs->drv->is_filter);
+            break;
         }
+        bs = c->bs;
     }
+    /*
+     * Note that this treats nodes with bs->drv == NULL as not being
+     * filters (bs->drv == NULL should be replaced by something else
+     * anyway).
+     * The advantage of this behavior is that this function will thus
+     * always return a non-NULL value (given a non-NULL @bs).
+     */
 
-out:
-    free_option_parameters(create_options);
-    free_option_parameters(param);
+    return bs;
+}
 
-    if (bs) {
-        bdrv_delete(bs);
-    }
+/*
+ * Return the first BDS that has not been added implicitly or that
+ * does not have a filtered child down the chain starting from @bs
+ * (including @bs itself).
+ */
+BlockDriverState *bdrv_skip_implicit_filters(BlockDriverState *bs)
+{
+    return bdrv_do_skip_filters(bs, true);
 }
 
-AioContext *bdrv_get_aio_context(BlockDriverState *bs)
+/*
+ * Return the first BDS that does not have a filtered child down the
+ * chain starting from @bs (including @bs itself).
+ */
+BlockDriverState *bdrv_skip_filters(BlockDriverState *bs)
 {
-    /* Currently BlockDriverState always uses the main loop AioContext */
-    return qemu_get_aio_context();
+    return bdrv_do_skip_filters(bs, false);
 }
 
-void bdrv_add_before_write_notifier(BlockDriverState *bs,
-                                    NotifierWithReturn *notifier)
+/*
+ * For a backing chain, return the first non-filter backing image of
+ * the first non-filter image.
+ */
+BlockDriverState *bdrv_backing_chain_next(BlockDriverState *bs)
 {
-    notifier_with_return_list_add(&bs->before_write_notifiers, notifier);
+    return bdrv_skip_filters(bdrv_cow_bs(bdrv_skip_filters(bs)));
 }