* [PATCH v11 1/5] docs/qcow2: add the zoned format feature
2026-06-01 21:44 [PATCH v11 0/5] Add full zoned storage emulation to the qcow2 driver Sam Li
@ 2026-06-01 21:44 ` Sam Li
2026-06-01 21:44 ` [PATCH v11 2/5] qcow2: add configurations for zoned format extension Sam Li
` (4 subsequent siblings)
5 siblings, 0 replies; 13+ messages in thread
From: Sam Li @ 2026-06-01 21:44 UTC (permalink / raw)
To: qemu-devel
Cc: Markus Armbruster, qemu-block, Eric Blake, Stefan Hajnoczi,
Pierrick Bouvier, dmitry.fomichev, Hanna Reitz, hare,
Michael S. Tsirkin, Kevin Wolf, cassel, dlemoal, Sam Li
Add the specs for the zoned format feature of the qcow2 driver.
The qcow2 file then can emulate real zoned devices, either passed
through by virtio-blk device or NVMe ZNS drive to the guest
given zoned information.
Signed-off-by: Sam Li <faithilikerun@gmail.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
---
docs/system/qemu-block-drivers.rst.inc | 47 ++++++++++++++++++++++++++
1 file changed, 47 insertions(+)
diff --git a/docs/system/qemu-block-drivers.rst.inc b/docs/system/qemu-block-drivers.rst.inc
index 675daa72f9..1af968c4fd 100644
--- a/docs/system/qemu-block-drivers.rst.inc
+++ b/docs/system/qemu-block-drivers.rst.inc
@@ -172,6 +172,53 @@ This section describes each format and the options that are supported for it.
filename`` to check if the NOCOW flag is set or not (Capital 'C' is
NOCOW flag).
+ .. option:: zone.mode
+ If this is set to ``host-managed``, the image is an emulated zoned
+ block device. This option is only valid to emulated zoned device files.
+
+ .. option:: zone.size
+
+ The size of a zone in bytes. The device is divided into zones of this
+ size with the exception of the last zone, which may be smaller.
+ Defaults to ``256 MiB``.
+
+ .. option:: zone.capacity
+
+ The initial capacity value, in bytes, for all zones. The capacity must
+ be less than or equal to zone size. If the last zone is smaller, then
+ its capacity is capped. Defaults to ``zone.size``.
+
+ The zone capacity is per zone and may be different between zones in real
+ devices. QCow2 sets all zones to the same capacity.
+
+ .. option:: zone.conventional_zones
+
+ The number of conventional zones of the zoned device. Defaults to ``0``
+ (all zones are sequential-write-required).
+
+ .. option:: zone.max_active_zones
+
+ The limit of the zones with implicit open, explicit open or closed state.
+
+ The max active zones must be less or equal to the number of SWR
+ (sequential write required) zones of the device. Defaults to ``0``,
+ meaning no limit.
+
+ .. option:: zone.max_open_zones
+
+ The maximal allowed open zones. The max open zones must not be larger than
+ the max active zones. Defaults to ``zone.max_active_zones`` if that is
+ set, otherwise ``0`` (no limit).
+
+ If the limits of open zones or active zones are equal to the number of
+ SWR zones, then it is the same as having no limits.
+
+ .. option:: zone.max_append_bytes
+
+ The number of bytes in a zone append request that can be issued to the
+ device. It must be 512-byte aligned and less than the zone capacity.
+ Defaults to ``64 KiB``.
+
.. program:: image-formats
.. option:: qed
--
2.53.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* [PATCH v11 2/5] qcow2: add configurations for zoned format extension
2026-06-01 21:44 [PATCH v11 0/5] Add full zoned storage emulation to the qcow2 driver Sam Li
2026-06-01 21:44 ` [PATCH v11 1/5] docs/qcow2: add the zoned format feature Sam Li
@ 2026-06-01 21:44 ` Sam Li
2026-06-02 20:03 ` Stefan Hajnoczi
2026-06-03 7:37 ` Markus Armbruster
2026-06-01 21:44 ` [PATCH v11 3/5] virtio-blk: do not merge writes across a zone boundary Sam Li
` (3 subsequent siblings)
5 siblings, 2 replies; 13+ messages in thread
From: Sam Li @ 2026-06-01 21:44 UTC (permalink / raw)
To: qemu-devel
Cc: Markus Armbruster, qemu-block, Eric Blake, Stefan Hajnoczi,
Pierrick Bouvier, dmitry.fomichev, Hanna Reitz, hare,
Michael S. Tsirkin, Kevin Wolf, cassel, dlemoal, Sam Li
To configure the zoned format feature on the qcow2 driver, it
requires settings as: the device size, zone model, zone size,
zone capacity, number of conventional zones, limits on zone
resources (max append bytes, max open zones, and max_active_zones).
To create a qcow2 image with zoned format feature, use command like
this:
qemu-img create -f qcow2 zbc.qcow2 -o size=768M \
-o zone.size=64M -o zone.capacity=64M -o zone.conventional_zones=0 \
-o zone.max_append_bytes=4096 -o zone.max_open_zones=6 \
-o zone.max_active_zones=8 -o zone.mode=host-managed
Signed-off-by: Sam Li <faithilikerun@gmail.com>
---
block/file-posix.c | 2 +-
block/qcow2.c | 329 ++++++++++++++++++++++++++++++-
block/qcow2.h | 83 +++++++-
docs/interop/qcow2.rst | 117 ++++++++++-
include/block/block_int-common.h | 15 +-
qapi/block-core.json | 91 ++++++++-
6 files changed, 628 insertions(+), 9 deletions(-)
diff --git a/block/file-posix.c b/block/file-posix.c
index e49b13d6ab..14278785b9 100644
--- a/block/file-posix.c
+++ b/block/file-posix.c
@@ -3607,7 +3607,7 @@ raw_co_zone_append(BlockDriverState *bs,
if (*offset & zone_size_mask) {
error_report("sector offset %" PRId64 " is not aligned to zone size "
- "%" PRId32 "", *offset / 512, bs->bl.zone_size / 512);
+ "%" PRId64 "", *offset / 512, bs->bl.zone_size / 512);
return -EINVAL;
}
diff --git a/block/qcow2.c b/block/qcow2.c
index 81fd299b4c..29eec33e34 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -73,6 +73,7 @@ typedef struct {
#define QCOW2_EXT_MAGIC_CRYPTO_HEADER 0x0537be77
#define QCOW2_EXT_MAGIC_BITMAPS 0x23852875
#define QCOW2_EXT_MAGIC_DATA_FILE 0x44415441
+#define QCOW2_EXT_MAGIC_ZONED_FORMAT 0x007a6264
static int coroutine_fn
qcow2_co_preadv_compressed(BlockDriverState *bs,
@@ -194,6 +195,93 @@ qcow2_extract_crypto_opts(QemuOpts *opts, const char *fmt, Error **errp)
return cryptoopts_qdict;
}
+/*
+ * Passing by the zoned device configurations by a zoned_header struct, check
+ * if the zone device options are under constraints. Return false when some
+ * option is invalid
+ */
+static bool
+qcow2_check_zone_options(Qcow2ZonedHeaderExtension *zone_opt, Error **errp)
+{
+ uint32_t sequential_zones;
+
+ assert(zone_opt != NULL);
+
+ if (zone_opt->zoned != QCOW2_Z_NONE && zone_opt->zoned != QCOW2_Z_HM) {
+ error_setg(errp, "Zoned extension header zoned field has unknown "
+ "value %" PRIu8, zone_opt->zoned);
+ return false;
+ }
+
+ if (!is_power_of_2(zone_opt->zone_size)) {
+ error_setg(errp, "Zoned extension header zone_size %" PRIu64
+ "B is not a power of 2", zone_opt->zone_size);
+ return false;
+ }
+
+ if (zone_opt->nr_zones > UINT32_MAX / 8) {
+ error_setg(errp, "Zoned extension header nr_zones %" PRIu32
+ " exceeds maximum %u",
+ zone_opt->nr_zones, UINT32_MAX / 8);
+ return false;
+ }
+
+ if (zone_opt->zone_capacity > zone_opt->zone_size) {
+ error_setg(errp, "zone capacity %" PRIu64 "B exceeds zone size "
+ "%" PRIu64 "B", zone_opt->zone_capacity,
+ zone_opt->zone_size);
+ return false;
+ }
+
+ if (!QEMU_IS_ALIGNED(zone_opt->max_append_bytes, BDRV_SECTOR_SIZE)) {
+ error_setg(errp, "max append bytes %" PRIu32 "B is not aligned "
+ "to %" PRIu64, zone_opt->max_append_bytes,
+ (uint64_t)BDRV_SECTOR_SIZE);
+ return false;
+ }
+
+ if (zone_opt->max_append_bytes + BDRV_SECTOR_SIZE >=
+ zone_opt->zone_capacity) {
+ error_setg(errp, "max append bytes %" PRIu32 "B exceeds zone "
+ "capacity %" PRIu64 "B by more than block size",
+ zone_opt->max_append_bytes, zone_opt->zone_capacity);
+ return false;
+ }
+
+ if (zone_opt->conventional_zones >= zone_opt->nr_zones) {
+ error_setg(errp, "Conventional_zones %" PRIu32 " exceeds "
+ "nr_zones %" PRIu32 ".",
+ zone_opt->conventional_zones, zone_opt->nr_zones);
+ return false;
+ }
+
+ if (zone_opt->max_active_zones > zone_opt->nr_zones) {
+ error_setg(errp, "Max_active_zones %" PRIu32 " exceeds "
+ "nr_zones %" PRIu32 ". Set it to nr_zones.",
+ zone_opt->max_active_zones, zone_opt->nr_zones);
+ zone_opt->max_active_zones = zone_opt->nr_zones;
+ }
+
+ sequential_zones = zone_opt->nr_zones - zone_opt->conventional_zones;
+ if (zone_opt->max_open_zones > sequential_zones) {
+ error_setg(errp, "Max_open_zones field can not be larger than"
+ "the number of SWR zones. Set it to number of SWR"
+ "zones %" PRIu32 ".", sequential_zones);
+ zone_opt->max_open_zones = sequential_zones;
+ }
+ if (zone_opt->max_active_zones != 0 &&
+ zone_opt->max_open_zones > zone_opt->max_active_zones) {
+ error_setg(errp, "Max_open_zones %" PRIu32 " exceeds "
+ "max_active_zones %" PRIu32 ". Set it to "
+ "max_active_zones.",
+ zone_opt->max_open_zones,
+ zone_opt->max_active_zones);
+ zone_opt->max_open_zones = zone_opt->max_active_zones;
+ }
+
+ return true;
+}
+
/*
* read qcow2 extension and fill bs
* start reading from start_offset
@@ -211,6 +299,7 @@ qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset,
uint64_t offset;
int ret;
Qcow2BitmapHeaderExt bitmaps_ext;
+ Qcow2ZonedHeaderExtension zoned_ext;
if (need_update_header != NULL) {
*need_update_header = false;
@@ -432,6 +521,50 @@ qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset,
break;
}
+ case QCOW2_EXT_MAGIC_ZONED_FORMAT:
+ {
+ if (ext.len != sizeof(zoned_ext)) {
+ error_setg(errp, "zoned_ext: unexpected len=%" PRIu32 " "
+ "(expected %zu)", ext.len, sizeof(zoned_ext));
+ return -EINVAL;
+ }
+ ret = bdrv_pread(bs->file, offset, ext.len, &zoned_ext, 0);
+ if (ret < 0) {
+ error_setg_errno(errp, -ret, "zoned_ext: "
+ "Could not read ext header");
+ return ret;
+ }
+
+ zoned_ext.zone_size = be64_to_cpu(zoned_ext.zone_size);
+ zoned_ext.zone_capacity = be64_to_cpu(zoned_ext.zone_capacity);
+ zoned_ext.conventional_zones =
+ be32_to_cpu(zoned_ext.conventional_zones);
+ zoned_ext.nr_zones = be32_to_cpu(zoned_ext.nr_zones);
+ zoned_ext.max_open_zones = be32_to_cpu(zoned_ext.max_open_zones);
+ zoned_ext.max_active_zones =
+ be32_to_cpu(zoned_ext.max_active_zones);
+ zoned_ext.max_append_bytes =
+ be32_to_cpu(zoned_ext.max_append_bytes);
+ s->zoned_header = zoned_ext;
+
+ /* refuse to open broken images */
+ if (zoned_ext.nr_zones != DIV_ROUND_UP(bs->total_sectors *
+ BDRV_SECTOR_SIZE, zoned_ext.zone_size)) {
+ error_setg(errp, "Zoned extension header nr_zones field "
+ "is wrong");
+ return -EINVAL;
+ }
+ if (!qcow2_check_zone_options(&zoned_ext, errp)) {
+ return -EINVAL;
+ }
+
+#ifdef DEBUG_EXT
+ printf("Qcow2: Got zoned format extension: "
+ "offset=%" PRIu32 "\n", offset);
+#endif
+ break;
+ }
+
default:
/* unknown magic - save it in case we need to rewrite the header */
/* If you add a new feature, make sure to also update the fast
@@ -2068,6 +2201,25 @@ static void qcow2_refresh_limits(BlockDriverState *bs, Error **errp)
}
bs->bl.pwrite_zeroes_alignment = s->subcluster_size;
bs->bl.pdiscard_alignment = s->cluster_size;
+
+ switch (s->zoned_header.zoned) {
+ case QCOW2_Z_HM:
+ bs->bl.zoned = BLK_Z_HM;
+ break;
+ case QCOW2_Z_NONE:
+ default:
+ bs->bl.zoned = BLK_Z_NONE;
+ break;
+ }
+
+ bs->bl.nr_zones = s->zoned_header.nr_zones;
+ bs->bl.max_append_sectors = s->zoned_header.max_append_bytes
+ >> BDRV_SECTOR_BITS;
+ bs->bl.max_active_zones = s->zoned_header.max_active_zones;
+ bs->bl.max_open_zones = s->zoned_header.max_open_zones;
+ bs->bl.zone_size = s->zoned_header.zone_size;
+ bs->bl.zone_capacity = s->zoned_header.zone_capacity;
+ bs->bl.write_granularity = BDRV_SECTOR_SIZE;
}
static int GRAPH_UNLOCKED
@@ -3170,6 +3322,11 @@ int qcow2_update_header(BlockDriverState *bs)
.bit = QCOW2_INCOMPAT_EXTL2_BITNR,
.name = "extended L2 entries",
},
+ {
+ .type = QCOW2_FEAT_TYPE_INCOMPATIBLE,
+ .bit = QCOW2_INCOMPAT_ZONED_FORMAT_BITNR,
+ .name = "zoned format",
+ },
{
.type = QCOW2_FEAT_TYPE_COMPATIBLE,
.bit = QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR,
@@ -3215,6 +3372,31 @@ int qcow2_update_header(BlockDriverState *bs)
buflen -= ret;
}
+ /* Zoned devices header extension */
+ if (s->zoned_header.zoned == QCOW2_Z_HM) {
+ Qcow2ZonedHeaderExtension zoned_header = {
+ .zoned = s->zoned_header.zoned,
+ .zone_size = cpu_to_be64(s->zoned_header.zone_size),
+ .zone_capacity = cpu_to_be64(s->zoned_header.zone_capacity),
+ .conventional_zones =
+ cpu_to_be32(s->zoned_header.conventional_zones),
+ .nr_zones = cpu_to_be32(s->zoned_header.nr_zones),
+ .max_open_zones = cpu_to_be32(s->zoned_header.max_open_zones),
+ .max_active_zones =
+ cpu_to_be32(s->zoned_header.max_active_zones),
+ .max_append_bytes =
+ cpu_to_be32(s->zoned_header.max_append_bytes)
+ };
+ ret = header_ext_add(buf, QCOW2_EXT_MAGIC_ZONED_FORMAT,
+ &zoned_header, sizeof(zoned_header),
+ buflen);
+ if (ret < 0) {
+ goto fail;
+ }
+ buf += ret;
+ buflen -= ret;
+ }
+
/* Keep unknown header extensions */
QLIST_FOREACH(uext, &s->unknown_header_ext, next) {
ret = header_ext_add(buf, uext->magic, uext->data, uext->len, buflen);
@@ -3589,6 +3771,8 @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp)
ERRP_GUARD();
BlockdevCreateOptionsQcow2 *qcow2_opts;
QDict *options;
+ Qcow2ZoneCreateOptions *zone_struct;
+ Qcow2ZoneHostManaged *zone_host_managed;
/*
* Open the image file and write a minimal qcow2 header.
@@ -3615,6 +3799,8 @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp)
assert(create_options->driver == BLOCKDEV_DRIVER_QCOW2);
qcow2_opts = &create_options->u.qcow2;
+ zone_struct = create_options->u.qcow2.zone;
+ zone_host_managed = &create_options->u.qcow2.zone->u.host_managed;
bs = bdrv_co_open_blockdev_ref(qcow2_opts->file, errp);
if (bs == NULL) {
@@ -3828,6 +4014,14 @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp)
header->incompatible_features |=
cpu_to_be64(QCOW2_INCOMPAT_DATA_FILE);
}
+ if (zone_struct && zone_struct->mode == QCOW2_ZONE_MODEL_HOST_MANAGED) {
+ /*
+ * The incompatible bit must be set when the zone model is
+ * host-managed
+ */
+ header->incompatible_features |=
+ cpu_to_be64(QCOW2_INCOMPAT_ZONED_FORMAT);
+ }
if (qcow2_opts->data_file_raw) {
header->autoclear_features |=
cpu_to_be64(QCOW2_AUTOCLEAR_DATA_FILE_RAW);
@@ -3885,10 +4079,9 @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp)
bdrv_graph_co_rdlock();
ret = qcow2_alloc_clusters(blk_bs(blk), 3 * cluster_size);
if (ret < 0) {
- bdrv_graph_co_rdunlock();
error_setg_errno(errp, -ret, "Could not allocate clusters for qcow2 "
"header and refcount table");
- goto out;
+ goto unlock;
} else if (ret != 0) {
error_report("Huh, first cluster in empty image is already in use?");
@@ -3896,11 +4089,76 @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp)
}
/* Set the external data file if necessary */
+ BDRVQcow2State *s = blk_bs(blk)->opaque;
if (data_bs) {
- BDRVQcow2State *s = blk_bs(blk)->opaque;
s->image_data_file = g_strdup(data_bs->filename);
}
+ if (zone_struct && zone_struct->mode == QCOW2_ZONE_MODEL_HOST_MANAGED) {
+ s->zoned_header.zoned = QCOW2_Z_HM;
+
+ if (zone_host_managed->has_size) {
+ s->zoned_header.zone_size = zone_host_managed->size;
+ } else {
+ s->zoned_header.zone_size = DEFAULT_ZONE_SIZE;
+ }
+
+ if (s->zoned_header.zone_size == 0) {
+ error_setg(errp, "Zoned extension header zone_size field "
+ "can not be 0");
+ s->zoned_header.zoned = QCOW2_Z_NONE;
+ ret = -EINVAL;
+ goto unlock;
+ }
+ s->zoned_header.nr_zones = DIV_ROUND_UP(qcow2_opts->size,
+ s->zoned_header.zone_size);
+
+ if (zone_host_managed->has_capacity) {
+ s->zoned_header.zone_capacity = zone_host_managed->capacity;
+ } else {
+ s->zoned_header.zone_capacity = s->zoned_header.zone_size;
+ }
+
+ if (zone_host_managed->has_conventional_zones) {
+ s->zoned_header.conventional_zones =
+ zone_host_managed->conventional_zones;
+ } else {
+ s->zoned_header.conventional_zones = 0;
+ }
+
+ if (zone_host_managed->has_max_active_zones) {
+ s->zoned_header.max_active_zones =
+ zone_host_managed->max_active_zones;
+ } else {
+ s->zoned_header.max_active_zones = 0;
+ }
+
+ if (zone_host_managed->has_max_open_zones) {
+ s->zoned_header.max_open_zones =
+ zone_host_managed->max_open_zones;
+ } else if (zone_host_managed->has_max_active_zones) {
+ s->zoned_header.max_open_zones =
+ zone_host_managed->max_active_zones;
+ } else {
+ s->zoned_header.max_open_zones = 0;
+ }
+
+ if (zone_host_managed->has_max_append_bytes) {
+ s->zoned_header.max_append_bytes =
+ zone_host_managed->max_append_bytes;
+ } else {
+ s->zoned_header.max_append_bytes = DEFAULT_ZONE_MAX_APPEND_BYTES;
+ }
+
+ if (!qcow2_check_zone_options(&s->zoned_header, errp)) {
+ s->zoned_header.zoned = QCOW2_Z_NONE;
+ ret = -EINVAL;
+ goto unlock;
+ }
+ } else {
+ s->zoned_header.zoned = QCOW2_Z_NONE;
+ }
+
/* Create a full header (including things like feature table) */
ret = qcow2_update_header(blk_bs(blk));
bdrv_graph_co_rdunlock();
@@ -3974,6 +4232,9 @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp)
}
ret = 0;
+ goto out;
+unlock:
+ bdrv_graph_co_rdunlock();
out:
blk_co_unref(blk);
bdrv_co_unref(bs);
@@ -4052,6 +4313,10 @@ qcow2_co_create_opts(BlockDriver *drv, const char *filename, QemuOpts *opts,
{ BLOCK_OPT_COMPAT_LEVEL, "version" },
{ BLOCK_OPT_DATA_FILE_RAW, "data-file-raw" },
{ BLOCK_OPT_COMPRESSION_TYPE, "compression-type" },
+ { BLOCK_OPT_CONVENTIONAL_ZONES, "zone.conventional-zones" },
+ { BLOCK_OPT_MAX_OPEN_ZONES, "zone.max-open-zones" },
+ { BLOCK_OPT_MAX_ACTIVE_ZONES, "zone.max-active-zones" },
+ { BLOCK_OPT_MAX_APPEND_BYTES, "zone.max-append-bytes" },
{ NULL, NULL },
};
@@ -5442,6 +5707,29 @@ qcow2_get_specific_info(BlockDriverState *bs, Error **errp)
.data_file_raw = data_file_is_raw(bs),
.compression_type = s->compression_type,
};
+ if (s->zoned_header.zoned == QCOW2_Z_HM) {
+ ImageInfoSpecificQCow2Zoned *z =
+ g_new0(ImageInfoSpecificQCow2Zoned, 1);
+ *z = (ImageInfoSpecificQCow2Zoned){
+ .mode = QCOW2_ZONE_MODEL_HOST_MANAGED,
+ .nr_zones = s->zoned_header.nr_zones,
+ .u.host_managed = {
+ .has_size = true,
+ .size = s->zoned_header.zone_size,
+ .has_capacity = true,
+ .capacity = s->zoned_header.zone_capacity,
+ .has_conventional_zones = true,
+ .conventional_zones = s->zoned_header.conventional_zones,
+ .has_max_open_zones = true,
+ .max_open_zones = s->zoned_header.max_open_zones,
+ .has_max_active_zones = true,
+ .max_active_zones = s->zoned_header.max_active_zones,
+ .has_max_append_bytes = true,
+ .max_append_bytes = s->zoned_header.max_append_bytes,
+ },
+ };
+ spec_info->u.qcow2.data->zone = z;
+ }
} else {
/* if this assertion fails, this probably means a new version was
* added without having it covered here */
@@ -6265,6 +6553,41 @@ static QemuOptsList qcow2_create_opts = {
.type = QEMU_OPT_BOOL, \
.help = "Assume the external data file already exists and " \
"do not overwrite it" \
+ }, \
+ { \
+ .name = BLOCK_OPT_ZONE_MODEL, \
+ .type = QEMU_OPT_STRING, \
+ .help = "zone model modes, mode choice: host-managed", \
+ }, \
+ { \
+ .name = BLOCK_OPT_ZONE_SIZE, \
+ .type = QEMU_OPT_SIZE, \
+ .help = "zone size", \
+ }, \
+ { \
+ .name = BLOCK_OPT_ZONE_CAPACITY, \
+ .type = QEMU_OPT_SIZE, \
+ .help = "zone capacity", \
+ }, \
+ { \
+ .name = BLOCK_OPT_CONVENTIONAL_ZONES, \
+ .type = QEMU_OPT_NUMBER, \
+ .help = "numbers of conventional zones", \
+ }, \
+ { \
+ .name = BLOCK_OPT_MAX_APPEND_BYTES, \
+ .type = QEMU_OPT_SIZE, \
+ .help = "max append bytes", \
+ }, \
+ { \
+ .name = BLOCK_OPT_MAX_ACTIVE_ZONES, \
+ .type = QEMU_OPT_NUMBER, \
+ .help = "max active zones", \
+ }, \
+ { \
+ .name = BLOCK_OPT_MAX_OPEN_ZONES, \
+ .type = QEMU_OPT_NUMBER, \
+ .help = "max open zones", \
},
QCOW_COMMON_OPTIONS,
{ /* end of list */ }
diff --git a/block/qcow2.h b/block/qcow2.h
index 192a45d596..66474a686f 100644
--- a/block/qcow2.h
+++ b/block/qcow2.h
@@ -129,6 +129,12 @@
#define DEFAULT_CLUSTER_SIZE 65536
+#define DEFAULT_ZONE_SIZE (256 * MiB)
+#define DEFAULT_ZONE_MAX_APPEND_BYTES (64 * KiB)
+
+#define QCOW2_Z_NONE 0 /* Regular block device */
+#define QCOW2_Z_HM 1 /* Host-managed zoned block device */
+
#define QCOW2_OPT_DATA_FILE "data-file"
#define QCOW2_OPT_LAZY_REFCOUNTS "lazy-refcounts"
#define QCOW2_OPT_DISCARD_REQUEST "pass-discard-request"
@@ -237,6 +243,60 @@ typedef struct Qcow2CryptoHeaderExtension {
uint64_t length;
} QEMU_PACKED Qcow2CryptoHeaderExtension;
+typedef struct Qcow2ZonedHeaderExtension {
+ /* Zoned device attributes */
+ uint8_t zoned;
+ uint8_t reserved[3];
+ uint32_t nr_zones;
+ uint64_t zone_size;
+ uint64_t zone_capacity;
+ uint32_t conventional_zones;
+ uint32_t max_active_zones;
+ uint32_t max_open_zones;
+ uint32_t max_append_bytes;
+ uint64_t zonedmeta_offset;
+} QEMU_PACKED Qcow2ZonedHeaderExtension;
+
+typedef struct Qcow2ZoneListEntry {
+ QTAILQ_ENTRY(Qcow2ZoneListEntry) exp_open_zone_entry;
+ QTAILQ_ENTRY(Qcow2ZoneListEntry) imp_open_zone_entry;
+ QTAILQ_ENTRY(Qcow2ZoneListEntry) closed_zone_entry;
+} Qcow2ZoneListEntry;
+
+/*
+ * Per-zone tracking for write pointer (WP) advance. A submitter
+ * reserves an LBA range, performs the data write outside the zone
+ * lock, then waits until the persisted WP catches up to its end
+ * offset before returning success to the guest. A failure aborts
+ * all higher-LBA peers in the same zone.
+ */
+typedef enum Qcow2WPReqState {
+ /* data write submitted, awaiting completion */
+ QCOW2_WP_REQ_INFLIGHT,
+ /* data write completed, persisted WP not yet advanced past lba + len */
+ QCOW2_WP_REQ_PENDING,
+ /* persisted WP covers this range, submitter may ack success */
+ QCOW2_WP_REQ_RESOLVED,
+ /* data write failed, or a lower-LBA peer in the same zone failed */
+ QCOW2_WP_REQ_ABORTED,
+} Qcow2WPReqState;
+
+typedef struct Qcow2WPReq {
+ uint64_t lba; /* start offset of the data write */
+ uint64_t len; /* length in bytes */
+ Qcow2WPReqState state;
+ /* submitter waits for the data write to reach a terminal state */
+ CoQueue wait;
+ QTAILQ_ENTRY(Qcow2WPReq) entry;
+} Qcow2WPReq;
+
+typedef struct Qcow2ZoneWPState {
+ /* INFLIGHT reqs */
+ QTAILQ_HEAD(, Qcow2WPReq) in_flight;
+ /* PENDING reqs, sorted by lba */
+ QTAILQ_HEAD(, Qcow2WPReq) completed_pending;
+} Qcow2ZoneWPState;
+
typedef struct Qcow2UnknownHeaderExtension {
uint32_t magic;
uint32_t len;
@@ -257,17 +317,20 @@ enum {
QCOW2_INCOMPAT_DATA_FILE_BITNR = 2,
QCOW2_INCOMPAT_COMPRESSION_BITNR = 3,
QCOW2_INCOMPAT_EXTL2_BITNR = 4,
+ QCOW2_INCOMPAT_ZONED_FORMAT_BITNR = 5,
QCOW2_INCOMPAT_DIRTY = 1 << QCOW2_INCOMPAT_DIRTY_BITNR,
QCOW2_INCOMPAT_CORRUPT = 1 << QCOW2_INCOMPAT_CORRUPT_BITNR,
QCOW2_INCOMPAT_DATA_FILE = 1 << QCOW2_INCOMPAT_DATA_FILE_BITNR,
QCOW2_INCOMPAT_COMPRESSION = 1 << QCOW2_INCOMPAT_COMPRESSION_BITNR,
QCOW2_INCOMPAT_EXTL2 = 1 << QCOW2_INCOMPAT_EXTL2_BITNR,
+ QCOW2_INCOMPAT_ZONED_FORMAT = 1 << QCOW2_INCOMPAT_ZONED_FORMAT_BITNR,
QCOW2_INCOMPAT_MASK = QCOW2_INCOMPAT_DIRTY
| QCOW2_INCOMPAT_CORRUPT
| QCOW2_INCOMPAT_DATA_FILE
| QCOW2_INCOMPAT_COMPRESSION
- | QCOW2_INCOMPAT_EXTL2,
+ | QCOW2_INCOMPAT_EXTL2
+ | QCOW2_INCOMPAT_ZONED_FORMAT,
};
/* Compatible feature bits */
@@ -426,6 +489,24 @@ typedef struct BDRVQcow2State {
* is to convert the image with the desired compression type set.
*/
Qcow2CompressionType compression_type;
+
+ /* States of zoned device */
+ Qcow2ZonedHeaderExtension zoned_header;
+ QTAILQ_HEAD(, Qcow2ZoneListEntry) exp_open_zones;
+ QTAILQ_HEAD(, Qcow2ZoneListEntry) imp_open_zones;
+ QTAILQ_HEAD(, Qcow2ZoneListEntry) closed_zones;
+ Qcow2ZoneListEntry *zone_list_entries;
+ uint32_t nr_zones_exp_open;
+ uint32_t nr_zones_imp_open;
+ uint32_t nr_zones_closed;
+
+ /* Per-zone in-memory WP tracking for zoned imgaes. */
+ Qcow2ZoneWPState *zone_wp_state;
+ /*
+ * Cache holding the on-disk zonedmeta cluster(s) for zoned image.
+ * WP advances go through this cache.
+ */
+ Qcow2Cache *wp_cache;
} BDRVQcow2State;
typedef struct Qcow2COWRegion {
diff --git a/docs/interop/qcow2.rst b/docs/interop/qcow2.rst
index 5948591107..1e2689b99e 100644
--- a/docs/interop/qcow2.rst
+++ b/docs/interop/qcow2.rst
@@ -128,7 +128,26 @@ the next fields through ``header_length``.
allows subcluster-based allocation. See the
Extended L2 Entries section for more details.
- Bits 5-63: Reserved (set to 0)
+ Bit 5: Zoned extension bit. If this bit is set then
+ the file is an emulated zoned device. The
+ zoned extension must be present.
+ Implementations that do not support zoned
+ emulation cannot open this file because it
+ generally only make sense to interpret the
+ data along with the zone information and
+ write pointers.
+
+ It is unsafe when any qcow2 user without
+ knowing the zoned extension reads or edits
+ a file with the zoned extension. The write
+ pointer tracking can be corrupted when a
+ writer edits a file, like overwriting beyond
+ the write pointer locations. Or a reader tries
+ to access a file without knowing write
+ pointers where the software setup will cause
+ invalid reads.
+
+ Bits 6-63: Reserved (set to 0)
80 - 87: compatible_features
Bitmask of compatible features. An implementation can
@@ -259,6 +278,7 @@ be stored. Each extension has a structure like the following::
0x23852875 - Bitmaps extension
0x0537be77 - Full disk encryption header pointer
0x44415441 - External data file name string
+ 0x007a6264 - Zoned extension
other - Unknown header extension, can be safely
ignored
@@ -344,6 +364,101 @@ The fields of the bitmaps extension are::
Offset into the image file at which the bitmap directory
starts. Must be aligned to a cluster boundary.
+Zoned extension
+---------------
+
+The zoned extension must be present if the incompatible_features bit 5 is set,
+and omitted when it is clear. It contains fields for emulating the zoned
+storage model (https://zonedstorage.io/). Currently only the host-managed zone
+model is supported.
+
+The fields of the zoned extension are::
+
+ Byte 0: zoned
+ The byte represents the zoned model of the device. 0 is for
+ a non-zoned device (all other information in this header
+ is ignored). 1 is for a host-managed device, which only
+ allows for sequential writes within each zone. Other
+ values may be added later, the implementation must refuse
+ to open a device containing an unknown zone model.
+
+ 1 - 3: Reserved, must be zero.
+
+ 4 - 7: nr_zones
+ The number of zones. It is the sum of conventional zones
+ and sequential zones. The maximum value for nr_zones is
+ (2^32 - 1)/8 = 536870911.
+
+ 8 - 15: zone_size
+ Total size of each zone, in bytes. The 64-bit field is to
+ satisfy the virtio-blk zone_size range and emulate a read
+ zoned device, whose maximum zone size can be as large as
+ 2TB.
+
+ The value must be power of 2. Linux currently requires
+ the zone size to be a power of 2 number of LBAs. Qcow2
+ following this is mainly to allow emulating a real
+ ZNS drive configuration. It is not relevant to the cluster
+ size.
+
+ 16 - 23: zone_capacity
+ The number of writable bytes within the zones. The bytes
+ between zone capacity and zone size are unusable: reads
+ will return 0s and writes will fail.
+
+ A zone capacity is always smaller or equal to the zone
+ size. It is for emulating a real ZNS drive configuration,
+ which has the constraint of aligning to some hardware erase
+ block size.
+
+ 24 - 27: conventional_zones
+ The number of conventional zones. The conventional zones
+ allow sequential writes and random writes. While the
+ sequential zones only allow sequential writes.
+
+ 28 - 31: max_active_zones
+ The number of the zones that can be in the implicit open,
+ explicit open or closed state. It cannot be larger than
+ nr_zones.
+
+ 32 - 35: max_open_zones
+ The maximal number of open (implicitly open or explicitly
+ open) zones. It cannot be larger than the number of SWR
+ zones of the device, nor larger than max_active_zones.
+
+ If the limits of open zones or active zones are equal to
+ the total number of SWR zones, then it's the same as having
+ no limits therefore max open zones and max active zones are
+ set to 0.
+
+ 36 - 39: max_append_bytes
+ The number of bytes of a zone append request that can be
+ issued to the device. It must be 512-byte aligned and less
+ than the zone capacity.
+
+ 40 - 47: zonedmeta_offset
+ The offset of zoned metadata structure in the contained
+ image, in bytes.
+
+The zonedmeta clusters contain a table of the zone write pointers. Each entry
+is a 64-bit value encoding both the zone type and the zone's write pointer::
+
+ Bit 0 - 58: Write pointer offset, in bytes, indicates the starting
+ point of the next write position in that zone.
+ For conventional zones the value is the zone's starting
+ offset.
+
+ 59: Zone type. 0 = SWR; 1 = conventional.
+
+ 60 - 63: Reserved, must be zero.
+
+
+Each zone's write pointer in the zonedmeta area is durable on disk only after
+the data it covers is durable. For a write to a SWR zone, the data is written
+back to disk first, then the non-zoned qcow2 L1/L2/refcount metadata is written,
+and finally the write pointers are written. This ordering ensures that after
+a power failure the on-disk write pointer never leads data being written.
+
Full disk encryption header pointer
-----------------------------------
diff --git a/include/block/block_int-common.h b/include/block/block_int-common.h
index 147c08155f..349876bc65 100644
--- a/include/block/block_int-common.h
+++ b/include/block/block_int-common.h
@@ -57,6 +57,13 @@
#define BLOCK_OPT_COMPRESSION_TYPE "compression_type"
#define BLOCK_OPT_EXTL2 "extended_l2"
#define BLOCK_OPT_KEEP_DATA_FILE "keep_data_file"
+#define BLOCK_OPT_ZONE_MODEL "zone.mode"
+#define BLOCK_OPT_ZONE_SIZE "zone.size"
+#define BLOCK_OPT_ZONE_CAPACITY "zone.capacity"
+#define BLOCK_OPT_CONVENTIONAL_ZONES "zone.conventional_zones"
+#define BLOCK_OPT_MAX_APPEND_BYTES "zone.max_append_bytes"
+#define BLOCK_OPT_MAX_ACTIVE_ZONES "zone.max_active_zones"
+#define BLOCK_OPT_MAX_OPEN_ZONES "zone.max_open_zones"
#define BLOCK_PROBE_BUF_SIZE 512
@@ -901,7 +908,13 @@ typedef struct BlockLimits {
BlockZoneModel zoned;
/* zone size expressed in bytes */
- uint32_t zone_size;
+ uint64_t zone_size;
+
+ /*
+ * The number of usable logical blocks within the zone, expressed
+ * in bytes. A zone capacity is smaller or equal to the zone size.
+ */
+ uint64_t zone_capacity;
/* total number of zones */
uint32_t nr_zones;
diff --git a/qapi/block-core.json b/qapi/block-core.json
index 1f87b07850..4542bfc8f5 100644
--- a/qapi/block-core.json
+++ b/qapi/block-core.json
@@ -93,6 +93,9 @@
#
# @compression-type: the image cluster compression method (since 5.1)
#
+# @zone: zoned configuration; only set when the image was created
+# with a zoned model (since 11.1)
+#
# Since: 1.7
##
{ 'struct': 'ImageInfoSpecificQCow2',
@@ -106,7 +109,8 @@
'refcount-bits': 'int',
'*encrypt': 'ImageInfoSpecificQCow2Encryption',
'*bitmaps': ['Qcow2BitmapInfo'],
- 'compression-type': 'Qcow2CompressionType'
+ 'compression-type': 'Qcow2CompressionType',
+ '*zone': 'ImageInfoSpecificQCow2Zoned'
} }
##
@@ -5221,6 +5225,85 @@
{ 'enum': 'Qcow2CompressionType',
'data': [ 'zlib', { 'name': 'zstd', 'if': 'CONFIG_ZSTD' } ] }
+##
+# @Qcow2ZoneModel:
+#
+# Zoned device model used in qcow2 image file
+#
+# @host-managed: The host-managed model only allows sequential write
+# over the device zones.
+#
+# Since: 11.1
+##
+{ 'enum': 'Qcow2ZoneModel',
+ 'data': [ 'host-managed'] }
+
+##
+# @Qcow2ZoneHostManaged:
+#
+# The host-managed zone model. It only allows sequential writes.
+#
+# @size: Total number of bytes within zones (default: 256 MB).
+#
+# @capacity: The usable space within each zone, in bytes. Always
+# smaller than or equal to @size (default: same as @size).
+#
+# @conventional-zones: The number of conventional zones of the
+# zoned device (default: 0).
+#
+# @max-open-zones: The maximal number of open zones. Must be less
+# than or equal to the number of non-conventional zones (i.e.
+# ``size / zone.size - @conventional-zones``), and less than or
+# equal to @max-active-zones (default: 0, meaning no limit).
+#
+# @max-active-zones: The maximal number of zones in the implicit
+# open, explicit open or closed state. It is less than or equal
+# to the number of zones (default: 0, meaning no limit).
+#
+# @max-append-bytes: The maximal size in bytes of a zone-append
+# request that can be issued to the device. Must be a multiple
+# of 512, and less than @capacity (default: 64 KB).
+#
+# Since: 11.1
+##
+{ 'struct': 'Qcow2ZoneHostManaged',
+ 'data': { '*size': 'size',
+ '*capacity': 'size',
+ '*conventional-zones': 'uint32',
+ '*max-open-zones': 'uint32',
+ '*max-active-zones': 'uint32',
+ '*max-append-bytes': 'size' } }
+
+##
+# @Qcow2ZoneCreateOptions:
+#
+# Creation options for zoned qcow2 images.
+#
+# @mode: The zone model used by the image.
+#
+# Since: 11.1
+##
+{ 'union': 'Qcow2ZoneCreateOptions',
+ 'base': { 'mode': 'Qcow2ZoneModel' },
+ 'discriminator': 'mode',
+ 'data': { 'host-managed': 'Qcow2ZoneHostManaged' } }
+
+##
+# @ImageInfoSpecificQCow2Zoned:
+#
+# @mode: The zone model used by the image.
+#
+# @nr-zones: total number of zones in the image, derived at create
+# time from the size and the zone size.
+#
+# Since: 11.1
+##
+{ 'union': 'ImageInfoSpecificQCow2Zoned',
+ 'base': { 'mode': 'Qcow2ZoneModel',
+ 'nr-zones': 'uint32' },
+ 'discriminator': 'mode',
+ 'data': { 'host-managed': 'Qcow2ZoneHostManaged' } }
+
##
# @BlockdevCreateOptionsQcow2:
#
@@ -5263,6 +5346,9 @@
# @compression-type: The image cluster compression method
# (default: zlib, since 5.1)
#
+# @zone: Options for zoned images. If absent, the device is not
+# zoned. (since 11.1)
+#
# Since: 2.12
##
{ 'struct': 'BlockdevCreateOptionsQcow2',
@@ -5279,7 +5365,8 @@
'*preallocation': 'PreallocMode',
'*lazy-refcounts': 'bool',
'*refcount-bits': 'int',
- '*compression-type':'Qcow2CompressionType' } }
+ '*compression-type':'Qcow2CompressionType',
+ '*zone': 'Qcow2ZoneCreateOptions' } }
##
# @BlockdevCreateOptionsQed:
--
2.53.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* Re: [PATCH v11 2/5] qcow2: add configurations for zoned format extension
2026-06-01 21:44 ` [PATCH v11 2/5] qcow2: add configurations for zoned format extension Sam Li
@ 2026-06-02 20:03 ` Stefan Hajnoczi
2026-06-03 7:37 ` Markus Armbruster
1 sibling, 0 replies; 13+ messages in thread
From: Stefan Hajnoczi @ 2026-06-02 20:03 UTC (permalink / raw)
To: Sam Li
Cc: qemu-devel, Markus Armbruster, qemu-block, Eric Blake,
Pierrick Bouvier, dmitry.fomichev, Hanna Reitz, hare,
Michael S. Tsirkin, Kevin Wolf, cassel, dlemoal
[-- Attachment #1: Type: text/plain, Size: 13639 bytes --]
On Mon, Jun 01, 2026 at 11:44:02PM +0200, Sam Li wrote:
> To configure the zoned format feature on the qcow2 driver, it
> requires settings as: the device size, zone model, zone size,
> zone capacity, number of conventional zones, limits on zone
> resources (max append bytes, max open zones, and max_active_zones).
>
> To create a qcow2 image with zoned format feature, use command like
> this:
> qemu-img create -f qcow2 zbc.qcow2 -o size=768M \
> -o zone.size=64M -o zone.capacity=64M -o zone.conventional_zones=0 \
> -o zone.max_append_bytes=4096 -o zone.max_open_zones=6 \
> -o zone.max_active_zones=8 -o zone.mode=host-managed
>
> Signed-off-by: Sam Li <faithilikerun@gmail.com>
> ---
> block/file-posix.c | 2 +-
> block/qcow2.c | 329 ++++++++++++++++++++++++++++++-
> block/qcow2.h | 83 +++++++-
> docs/interop/qcow2.rst | 117 ++++++++++-
> include/block/block_int-common.h | 15 +-
> qapi/block-core.json | 91 ++++++++-
> 6 files changed, 628 insertions(+), 9 deletions(-)
>
> diff --git a/block/file-posix.c b/block/file-posix.c
> index e49b13d6ab..14278785b9 100644
> --- a/block/file-posix.c
> +++ b/block/file-posix.c
> @@ -3607,7 +3607,7 @@ raw_co_zone_append(BlockDriverState *bs,
>
> if (*offset & zone_size_mask) {
> error_report("sector offset %" PRId64 " is not aligned to zone size "
> - "%" PRId32 "", *offset / 512, bs->bl.zone_size / 512);
> + "%" PRId64 "", *offset / 512, bs->bl.zone_size / 512);
> return -EINVAL;
> }
>
> diff --git a/block/qcow2.c b/block/qcow2.c
> index 81fd299b4c..29eec33e34 100644
> --- a/block/qcow2.c
> +++ b/block/qcow2.c
> @@ -73,6 +73,7 @@ typedef struct {
> #define QCOW2_EXT_MAGIC_CRYPTO_HEADER 0x0537be77
> #define QCOW2_EXT_MAGIC_BITMAPS 0x23852875
> #define QCOW2_EXT_MAGIC_DATA_FILE 0x44415441
> +#define QCOW2_EXT_MAGIC_ZONED_FORMAT 0x007a6264
>
> static int coroutine_fn
> qcow2_co_preadv_compressed(BlockDriverState *bs,
> @@ -194,6 +195,93 @@ qcow2_extract_crypto_opts(QemuOpts *opts, const char *fmt, Error **errp)
> return cryptoopts_qdict;
> }
>
> +/*
> + * Passing by the zoned device configurations by a zoned_header struct, check
> + * if the zone device options are under constraints. Return false when some
> + * option is invalid
> + */
I have trouble parsing the first sentence. Maybe replace this doc
comment with "Returns true if zone_opt is valid, false otherwise"?
> +static bool
> +qcow2_check_zone_options(Qcow2ZonedHeaderExtension *zone_opt, Error **errp)
> +{
> + uint32_t sequential_zones;
> +
> + assert(zone_opt != NULL);
> +
> + if (zone_opt->zoned != QCOW2_Z_NONE && zone_opt->zoned != QCOW2_Z_HM) {
> + error_setg(errp, "Zoned extension header zoned field has unknown "
> + "value %" PRIu8, zone_opt->zoned);
> + return false;
> + }
> +
> + if (!is_power_of_2(zone_opt->zone_size)) {
> + error_setg(errp, "Zoned extension header zone_size %" PRIu64
> + "B is not a power of 2", zone_opt->zone_size);
> + return false;
> + }
> +
> + if (zone_opt->nr_zones > UINT32_MAX / 8) {
> + error_setg(errp, "Zoned extension header nr_zones %" PRIu32
> + " exceeds maximum %u",
> + zone_opt->nr_zones, UINT32_MAX / 8);
> + return false;
> + }
> +
> + if (zone_opt->zone_capacity > zone_opt->zone_size) {
> + error_setg(errp, "zone capacity %" PRIu64 "B exceeds zone size "
> + "%" PRIu64 "B", zone_opt->zone_capacity,
> + zone_opt->zone_size);
> + return false;
> + }
> +
> + if (!QEMU_IS_ALIGNED(zone_opt->max_append_bytes, BDRV_SECTOR_SIZE)) {
> + error_setg(errp, "max append bytes %" PRIu32 "B is not aligned "
> + "to %" PRIu64, zone_opt->max_append_bytes,
> + (uint64_t)BDRV_SECTOR_SIZE);
> + return false;
> + }
> +
> + if (zone_opt->max_append_bytes + BDRV_SECTOR_SIZE >=
> + zone_opt->zone_capacity) {
> + error_setg(errp, "max append bytes %" PRIu32 "B exceeds zone "
> + "capacity %" PRIu64 "B by more than block size",
> + zone_opt->max_append_bytes, zone_opt->zone_capacity);
> + return false;
> + }
> +
> + if (zone_opt->conventional_zones >= zone_opt->nr_zones) {
> + error_setg(errp, "Conventional_zones %" PRIu32 " exceeds "
> + "nr_zones %" PRIu32 ".",
> + zone_opt->conventional_zones, zone_opt->nr_zones);
> + return false;
> + }
> +
> + if (zone_opt->max_active_zones > zone_opt->nr_zones) {
> + error_setg(errp, "Max_active_zones %" PRIu32 " exceeds "
> + "nr_zones %" PRIu32 ". Set it to nr_zones.",
> + zone_opt->max_active_zones, zone_opt->nr_zones);
> + zone_opt->max_active_zones = zone_opt->nr_zones;
> + }
errp is an error that must be handled by the caller, but the function
will return true. The caller may not notice the Error object and it may
be leaked.
Another issue with this approach is that error_setg(errp, "a");
error_setg(errp, "b"); is not allowed (see assert(*errp == NULL) in
error_setv()). So if two of these conditionals are taken then QEMU will
abort.
I think the intention is to print a warning rather than to set an error
here. Use warn_report() without touching errp to avoid these issues.
> +
> + sequential_zones = zone_opt->nr_zones - zone_opt->conventional_zones;
> + if (zone_opt->max_open_zones > sequential_zones) {
> + error_setg(errp, "Max_open_zones field can not be larger than"
> + "the number of SWR zones. Set it to number of SWR"
> + "zones %" PRIu32 ".", sequential_zones);
> + zone_opt->max_open_zones = sequential_zones;
> + }
> + if (zone_opt->max_active_zones != 0 &&
> + zone_opt->max_open_zones > zone_opt->max_active_zones) {
> + error_setg(errp, "Max_open_zones %" PRIu32 " exceeds "
> + "max_active_zones %" PRIu32 ". Set it to "
> + "max_active_zones.",
> + zone_opt->max_open_zones,
> + zone_opt->max_active_zones);
> + zone_opt->max_open_zones = zone_opt->max_active_zones;
> + }
> +
> + return true;
> +}
> +
> /*
> * read qcow2 extension and fill bs
> * start reading from start_offset
> @@ -211,6 +299,7 @@ qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset,
> uint64_t offset;
> int ret;
> Qcow2BitmapHeaderExt bitmaps_ext;
> + Qcow2ZonedHeaderExtension zoned_ext;
>
> if (need_update_header != NULL) {
> *need_update_header = false;
> @@ -432,6 +521,50 @@ qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset,
> break;
> }
>
> + case QCOW2_EXT_MAGIC_ZONED_FORMAT:
> + {
> + if (ext.len != sizeof(zoned_ext)) {
> + error_setg(errp, "zoned_ext: unexpected len=%" PRIu32 " "
> + "(expected %zu)", ext.len, sizeof(zoned_ext));
> + return -EINVAL;
> + }
> + ret = bdrv_pread(bs->file, offset, ext.len, &zoned_ext, 0);
> + if (ret < 0) {
> + error_setg_errno(errp, -ret, "zoned_ext: "
> + "Could not read ext header");
> + return ret;
> + }
> +
> + zoned_ext.zone_size = be64_to_cpu(zoned_ext.zone_size);
> + zoned_ext.zone_capacity = be64_to_cpu(zoned_ext.zone_capacity);
> + zoned_ext.conventional_zones =
> + be32_to_cpu(zoned_ext.conventional_zones);
> + zoned_ext.nr_zones = be32_to_cpu(zoned_ext.nr_zones);
> + zoned_ext.max_open_zones = be32_to_cpu(zoned_ext.max_open_zones);
> + zoned_ext.max_active_zones =
> + be32_to_cpu(zoned_ext.max_active_zones);
> + zoned_ext.max_append_bytes =
> + be32_to_cpu(zoned_ext.max_append_bytes);
> + s->zoned_header = zoned_ext;
> +
> + /* refuse to open broken images */
> + if (zoned_ext.nr_zones != DIV_ROUND_UP(bs->total_sectors *
> + BDRV_SECTOR_SIZE, zoned_ext.zone_size)) {
This is not safe when validating untrusted qcow2 files because
zoned_ext.zone_size could be 0 (division-by-zero kills the process) and
bs->total_sectors * BDRV_SECTOR_SIZE can integer overflow (unlikely
because the file would have to be extremely large).
Can you express this in a way that is safe? If necessary, split it into
multiple checks.
> + error_setg(errp, "Zoned extension header nr_zones field "
> + "is wrong");
> + return -EINVAL;
> + }
> + if (!qcow2_check_zone_options(&zoned_ext, errp)) {
> + return -EINVAL;
> + }
> +
> +#ifdef DEBUG_EXT
> + printf("Qcow2: Got zoned format extension: "
> + "offset=%" PRIu32 "\n", offset);
> +#endif
> + break;
> + }
> +
> default:
> /* unknown magic - save it in case we need to rewrite the header */
> /* If you add a new feature, make sure to also update the fast
> @@ -2068,6 +2201,25 @@ static void qcow2_refresh_limits(BlockDriverState *bs, Error **errp)
> }
> bs->bl.pwrite_zeroes_alignment = s->subcluster_size;
> bs->bl.pdiscard_alignment = s->cluster_size;
> +
> + switch (s->zoned_header.zoned) {
> + case QCOW2_Z_HM:
> + bs->bl.zoned = BLK_Z_HM;
> + break;
> + case QCOW2_Z_NONE:
> + default:
> + bs->bl.zoned = BLK_Z_NONE;
> + break;
> + }
> +
> + bs->bl.nr_zones = s->zoned_header.nr_zones;
> + bs->bl.max_append_sectors = s->zoned_header.max_append_bytes
> + >> BDRV_SECTOR_BITS;
> + bs->bl.max_active_zones = s->zoned_header.max_active_zones;
> + bs->bl.max_open_zones = s->zoned_header.max_open_zones;
> + bs->bl.zone_size = s->zoned_header.zone_size;
> + bs->bl.zone_capacity = s->zoned_header.zone_capacity;
> + bs->bl.write_granularity = BDRV_SECTOR_SIZE;
> }
>
> static int GRAPH_UNLOCKED
> @@ -3170,6 +3322,11 @@ int qcow2_update_header(BlockDriverState *bs)
> .bit = QCOW2_INCOMPAT_EXTL2_BITNR,
> .name = "extended L2 entries",
> },
> + {
> + .type = QCOW2_FEAT_TYPE_INCOMPATIBLE,
> + .bit = QCOW2_INCOMPAT_ZONED_FORMAT_BITNR,
> + .name = "zoned format",
> + },
> {
> .type = QCOW2_FEAT_TYPE_COMPATIBLE,
> .bit = QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR,
> @@ -3215,6 +3372,31 @@ int qcow2_update_header(BlockDriverState *bs)
> buflen -= ret;
> }
>
> + /* Zoned devices header extension */
> + if (s->zoned_header.zoned == QCOW2_Z_HM) {
> + Qcow2ZonedHeaderExtension zoned_header = {
> + .zoned = s->zoned_header.zoned,
> + .zone_size = cpu_to_be64(s->zoned_header.zone_size),
> + .zone_capacity = cpu_to_be64(s->zoned_header.zone_capacity),
> + .conventional_zones =
> + cpu_to_be32(s->zoned_header.conventional_zones),
> + .nr_zones = cpu_to_be32(s->zoned_header.nr_zones),
> + .max_open_zones = cpu_to_be32(s->zoned_header.max_open_zones),
> + .max_active_zones =
> + cpu_to_be32(s->zoned_header.max_active_zones),
> + .max_append_bytes =
> + cpu_to_be32(s->zoned_header.max_append_bytes)
> + };
> + ret = header_ext_add(buf, QCOW2_EXT_MAGIC_ZONED_FORMAT,
> + &zoned_header, sizeof(zoned_header),
> + buflen);
> + if (ret < 0) {
> + goto fail;
> + }
> + buf += ret;
> + buflen -= ret;
> + }
> +
> /* Keep unknown header extensions */
> QLIST_FOREACH(uext, &s->unknown_header_ext, next) {
> ret = header_ext_add(buf, uext->magic, uext->data, uext->len, buflen);
> @@ -3589,6 +3771,8 @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp)
> ERRP_GUARD();
> BlockdevCreateOptionsQcow2 *qcow2_opts;
> QDict *options;
> + Qcow2ZoneCreateOptions *zone_struct;
> + Qcow2ZoneHostManaged *zone_host_managed;
>
> /*
> * Open the image file and write a minimal qcow2 header.
> @@ -3615,6 +3799,8 @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp)
>
> assert(create_options->driver == BLOCKDEV_DRIVER_QCOW2);
> qcow2_opts = &create_options->u.qcow2;
> + zone_struct = create_options->u.qcow2.zone;
A slightly more concise way of writing this:
zone_struct = qcow2_opts->zone;
> + zone_host_managed = &create_options->u.qcow2.zone->u.host_managed;
This must be moved down to avoid deferencing a NULL
create_options->u.qcow2.zone pointer:
if (zone_struct && zone_struct->mode == QCOW2_ZONE_MODEL_HOST_MANAGED) {
Qcow2ZoneHostManaged *zone_host_managed = &zone_struct->u.host_managed;
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: [PATCH v11 2/5] qcow2: add configurations for zoned format extension
2026-06-01 21:44 ` [PATCH v11 2/5] qcow2: add configurations for zoned format extension Sam Li
2026-06-02 20:03 ` Stefan Hajnoczi
@ 2026-06-03 7:37 ` Markus Armbruster
1 sibling, 0 replies; 13+ messages in thread
From: Markus Armbruster @ 2026-06-03 7:37 UTC (permalink / raw)
To: Sam Li
Cc: qemu-devel, qemu-block, Eric Blake, Stefan Hajnoczi,
Pierrick Bouvier, dmitry.fomichev, Hanna Reitz, hare,
Michael S. Tsirkin, Kevin Wolf, cassel, dlemoal
Sam Li <faithilikerun@gmail.com> writes:
> To configure the zoned format feature on the qcow2 driver, it
> requires settings as: the device size, zone model, zone size,
> zone capacity, number of conventional zones, limits on zone
> resources (max append bytes, max open zones, and max_active_zones).
>
> To create a qcow2 image with zoned format feature, use command like
> this:
> qemu-img create -f qcow2 zbc.qcow2 -o size=768M \
> -o zone.size=64M -o zone.capacity=64M -o zone.conventional_zones=0 \
> -o zone.max_append_bytes=4096 -o zone.max_open_zones=6 \
> -o zone.max_active_zones=8 -o zone.mode=host-managed
>
> Signed-off-by: Sam Li <faithilikerun@gmail.com>
QAPI schema
Acked-by: Markus Armbruster <armbru@redhat.com>
^ permalink raw reply [flat|nested] 13+ messages in thread
* [PATCH v11 3/5] virtio-blk: do not merge writes across a zone boundary
2026-06-01 21:44 [PATCH v11 0/5] Add full zoned storage emulation to the qcow2 driver Sam Li
2026-06-01 21:44 ` [PATCH v11 1/5] docs/qcow2: add the zoned format feature Sam Li
2026-06-01 21:44 ` [PATCH v11 2/5] qcow2: add configurations for zoned format extension Sam Li
@ 2026-06-01 21:44 ` Sam Li
2026-06-03 20:41 ` Stefan Hajnoczi
2026-06-01 21:44 ` [PATCH v11 4/5] qcow2: add zoned emulation capability Sam Li
` (2 subsequent siblings)
5 siblings, 1 reply; 13+ messages in thread
From: Sam Li @ 2026-06-01 21:44 UTC (permalink / raw)
To: qemu-devel
Cc: Markus Armbruster, qemu-block, Eric Blake, Stefan Hajnoczi,
Pierrick Bouvier, dmitry.fomichev, Hanna Reitz, hare,
Michael S. Tsirkin, Kevin Wolf, cassel, dlemoal, Sam Li
virtio_blk_submit_multireq() fuses adjacent in-zone writes into a
single request. On a zoned backend, a merged zone append request
that straddles a zone boundary is rejected by the device because
each write must stay within a single zone.
Add a bail condition to the merge coalescer: if combining the
candidate request into the current batch would cross a zone
boundary, flush the current batch and start a new one.
Signed-off-by: Sam Li <faithilikerun@gmail.com>
---
hw/block/virtio-blk.c | 22 +++++++++++++++++++++-
1 file changed, 21 insertions(+), 1 deletion(-)
diff --git a/hw/block/virtio-blk.c b/hw/block/virtio-blk.c
index 9cb9f1fb2b..285db19ac7 100644
--- a/hw/block/virtio-blk.c
+++ b/hw/block/virtio-blk.c
@@ -288,6 +288,9 @@ static void virtio_blk_submit_multireq(VirtIOBlock *s, MultiReqBuffer *mrb)
int i = 0, start = 0, num_reqs = 0, niov = 0, nb_sectors = 0;
uint32_t max_transfer;
int64_t sector_num = 0;
+ BlockDriverState *bs = blk_bs(s->blk);
+ bool zone_cross;
+ int64_t zone_sector, end_sector;
if (mrb->num_reqs == 1) {
submit_requests(s, mrb, 0, 1, -1);
@@ -303,17 +306,34 @@ static void virtio_blk_submit_multireq(VirtIOBlock *s, MultiReqBuffer *mrb)
for (i = 0; i < mrb->num_reqs; i++) {
VirtIOBlockReq *req = mrb->reqs[i];
if (num_reqs > 0) {
+ zone_cross = false;
+
+ /*
+ * On zoned backends, a single backend write must not span a zone
+ * boundary. Bail out of merging if combining req into the current
+ * batch would straddle a zone.
+ */
+ if (bs && bs->bl.zone_size > 0) {
+ zone_sector = bs->bl.zone_size / BDRV_SECTOR_SIZE;
+ end_sector = req->sector_num
+ + req->qiov.size / BDRV_SECTOR_SIZE - 1;
+ zone_cross = (sector_num / zone_sector) !=
+ (end_sector / zone_sector);
+ }
+
/*
* NOTE: We cannot merge the requests in below situations:
* 1. requests are not sequential
* 2. merge would exceed maximum number of IOVs
* 3. merge would exceed maximum transfer length of backend device
+ * 4. merge would cross a zone boundary on a zoned backend
*/
if (sector_num + nb_sectors != req->sector_num ||
niov > blk_get_max_iov(s->blk) - req->qiov.niov ||
req->qiov.size > max_transfer ||
nb_sectors > (max_transfer -
- req->qiov.size) / BDRV_SECTOR_SIZE) {
+ req->qiov.size) / BDRV_SECTOR_SIZE ||
+ zone_cross) {
submit_requests(s, mrb, start, num_reqs, niov);
num_reqs = 0;
}
--
2.53.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* Re: [PATCH v11 3/5] virtio-blk: do not merge writes across a zone boundary
2026-06-01 21:44 ` [PATCH v11 3/5] virtio-blk: do not merge writes across a zone boundary Sam Li
@ 2026-06-03 20:41 ` Stefan Hajnoczi
0 siblings, 0 replies; 13+ messages in thread
From: Stefan Hajnoczi @ 2026-06-03 20:41 UTC (permalink / raw)
To: Sam Li
Cc: qemu-devel, Markus Armbruster, qemu-block, Eric Blake,
Pierrick Bouvier, dmitry.fomichev, Hanna Reitz, hare,
Michael S. Tsirkin, Kevin Wolf, cassel, dlemoal
[-- Attachment #1: Type: text/plain, Size: 1362 bytes --]
On Mon, Jun 01, 2026 at 11:44:03PM +0200, Sam Li wrote:
> diff --git a/hw/block/virtio-blk.c b/hw/block/virtio-blk.c
> index 9cb9f1fb2b..285db19ac7 100644
> --- a/hw/block/virtio-blk.c
> +++ b/hw/block/virtio-blk.c
> @@ -288,6 +288,9 @@ static void virtio_blk_submit_multireq(VirtIOBlock *s, MultiReqBuffer *mrb)
> int i = 0, start = 0, num_reqs = 0, niov = 0, nb_sectors = 0;
> uint32_t max_transfer;
> int64_t sector_num = 0;
> + BlockDriverState *bs = blk_bs(s->blk);
Please add a blk_get_zone_size() API in block-backend.c along the lines
of the other accessor APIs (blk_get_request_alignment(), etc) instead of
using bs directly.
> + bool zone_cross;
> + int64_t zone_sector, end_sector;
>
> if (mrb->num_reqs == 1) {
> submit_requests(s, mrb, 0, 1, -1);
> @@ -303,17 +306,34 @@ static void virtio_blk_submit_multireq(VirtIOBlock *s, MultiReqBuffer *mrb)
> for (i = 0; i < mrb->num_reqs; i++) {
> VirtIOBlockReq *req = mrb->reqs[i];
> if (num_reqs > 0) {
> + zone_cross = false;
> +
> + /*
> + * On zoned backends, a single backend write must not span a zone
This code handles both reads and writes but the comment only mentions
writes. I think reads are allowed to span zones, so should there be a
check that mrb->is_write?
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]
^ permalink raw reply [flat|nested] 13+ messages in thread
* [PATCH v11 4/5] qcow2: add zoned emulation capability
2026-06-01 21:44 [PATCH v11 0/5] Add full zoned storage emulation to the qcow2 driver Sam Li
` (2 preceding siblings ...)
2026-06-01 21:44 ` [PATCH v11 3/5] virtio-blk: do not merge writes across a zone boundary Sam Li
@ 2026-06-01 21:44 ` Sam Li
2026-06-04 20:51 ` Stefan Hajnoczi
2026-06-01 21:44 ` [PATCH v11 5/5] iotests: test the zoned format feature for qcow2 file Sam Li
2026-06-02 20:06 ` [PATCH v11 0/5] Add full zoned storage emulation to the qcow2 driver Stefan Hajnoczi
5 siblings, 1 reply; 13+ messages in thread
From: Sam Li @ 2026-06-01 21:44 UTC (permalink / raw)
To: qemu-devel
Cc: Markus Armbruster, qemu-block, Eric Blake, Stefan Hajnoczi,
Pierrick Bouvier, dmitry.fomichev, Hanna Reitz, hare,
Michael S. Tsirkin, Kevin Wolf, cassel, dlemoal, Sam Li
By adding zone operations and zoned metadata, the zoned emulation
capability enables full emulation support of zoned device using
a qcow2 file. The zoned device metadata includes zone type,
zoned device state and write pointer (WP) of each zone, which is
stored to an array of unsigned integers.
WP accessor (qcow2_rw_wp_at) routes reads and writes of an 8-byte
WP slot through the write pointer cache. The write pointer cache is
written to disk after the qcow2 metadata is written, thus guaranteeing
that the write pointer is updated after the corresponding data is
written. Per-completion cache flush is deferred. The WP cluster reaches
disk on the next flush.
Each zone of a zoned device makes state transitions following
the zone state machine. The zone state machine mainly describes
five states, IMPLICIT OPEN, EXPLICIT OPEN, FULL, EMPTY and CLOSED.
READ ONLY and OFFLINE states will generally be affected by device
internal events. The operations on zones cause corresponding state
changing.
Zoned devices have limits on zone resources, which put constraints on
write operations on zones. It is managed by active zone queues
following LRU policy.
Signed-off-by: Sam Li <faithilikerun@gmail.com>
---
block/qcow2-cache.c | 8 +
block/qcow2-refcount.c | 7 +
block/qcow2.c | 1137 +++++++++++++++++++++++++++++++++++++++-
block/trace-events | 2 +
4 files changed, 1149 insertions(+), 5 deletions(-)
diff --git a/block/qcow2-cache.c b/block/qcow2-cache.c
index 23d9588b08..bdfb11ce88 100644
--- a/block/qcow2-cache.c
+++ b/block/qcow2-cache.c
@@ -275,6 +275,14 @@ int qcow2_cache_set_dependency(BlockDriverState *bs, Qcow2Cache *c,
{
int ret;
+ /*
+ * If the dependency graph is unchanged, nothing to do. This avoids
+ * a synchronous flush on every call below.
+ */
+ if (c->depends == dependency) {
+ return 0;
+ }
+
if (dependency->depends) {
ret = qcow2_cache_flush_dependency(bs, dependency);
if (ret < 0) {
diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index 6512cda407..f551726609 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -1239,6 +1239,13 @@ int qcow2_write_caches(BlockDriverState *bs)
}
}
+ if (s->wp_cache) {
+ ret = qcow2_cache_write(bs, s->wp_cache);
+ if (ret < 0) {
+ return ret;
+ }
+ }
+
return 0;
}
diff --git a/block/qcow2.c b/block/qcow2.c
index 29eec33e34..bdc8923b71 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -195,6 +195,300 @@ qcow2_extract_crypto_opts(QemuOpts *opts, const char *fmt, Error **errp)
return cryptoopts_qdict;
}
+#define QCOW2_ZT_IS_CONV(wp) (wp & 1ULL << 59)
+#define QCOW2_GET_WP(wp) ((wp << 5) >> 5)
+
+/*
+ * To emulate a real zoned device, closed, empty and full states are
+ * preserved after a power cycle. The open states are in-memory and will
+ * be lost after closing the device. Read-only and offline states are
+ * device-internal events, which are not considered for simplicity.
+ */
+static inline BlockZoneState qcow2_get_zone_state(BlockDriverState *bs,
+ uint32_t index)
+{
+ BDRVQcow2State *s = bs->opaque;
+ Qcow2ZoneListEntry *zone_entry = &s->zone_list_entries[index];
+ uint64_t zone_wp = bs->wps->wp[index];
+ uint64_t zone_start;
+
+ if (QCOW2_ZT_IS_CONV(zone_wp)) {
+ return BLK_ZS_NOT_WP;
+ }
+
+ if (QTAILQ_IN_USE(zone_entry, exp_open_zone_entry)) {
+ return BLK_ZS_EOPEN;
+ }
+ if (QTAILQ_IN_USE(zone_entry, imp_open_zone_entry)) {
+ return BLK_ZS_IOPEN;
+ }
+
+ zone_start = index * bs->bl.zone_size;
+ if (zone_wp == zone_start) {
+ return BLK_ZS_EMPTY;
+ }
+ if (zone_wp >= zone_start + bs->bl.zone_capacity) {
+ return BLK_ZS_FULL;
+ }
+ if (zone_wp > zone_start) {
+ if (!QTAILQ_IN_USE(zone_entry, closed_zone_entry)) {
+ /*
+ * The number of closed zones is not always updated in time when
+ * the device is closed. However, it only matters when doing
+ * zone report. Refresh the count and list of closed zones to
+ * provide correct zone states for zone report.
+ */
+ QTAILQ_INSERT_HEAD(&s->closed_zones, zone_entry, closed_zone_entry);
+ s->nr_zones_closed++;
+ }
+ return BLK_ZS_CLOSED;
+ }
+ return BLK_ZS_NOT_WP;
+}
+
+static void qcow2_rm_exp_open_zone(BDRVQcow2State *s,
+ uint32_t index)
+{
+ Qcow2ZoneListEntry *zone_entry = &s->zone_list_entries[index];
+
+ QTAILQ_REMOVE(&s->exp_open_zones, zone_entry, exp_open_zone_entry);
+ s->nr_zones_exp_open--;
+}
+
+static void qcow2_rm_imp_open_zone(BDRVQcow2State *s,
+ int32_t index)
+{
+ Qcow2ZoneListEntry *zone_entry;
+ if (index < 0) {
+ /* Apply LRU when the index is not specified. */
+ zone_entry = QTAILQ_LAST(&s->imp_open_zones);
+ } else {
+ zone_entry = &s->zone_list_entries[index];
+ }
+
+ QTAILQ_REMOVE(&s->imp_open_zones, zone_entry, imp_open_zone_entry);
+ s->nr_zones_imp_open--;
+}
+
+static void qcow2_rm_open_zone(BDRVQcow2State *s,
+ uint32_t index)
+{
+ Qcow2ZoneListEntry *zone_entry = &s->zone_list_entries[index];
+
+ if (QTAILQ_IN_USE(zone_entry, exp_open_zone_entry)) {
+ qcow2_rm_exp_open_zone(s, index);
+ } else if (QTAILQ_IN_USE(zone_entry, imp_open_zone_entry)) {
+ qcow2_rm_imp_open_zone(s, index);
+ }
+}
+
+static void qcow2_rm_closed_zone(BDRVQcow2State *s,
+ uint32_t index)
+{
+ Qcow2ZoneListEntry *zone_entry = &s->zone_list_entries[index];
+
+ QTAILQ_REMOVE(&s->closed_zones, zone_entry, closed_zone_entry);
+ s->nr_zones_closed--;
+}
+
+static void qcow2_do_imp_open_zone(BDRVQcow2State *s,
+ uint32_t index,
+ BlockZoneState zs)
+{
+ Qcow2ZoneListEntry *zone_entry = &s->zone_list_entries[index];
+
+ switch (zs) {
+ case BLK_ZS_EMPTY:
+ break;
+ case BLK_ZS_CLOSED:
+ qcow2_rm_closed_zone(s, index);
+ break;
+ case BLK_ZS_IOPEN:
+ /*
+ * The LRU policy: update the zone that is most recently
+ * used to the head of the zone list
+ */
+ if (zone_entry == QTAILQ_FIRST(&s->imp_open_zones)) {
+ return;
+ }
+ QTAILQ_REMOVE(&s->imp_open_zones, zone_entry, imp_open_zone_entry);
+ s->nr_zones_imp_open--;
+ break;
+ default:
+ return;
+ }
+
+ QTAILQ_INSERT_HEAD(&s->imp_open_zones, zone_entry, imp_open_zone_entry);
+ s->nr_zones_imp_open++;
+}
+
+static void qcow2_do_exp_open_zone(BDRVQcow2State *s,
+ uint32_t index)
+{
+ Qcow2ZoneListEntry *zone_entry = &s->zone_list_entries[index];
+
+ QTAILQ_INSERT_HEAD(&s->exp_open_zones, zone_entry, exp_open_zone_entry);
+ s->nr_zones_exp_open++;
+}
+
+/*
+ * The list of zones is managed using an LRU policy: the last
+ * zone of the list is always the one that was least recently used
+ * for writing and is chosen as the zone to close to be able to
+ * implicitly open another zone.
+ *
+ * We can only close the open zones. The index is not specified
+ * when it is less than 0.
+ */
+static void qcow2_do_close_zone(BlockDriverState *bs,
+ int32_t index,
+ BlockZoneState zs)
+{
+ BDRVQcow2State *s = bs->opaque;
+ Qcow2ZoneListEntry *zone_entry;
+
+ if (index >= 0) {
+ zone_entry = &s->zone_list_entries[index];
+ } else {
+ /* before removal of the last implicitly open zone */
+ zone_entry = QTAILQ_LAST(&s->imp_open_zones);
+ }
+
+ if (zs == BLK_ZS_IOPEN) {
+ qcow2_rm_imp_open_zone(s, index);
+ goto close_zone;
+ }
+
+ if (index >= 0 && zs == BLK_ZS_EOPEN) {
+ qcow2_rm_exp_open_zone(s, index);
+ /*
+ * The zone state changes when the zone is removed from the list of
+ * open zones (explicitly open -> empty). The closed zone list is
+ * refreshed during get_zone_state().
+ */
+ qcow2_get_zone_state(bs, index);
+ }
+ return;
+
+close_zone:
+ QTAILQ_INSERT_HEAD(&s->closed_zones, zone_entry, closed_zone_entry);
+ s->nr_zones_closed++;
+}
+
+/*
+ * Read/Write the new wp value for zone `index` through write pointe
+ * cache. Reads return the value currently held in the cache, which may
+ * be ahead of the on-disk value if the cache hasn't been flushed yet.
+ * Writes update the cache and mark the entry dirty.
+ */
+static int coroutine_fn GRAPH_RDLOCK
+qcow2_rw_wp_at(BlockDriverState *bs, uint64_t *wp,
+ int32_t index, bool is_write) {
+ BDRVQcow2State *s = bs->opaque;
+ uint64_t wp_byte_off = sizeof(uint64_t) * index;
+ uint64_t cluster_file_off =
+ s->zoned_header.zonedmeta_offset +
+ (wp_byte_off & ~((uint64_t)s->cluster_size - 1));
+ size_t off_in_cluster = wp_byte_off & (s->cluster_size - 1);
+ void *cluster_buf;
+ uint64_t *slot;
+ int ret;
+
+ assert(s->wp_cache != NULL);
+
+ qemu_co_mutex_lock(&s->lock);
+ ret = qcow2_cache_get(bs, s->wp_cache, cluster_file_off, &cluster_buf);
+ if (ret < 0) {
+ qemu_co_mutex_unlock(&s->lock);
+ error_report("Failed to %s WP slot (zone %d): %s",
+ is_write ? "write" : "read", index, strerror(-ret));
+ return ret;
+ }
+
+ slot = (uint64_t *)((char *)cluster_buf + off_in_cluster);
+ if (is_write) {
+ *slot = *wp;
+ qcow2_cache_entry_mark_dirty(s->wp_cache, cluster_buf);
+ } else {
+ *wp = *slot;
+ }
+ qcow2_cache_put(s->wp_cache, &cluster_buf);
+ qemu_co_mutex_unlock(&s->lock);
+
+ trace_qcow2_wp_tracking(index, *wp >> BDRV_SECTOR_BITS);
+ return 0;
+}
+
+static bool qcow2_can_activate_zone(BlockDriverState *bs)
+{
+ BDRVQcow2State *s = bs->opaque;
+
+ /* When the max active zone is zero, there is no limit on active zones */
+ if (!s->zoned_header.max_active_zones) {
+ return true;
+ }
+
+ /* Active zones are zones that are open or closed */
+ return s->nr_zones_exp_open + s->nr_zones_imp_open + s->nr_zones_closed
+ < s->zoned_header.max_active_zones;
+}
+
+/*
+ * This function manages open zones under active zones limit. It checks
+ * if a zone can transition to open state while maintaining max open and
+ * active zone limits.
+ */
+static bool qcow2_can_open_zone(BlockDriverState *bs)
+{
+ BDRVQcow2State *s = bs->opaque;
+
+ /* When the max open zone is zero, there is no limit on open zones */
+ if (!s->zoned_header.max_open_zones) {
+ return true;
+ }
+
+ /*
+ * The open zones are zones with the states of explicitly and
+ * implicitly open.
+ */
+ if (s->nr_zones_imp_open + s->nr_zones_exp_open <
+ s->zoned_header.max_open_zones) {
+ return true;
+ }
+
+ /*
+ * Zones are managed one at a time. Thus, the number of implicitly open
+ * zone can never be over the open zone limit. When the active zone limit
+ * is not reached, close only one implicitly open zone.
+ */
+ if (qcow2_can_activate_zone(bs)) {
+ qcow2_do_close_zone(bs, -1, BLK_ZS_IOPEN);
+ trace_qcow2_imp_open_zones(0x23, s->nr_zones_imp_open);
+ return true;
+ }
+ return false;
+}
+
+static inline int coroutine_fn GRAPH_RDLOCK
+qcow2_refresh_zonedmeta(BlockDriverState *bs)
+{
+ int ret;
+ BDRVQcow2State *s = bs->opaque;
+ uint64_t wps_size = s->zoned_header.nr_zones * sizeof(uint64_t);
+ g_autofree uint64_t *temp = NULL;
+
+ temp = g_new(uint64_t, s->zoned_header.nr_zones);
+ ret = bdrv_pread(bs->file, s->zoned_header.zonedmeta_offset,
+ wps_size, temp, 0);
+ if (ret < 0) {
+ error_report("Cannot read metadata");
+ return ret;
+ }
+
+ memcpy(bs->wps->wp, temp, wps_size);
+ return 0;
+}
+
/*
* Passing by the zoned device configurations by a zoned_header struct, check
* if the zone device options are under constraints. Return false when some
@@ -545,7 +839,37 @@ qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset,
be32_to_cpu(zoned_ext.max_active_zones);
zoned_ext.max_append_bytes =
be32_to_cpu(zoned_ext.max_append_bytes);
+ zoned_ext.zonedmeta_offset =
+ be64_to_cpu(zoned_ext.zonedmeta_offset);
s->zoned_header = zoned_ext;
+ bs->wps = g_malloc(sizeof(BlockZoneWps)
+ + s->zoned_header.nr_zones * sizeof(uint64_t));
+ ret = qcow2_refresh_zonedmeta(bs);
+ if (ret < 0) {
+ return ret;
+ }
+
+ s->zone_list_entries = g_new0(Qcow2ZoneListEntry,
+ zoned_ext.nr_zones);
+ QTAILQ_INIT(&s->exp_open_zones);
+ QTAILQ_INIT(&s->imp_open_zones);
+ QTAILQ_INIT(&s->closed_zones);
+ qemu_co_mutex_init(&bs->wps->colock);
+
+ s->zone_wp_state = g_new0(Qcow2ZoneWPState, zoned_ext.nr_zones);
+ for (uint32_t i = 0; i < zoned_ext.nr_zones; i++) {
+ QTAILQ_INIT(&s->zone_wp_state[i].in_flight);
+ QTAILQ_INIT(&s->zone_wp_state[i].completed_pending);
+ }
+
+ s->wp_cache = qcow2_cache_create(bs,
+ DIV_ROUND_UP(zoned_ext.nr_zones * sizeof(uint64_t),
+ s->cluster_size),
+ s->cluster_size);
+ if (!s->wp_cache) {
+ error_setg(errp, "Could not allocate the write pointer cache");
+ return -ENOMEM;
+ }
/* refuse to open broken images */
if (zoned_ext.nr_zones != DIV_ROUND_UP(bs->total_sectors *
@@ -2911,21 +3235,277 @@ static coroutine_fn GRAPH_RDLOCK int qcow2_co_pwritev_task_entry(AioTask *task)
t->l2meta);
}
+/*
+ * Walk the per-zone completed_pending list while the head's LBA equals
+ * the cached persisted-WP, advancing persisted-WP and marking each
+ * covered request RESOLVED. On any advance, the new persisted-WP is
+ * written into the WP cache and depends-on qcow2 L2 table cache is
+ * established so write pointer is updated after the corresponding write
+ * completes.
+ *
+ * Caller holds bs->wps->colock.
+ */
static int coroutine_fn GRAPH_RDLOCK
-qcow2_co_pwritev_part(BlockDriverState *bs, int64_t offset, int64_t bytes,
- QEMUIOVector *qiov, size_t qiov_offset,
- BdrvRequestFlags flags)
+qcow2_wp_greedy_advance_locked(BlockDriverState *bs, uint32_t index)
+{
+ BDRVQcow2State *s = bs->opaque;
+ Qcow2ZoneWPState *zs = &s->zone_wp_state[index];
+ uint64_t wp_byte_off = sizeof(uint64_t) * index;
+ uint64_t cluster_file_off =
+ s->zoned_header.zonedmeta_offset +
+ (wp_byte_off & ~((uint64_t)s->cluster_size - 1));
+ size_t off_in_cluster = wp_byte_off & (s->cluster_size - 1);
+ void *cluster_buf;
+ uint64_t *slot;
+ uint64_t persisted_wp, new_persisted_wp;
+ Qcow2WPReq *head;
+ bool any_advanced = false;
+ int ret;
+
+ qemu_co_mutex_lock(&s->lock);
+ ret = qcow2_cache_get(bs, s->wp_cache, cluster_file_off, &cluster_buf);
+ if (ret < 0) {
+ qemu_co_mutex_unlock(&s->lock);
+ return ret;
+ }
+ slot = (uint64_t *)((char *)cluster_buf + off_in_cluster);
+ persisted_wp = *slot;
+ new_persisted_wp = persisted_wp;
+
+ QTAILQ_FOREACH(head, &zs->completed_pending, entry) {
+ if (head->lba != new_persisted_wp) {
+ break;
+ }
+ new_persisted_wp += head->len;
+ any_advanced = true;
+ }
+
+ if (!any_advanced) {
+ qcow2_cache_put(s->wp_cache, &cluster_buf);
+ qemu_co_mutex_unlock(&s->lock);
+ return 0;
+ }
+
+ *slot = new_persisted_wp;
+ qcow2_cache_entry_mark_dirty(s->wp_cache, cluster_buf);
+ qcow2_cache_put(s->wp_cache, &cluster_buf);
+
+ qcow2_cache_set_dependency(bs, s->wp_cache, s->l2_table_cache);
+ qemu_co_mutex_unlock(&s->lock);
+
+ while (!QTAILQ_EMPTY(&zs->completed_pending)) {
+ head = QTAILQ_FIRST(&zs->completed_pending);
+ if (head->lba >= new_persisted_wp) {
+ break;
+ }
+ QTAILQ_REMOVE(&zs->completed_pending, head, entry);
+ head->state = QCOW2_WP_REQ_RESOLVED;
+ qemu_co_queue_restart_all(&head->wait);
+ }
+
+ return 0;
+}
+
+/*
+ * Insert req into the zone's completed_pending list in ascending LBA
+ * order. Data writes may complete out of order; greedy_advance walks
+ * the list head while head->lba == persisted_wp, so the list must
+ * stay sorted.
+ *
+ * Caller holds bs->wps->colock.
+ */
+static void
+qcow2_wp_insert_pending_locked(Qcow2ZoneWPState *zs, Qcow2WPReq *req)
+{
+ Qcow2WPReq *iter;
+
+ req->state = QCOW2_WP_REQ_PENDING;
+
+ QTAILQ_FOREACH(iter, &zs->completed_pending, entry) {
+ if (iter->lba > req->lba) {
+ QTAILQ_INSERT_BEFORE(iter, req, entry);
+ return;
+ }
+ }
+ QTAILQ_INSERT_TAIL(&zs->completed_pending, req, entry);
+}
+
+/*
+ * Peer requests with lba > failed_lba are marked ABORTED. Survivors
+ * with lba < failed_lba are left untouched. Logical WP is rolled back
+ * to the contiguous extent above persisted-WP that remains covered by
+ * survivor requests.
+ *
+ * Caller holds bs->wps->colock.
+ */
+static void
+qcow2_wp_abort_higher_peers_locked(BlockDriverState *bs, uint32_t index,
+ uint64_t failed_lba)
+{
+ BDRVQcow2State *s = bs->opaque;
+ Qcow2ZoneWPState *zs = &s->zone_wp_state[index];
+ Qcow2WPReq *r, *next;
+ uint64_t logical_wp;
+
+ QTAILQ_FOREACH(r, &zs->in_flight, entry) {
+ if (r->lba > failed_lba && r->state != QCOW2_WP_REQ_ABORTED) {
+ r->state = QCOW2_WP_REQ_ABORTED;
+ qemu_co_queue_restart_all(&r->wait);
+ }
+ }
+ QTAILQ_FOREACH_SAFE(r, &zs->completed_pending, entry, next) {
+ if (r->lba > failed_lba) {
+ r->state = QCOW2_WP_REQ_ABORTED;
+ QTAILQ_REMOVE(&zs->completed_pending, r, entry);
+ qemu_co_queue_restart_all(&r->wait);
+ }
+ }
+
+ /* Rewind logical_wp to the highest survivor end_offset. */
+ logical_wp = failed_lba;
+ QTAILQ_FOREACH(r, &zs->in_flight, entry) {
+ if (r->lba < failed_lba && r->lba + r->len > logical_wp) {
+ logical_wp = r->lba + r->len;
+ }
+ }
+ QTAILQ_FOREACH(r, &zs->completed_pending, entry) {
+ if (r->lba < failed_lba && r->lba + r->len > logical_wp) {
+ logical_wp = r->lba + r->len;
+ }
+ }
+ bs->wps->wp[index] = logical_wp;
+}
+
+/*
+ * If it is an append write request, the offset pointer needs to be updated to
+ * the wp value of that zone after the IO completion. The unique pointer is
+ * passed on to this function to prevent the value being changed in condition of
+ * multiple concurrent writes.
+ */
+static int coroutine_fn GRAPH_RDLOCK
+qcow2_co_pwv_part(BlockDriverState *bs, int64_t *offset_ptr, int64_t bytes,
+ QEMUIOVector *qiov, size_t qiov_offset, bool is_append,
+ BdrvRequestFlags flags)
{
BDRVQcow2State *s = bs->opaque;
int offset_in_cluster;
int ret;
unsigned int cur_bytes; /* number of sectors in current iteration */
uint64_t host_offset;
+ int64_t offset = *offset_ptr;
QCowL2Meta *l2meta = NULL;
AioTaskPool *aio = NULL;
+ int64_t start_offset, start_bytes;
+ BlockZoneState zs;
+ int64_t end_zone, end_offset;
+ uint64_t *wp;
+ int64_t zone_size = bs->bl.zone_size;
+ int64_t zone_capacity = bs->bl.zone_capacity;
+ int index = 0;
+ Qcow2WPReq req;
+ Qcow2WPReq *wp_req = NULL;
trace_qcow2_writev_start_req(qemu_coroutine_self(), offset, bytes);
+ start_offset = offset;
+ start_bytes = bytes;
+ if (bs->bl.zoned == BLK_Z_HM) {
+ index = start_offset / zone_size;
+ wp = &bs->wps->wp[index];
+ if (!QCOW2_ZT_IS_CONV(*wp)) {
+ if (offset != *wp && !is_append) {
+ /* The write offset must be equal to the zone write pointer */
+ error_report("Offset 0x%" PRIx64 " of regular writes must be "
+ "equal to the zone write pointer 0x%" PRIx64 "",
+ offset, *wp);
+ return -EINVAL;
+ }
+
+ if (is_append) {
+ /*
+ * The offset of append write is the write pointer value of
+ * that zone.
+ */
+ start_offset = *wp;
+ }
+
+ end_offset = start_offset + start_bytes;
+
+ /* Only allow writes when there are zone resources left */
+ zs = qcow2_get_zone_state(bs, index);
+ if (zs == BLK_ZS_CLOSED || zs == BLK_ZS_EMPTY) {
+ if (!qcow2_can_open_zone(bs)) {
+ error_report("no more open zones available");
+ return -EINVAL;
+ }
+ }
+
+ /*
+ * Align up (start_offset, zone_size), the start offset is not
+ * necessarily power of two.
+ */
+ end_zone = index * zone_size + zone_capacity;
+ /* Write cannot exceed the zone capacity. */
+ if (end_offset > end_zone) {
+ error_report("write exceeds zone capacity with end_offset:"
+ "0x%lx, end_zone: 0x%lx",
+ end_offset / 512, end_zone / 512);
+ return -EINVAL;
+ }
+
+ /*
+ * Real drives change states before it can write to the zone. If
+ * the write fails, the zone state may have changed.
+ *
+ * The zone state transitions to implicit open when the original
+ * state is empty or closed. When the wp reaches the end, the
+ * open states (explicit open, implicit open) become full.
+ */
+ zs = qcow2_get_zone_state(bs, index);
+ if (!(end_offset & (zone_capacity - 1))) {
+ /* Being aligned to zone capacity implies full state */
+ qcow2_rm_open_zone(s, index);
+ trace_qcow2_imp_open_zones(0x24,
+ s->nr_zones_imp_open);
+ } else {
+ qcow2_do_imp_open_zone(s, index, zs);
+ trace_qcow2_imp_open_zones(0x24,
+ s->nr_zones_imp_open);
+ }
+
+ /*
+ * Submission for a zone append write. The logical-WP is updated
+ * while the on-disk WP is not touched.
+ */
+ qemu_co_mutex_lock(&bs->wps->colock);
+ if (is_append) {
+ start_offset = *wp;
+ end_offset = start_offset + start_bytes;
+ end_zone = (uint64_t)index * zone_size + zone_capacity;
+ if (end_offset > end_zone) {
+ qemu_co_mutex_unlock(&bs->wps->colock);
+ error_report("append: end_offset 0x%" PRIx64
+ " > end_zone 0x%" PRIx64,
+ end_offset, end_zone);
+ return -EINVAL;
+ }
+ *offset_ptr = start_offset;
+ offset = start_offset;
+ }
+
+ wp_req = &req;
+ wp_req->lba = start_offset;
+ wp_req->len = start_bytes;
+ wp_req->state = QCOW2_WP_REQ_INFLIGHT;
+ qemu_co_queue_init(&wp_req->wait);
+ QTAILQ_INSERT_TAIL(&s->zone_wp_state[index].in_flight,
+ wp_req, entry);
+
+ *wp = end_offset;
+ qemu_co_mutex_unlock(&bs->wps->colock);
+ }
+ }
+
while (bytes != 0 && aio_task_pool_status(aio) == 0) {
l2meta = NULL;
@@ -2989,11 +3569,75 @@ fail_nometa:
g_free(aio);
}
+ if (wp_req != NULL) {
+ qemu_co_mutex_lock(&bs->wps->colock);
+ QTAILQ_REMOVE(&s->zone_wp_state[index].in_flight, wp_req, entry);
+
+ if (wp_req->state == QCOW2_WP_REQ_ABORTED) {
+ /*
+ * A peer's failure handler aborted us. Whether our data
+ * write itself succeeded or not, reject it.
+ */
+ qemu_co_mutex_unlock(&bs->wps->colock);
+ ret = ret < 0 ? ret : -EIO;
+ goto wp_done;
+ }
+
+ if (ret < 0) {
+ /*
+ * This req's data write failed. Higher-LBA peers (still
+ * in_flight or already completed_pending) are marked ABORTED,
+ * their waiters woken; logical-WP rewinds to the highest
+ * surviving end_offset below this LBA. The zone remains usable.
+ */
+ qcow2_wp_abort_higher_peers_locked(bs, index, wp_req->lba);
+ qemu_co_mutex_unlock(&bs->wps->colock);
+ goto wp_done;
+ }
+
+ qcow2_wp_insert_pending_locked(&s->zone_wp_state[index], wp_req);
+
+ ret = qcow2_wp_greedy_advance_locked(bs, index);
+ if (ret < 0) {
+ if (wp_req->state == QCOW2_WP_REQ_PENDING) {
+ QTAILQ_REMOVE(&s->zone_wp_state[index].completed_pending,
+ wp_req, entry);
+ }
+ qcow2_wp_abort_higher_peers_locked(bs, index, wp_req->lba);
+ qemu_co_mutex_unlock(&bs->wps->colock);
+ goto wp_done;
+ }
+
+ /* Block until this write pointer req is RESOLVED or ABORTED. */
+ while (wp_req->state == QCOW2_WP_REQ_PENDING) {
+ qemu_co_queue_wait(&wp_req->wait, &bs->wps->colock);
+ }
+
+ if (wp_req->state == QCOW2_WP_REQ_ABORTED) {
+ ret = -EIO;
+ } else {
+ assert(wp_req->state == QCOW2_WP_REQ_RESOLVED);
+ ret = 0;
+ }
+ qemu_co_mutex_unlock(&bs->wps->colock);
+ }
+
+wp_done:
trace_qcow2_writev_done_req(qemu_coroutine_self(), ret);
return ret;
}
+static int coroutine_fn GRAPH_RDLOCK
+qcow2_co_pwritev_part(BlockDriverState *bs, int64_t offset, int64_t bytes,
+ QEMUIOVector *qiov, size_t qiov_offset,
+ BdrvRequestFlags flags)
+{
+ return qcow2_co_pwv_part(bs, &offset, bytes, qiov, qiov_offset, false,
+ flags);
+}
+
+
static int GRAPH_RDLOCK qcow2_inactivate(BlockDriverState *bs)
{
BDRVQcow2State *s = bs->opaque;
@@ -3022,6 +3666,15 @@ static int GRAPH_RDLOCK qcow2_inactivate(BlockDriverState *bs)
strerror(-ret));
}
+ if (s->wp_cache) {
+ ret = qcow2_cache_flush(bs, s->wp_cache);
+ if (ret) {
+ result = ret;
+ error_report("Failed to flush the WP cache: %s",
+ strerror(-ret));
+ }
+ }
+
if (result == 0) {
qcow2_mark_clean(bs);
}
@@ -3029,6 +3682,25 @@ static int GRAPH_RDLOCK qcow2_inactivate(BlockDriverState *bs)
return result;
}
+static void qcow2_do_close_all_zone(BDRVQcow2State *s)
+{
+ Qcow2ZoneListEntry *zone_entry, *next;
+
+ QTAILQ_FOREACH_SAFE(zone_entry, &s->imp_open_zones, imp_open_zone_entry,
+ next) {
+ QTAILQ_REMOVE(&s->imp_open_zones, zone_entry, imp_open_zone_entry);
+ s->nr_zones_imp_open--;
+ }
+
+ QTAILQ_FOREACH_SAFE(zone_entry, &s->exp_open_zones, exp_open_zone_entry,
+ next) {
+ QTAILQ_REMOVE(&s->exp_open_zones, zone_entry, exp_open_zone_entry);
+ s->nr_zones_exp_open--;
+ }
+
+ assert(s->nr_zones_imp_open + s->nr_zones_exp_open == 0);
+}
+
static void coroutine_mixed_fn GRAPH_RDLOCK
qcow2_do_close(BlockDriverState *bs, bool close_data_file)
{
@@ -3044,6 +3716,10 @@ qcow2_do_close(BlockDriverState *bs, bool close_data_file)
cache_clean_timer_del_and_wait(bs);
qcow2_cache_destroy(s->l2_table_cache);
qcow2_cache_destroy(s->refcount_block_cache);
+ if (s->wp_cache) {
+ qcow2_cache_destroy(s->wp_cache);
+ s->wp_cache = NULL;
+ }
qcrypto_block_free(s->crypto);
s->crypto = NULL;
@@ -3068,6 +3744,12 @@ qcow2_do_close(BlockDriverState *bs, bool close_data_file)
qcow2_refcount_close(bs);
qcow2_free_snapshots(bs);
+ qcow2_do_close_all_zone(s);
+ g_free(s->zone_list_entries);
+ s->zone_list_entries = NULL;
+ g_free(s->zone_wp_state);
+ s->zone_wp_state = NULL;
+ g_free(bs->wps);
}
static void GRAPH_UNLOCKED qcow2_close(BlockDriverState *bs)
@@ -3385,7 +4067,9 @@ int qcow2_update_header(BlockDriverState *bs)
.max_active_zones =
cpu_to_be32(s->zoned_header.max_active_zones),
.max_append_bytes =
- cpu_to_be32(s->zoned_header.max_append_bytes)
+ cpu_to_be32(s->zoned_header.max_append_bytes),
+ .zonedmeta_offset =
+ cpu_to_be64(s->zoned_header.zonedmeta_offset),
};
ret = header_ext_add(buf, QCOW2_EXT_MAGIC_ZONED_FORMAT,
&zoned_header, sizeof(zoned_header),
@@ -3794,7 +4478,8 @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp)
int version;
int refcount_order;
uint64_t *refcount_table;
- int ret;
+ uint64_t zoned_meta_size, zoned_clusterlen;
+ int ret, offset, i;
uint8_t compression_type = QCOW2_COMPRESSION_TYPE_ZLIB;
assert(create_options->driver == BLOCKDEV_DRIVER_QCOW2);
@@ -4155,6 +4840,41 @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp)
ret = -EINVAL;
goto unlock;
}
+
+ uint32_t nrz = s->zoned_header.nr_zones;
+ zoned_meta_size = sizeof(uint64_t) * nrz;
+ g_autofree uint64_t *meta = NULL;
+ meta = g_new0(uint64_t, nrz);
+
+ for (i = 0; i < s->zoned_header.conventional_zones; ++i) {
+ meta[i] = i * s->zoned_header.zone_size;
+ meta[i] |= 1ULL << 59;
+ }
+
+ for (; i < nrz; ++i) {
+ meta[i] = i * s->zoned_header.zone_size;
+ }
+
+ offset = qcow2_alloc_clusters(blk_bs(blk), zoned_meta_size);
+ if (offset < 0) {
+ ret = offset;
+ error_setg_errno(errp, -ret, "Could not allocate clusters "
+ "for zoned metadata size");
+ goto unlock;
+ }
+ s->zoned_header.zonedmeta_offset = offset;
+
+ zoned_clusterlen = size_to_clusters(s, zoned_meta_size)
+ * s->cluster_size;
+ ret = qcow2_pre_write_overlap_check(blk_bs(blk), 0, offset,
+ zoned_clusterlen, false);
+ assert(ret == 0);
+ ret = bdrv_pwrite(blk_bs(blk)->file, offset, zoned_meta_size, meta, 0);
+ if (ret < 0) {
+ error_setg_errno(errp, -ret, "Could not write zoned metadata "
+ "to disk");
+ goto unlock;
+ }
} else {
s->zoned_header.zoned = QCOW2_Z_NONE;
}
@@ -4554,6 +5274,409 @@ qcow2_co_pdiscard(BlockDriverState *bs, int64_t offset, int64_t bytes)
return ret;
}
+static int coroutine_fn
+qcow2_co_zone_report(BlockDriverState *bs, int64_t offset,
+ unsigned int *nr_zones, BlockZoneDescriptor *zones)
+{
+ BDRVQcow2State *s = bs->opaque;
+ uint64_t zone_size = s->zoned_header.zone_size;
+ int64_t capacity = bs->total_sectors << BDRV_SECTOR_BITS;
+ int64_t size = bs->bl.nr_zones * zone_size;
+ unsigned int nrz;
+ int i = 0;
+ int si;
+
+ if (offset >= capacity) {
+ error_report("offset %" PRId64 " is equal to or greater than the "
+ "device capacity %" PRId64 "", offset, capacity);
+ return -EINVAL;
+ }
+
+ nrz = ((*nr_zones) < bs->bl.nr_zones) ? (*nr_zones) : bs->bl.nr_zones;
+ si = offset / zone_size; /* Zone size cannot be 0 for zoned device */
+ qemu_co_mutex_lock(&bs->wps->colock);
+ for (; i < nrz; ++i) {
+ if (i + si >= bs->bl.nr_zones) {
+ break;
+ }
+
+ zones[i].start = (si + i) * zone_size;
+
+ /* The last zone can be smaller than the zone size */
+ if ((si + i + 1) == bs->bl.nr_zones && size > capacity) {
+ uint32_t l = zone_size - (size - capacity);
+ zones[i].length = l;
+ zones[i].cap = l;
+ } else {
+ zones[i].length = zone_size;
+ zones[i].cap = zone_size;
+ }
+
+ uint64_t wp = bs->wps->wp[si + i];
+ if (QCOW2_ZT_IS_CONV(wp)) {
+ zones[i].type = BLK_ZT_CONV;
+ zones[i].state = BLK_ZS_NOT_WP;
+ /* Clear masking bits */
+ wp = QCOW2_GET_WP(wp);
+ } else {
+ zones[i].type = BLK_ZT_SWR;
+ zones[i].state = qcow2_get_zone_state(bs, si + i);
+ }
+ zones[i].wp = wp;
+ }
+ qemu_co_mutex_unlock(&bs->wps->colock);
+ *nr_zones = i;
+ return 0;
+}
+
+static int coroutine_fn GRAPH_RDLOCK
+qcow2_open_zone(BlockDriverState *bs, uint32_t index) {
+ BDRVQcow2State *s = bs->opaque;
+ int ret;
+
+ qemu_co_mutex_lock(&bs->wps->colock);
+ BlockZoneState zs = qcow2_get_zone_state(bs, index);
+ trace_qcow2_imp_open_zones(BLK_ZO_OPEN, s->nr_zones_imp_open);
+
+ switch (zs) {
+ case BLK_ZS_EMPTY:
+ if (!qcow2_can_activate_zone(bs)) {
+ ret = -EBUSY;
+ goto unlock;
+ }
+ break;
+ case BLK_ZS_IOPEN:
+ qcow2_rm_imp_open_zone(s, index);
+ break;
+ case BLK_ZS_EOPEN:
+ return 0;
+ case BLK_ZS_CLOSED:
+ if (!qcow2_can_open_zone(bs)) {
+ ret = -EINVAL;
+ goto unlock;
+ }
+ qcow2_rm_closed_zone(s, index);
+ break;
+ case BLK_ZS_FULL:
+ break;
+ default:
+ ret = -EINVAL;
+ goto unlock;
+ }
+
+ qcow2_do_exp_open_zone(s, index);
+ ret = 0;
+
+unlock:
+ qemu_co_mutex_unlock(&bs->wps->colock);
+ return ret;
+}
+
+static int qcow2_close_zone(BlockDriverState *bs, uint32_t index)
+{
+ int ret;
+
+ qemu_co_mutex_lock(&bs->wps->colock);
+ BlockZoneState zs = qcow2_get_zone_state(bs, index);
+
+ switch (zs) {
+ case BLK_ZS_EMPTY:
+ break;
+ case BLK_ZS_IOPEN:
+ break;
+ case BLK_ZS_EOPEN:
+ break;
+ case BLK_ZS_CLOSED:
+ /* Closing a closed zone is not an error */
+ ret = 0;
+ goto unlock;
+ case BLK_ZS_FULL:
+ break;
+ default:
+ ret = -EINVAL;
+ goto unlock;
+ }
+ qcow2_do_close_zone(bs, index, zs);
+ ret = 0;
+
+unlock:
+ qemu_co_mutex_unlock(&bs->wps->colock);
+ return ret;
+}
+
+static int coroutine_fn GRAPH_RDLOCK
+qcow2_finish_zone(BlockDriverState *bs, uint32_t index) {
+ BDRVQcow2State *s = bs->opaque;
+ int ret;
+
+ qemu_co_mutex_lock(&bs->wps->colock);
+ uint64_t *wp = &bs->wps->wp[index];
+ BlockZoneState zs = qcow2_get_zone_state(bs, index);
+
+ switch (zs) {
+ case BLK_ZS_EMPTY:
+ if (!qcow2_can_activate_zone(bs)) {
+ ret = -EBUSY;
+ goto unlock;
+ }
+ break;
+ case BLK_ZS_IOPEN:
+ qcow2_rm_imp_open_zone(s, index);
+ trace_qcow2_imp_open_zones(BLK_ZO_FINISH, s->nr_zones_imp_open);
+ break;
+ case BLK_ZS_EOPEN:
+ qcow2_rm_exp_open_zone(s, index);
+ break;
+ case BLK_ZS_CLOSED:
+ if (!qcow2_can_open_zone(bs)) {
+ ret = -EINVAL;
+ goto unlock;
+ }
+ qcow2_rm_closed_zone(s, index);
+ break;
+ case BLK_ZS_FULL:
+ ret = 0;
+ goto unlock;
+ default:
+ ret = -EINVAL;
+ goto unlock;
+ }
+
+ /* Reject if any append write is still in flight for this zone. */
+ if (s->zone_wp_state &&
+ !QTAILQ_EMPTY(&s->zone_wp_state[index].in_flight)) {
+ ret = -EBUSY;
+ goto unlock;
+ }
+
+ *wp = ((uint64_t)index + 1) * s->zoned_header.zone_size;
+ ret = qcow2_rw_wp_at(bs, wp, index, true);
+ if (ret < 0) {
+ goto unlock;
+ }
+
+ /*
+ * Flush the WP cache so the on-disk write pointer reflects the new state
+ * on return.
+ */
+ qemu_co_mutex_lock(&s->lock);
+ ret = qcow2_cache_flush(bs, s->wp_cache);
+ qemu_co_mutex_unlock(&s->lock);
+
+unlock:
+ qemu_co_mutex_unlock(&bs->wps->colock);
+ return ret;
+}
+
+static int coroutine_fn GRAPH_RDLOCK
+qcow2_reset_zone(BlockDriverState *bs, uint32_t index,
+ int64_t len) {
+ BDRVQcow2State *s = bs->opaque;
+ int nrz = bs->bl.nr_zones;
+ int zone_size = bs->bl.zone_size;
+ int n, ret = 0;
+ bool any_dirtied = false;
+
+ qemu_co_mutex_lock(&bs->wps->colock);
+ uint64_t *wp = &bs->wps->wp[index];
+ if (len == bs->total_sectors << BDRV_SECTOR_BITS) {
+ n = nrz;
+ index = 0;
+ wp = &bs->wps->wp[0];
+ } else {
+ n = len / zone_size;
+ }
+
+ for (int i = 0; i < n; ++i) {
+ uint64_t *wp_i = (uint64_t *)(wp + i);
+ uint64_t wpi_v = *wp_i;
+ if (QCOW2_ZT_IS_CONV(wpi_v)) {
+ continue;
+ }
+
+ /* Reject if any write is in flight for this zone. */
+ if (s->zone_wp_state &&
+ !QTAILQ_EMPTY(&s->zone_wp_state[index + i].in_flight)) {
+ ret = -EBUSY;
+ goto unlock;
+ }
+
+ BlockZoneState zs = qcow2_get_zone_state(bs, index + i);
+ switch (zs) {
+ case BLK_ZS_EMPTY:
+ break;
+ case BLK_ZS_IOPEN:
+ qcow2_rm_imp_open_zone(s, index + i);
+ trace_qcow2_imp_open_zones(BLK_ZO_RESET, s->nr_zones_imp_open);
+ break;
+ case BLK_ZS_EOPEN:
+ qcow2_rm_exp_open_zone(s, index + i);
+ break;
+ case BLK_ZS_CLOSED:
+ qcow2_rm_closed_zone(s, index + i);
+ break;
+ case BLK_ZS_FULL:
+ break;
+ default:
+ ret = -EINVAL;
+ goto unlock;
+ }
+
+ if (zs == BLK_ZS_EMPTY) {
+ continue;
+ }
+
+ /*
+ * Zero the data extent first. Date write fires before the WP cluster
+ * hits disk. So the wp advance cannot become durable while stale data
+ * is still readable.
+ */
+ ret = qcow2_co_pwrite_zeroes(bs, (uint64_t)(index + i) * zone_size,
+ zone_size, 0);
+ if (ret < 0) {
+ error_report("Failed to clear zone data at zone %u",
+ index + i);
+ goto unlock;
+ }
+
+ qemu_co_mutex_lock(&s->lock);
+ qcow2_cache_depends_on_flush(s->wp_cache);
+ qemu_co_mutex_unlock(&s->lock);
+
+ *wp_i = (uint64_t)(index + i) * zone_size;
+ ret = qcow2_rw_wp_at(bs, wp_i, index + i, true);
+ if (ret < 0) {
+ goto unlock;
+ }
+ any_dirtied = true;
+ }
+
+ if (any_dirtied) {
+ /* Single flush at the end. */
+ qemu_co_mutex_lock(&s->lock);
+ ret = qcow2_cache_flush(bs, s->wp_cache);
+ qemu_co_mutex_unlock(&s->lock);
+ }
+
+unlock:
+ qemu_co_mutex_unlock(&bs->wps->colock);
+ return ret;
+}
+
+static int coroutine_fn GRAPH_RDLOCK
+qcow2_co_zone_mgmt(BlockDriverState *bs, BlockZoneOp op,
+ int64_t offset, int64_t len)
+{
+ BDRVQcow2State *s = bs->opaque;
+ int ret = 0;
+ int64_t capacity = bs->total_sectors << BDRV_SECTOR_BITS;
+ int64_t zone_size = s->zoned_header.zone_size;
+ int64_t zone_size_mask = zone_size - 1;
+ uint32_t index = offset / zone_size;
+ BlockZoneWps *wps = bs->wps;
+
+ if (offset >= capacity) {
+ error_report("offset %" PRId64 " is equal to or greater than the"
+ "device capacity %" PRId64 "", offset, capacity);
+ return -EINVAL;
+ }
+
+ if (offset & zone_size_mask) {
+ error_report("sector offset %" PRId64 " is not aligned to zone size"
+ " %" PRId64 "", offset / 512, zone_size / 512);
+ return -EINVAL;
+ }
+
+ if (((offset + len) < capacity && len & zone_size_mask) ||
+ offset + len > capacity) {
+ error_report("number of sectors %" PRId64 " is not aligned to zone"
+ " size %" PRId64 "", len / 512, zone_size / 512);
+ return -EINVAL;
+ }
+
+ qemu_co_mutex_lock(&wps->colock);
+ uint64_t wpv = wps->wp[index];
+ qemu_co_mutex_unlock(&wps->colock);
+
+ if (QCOW2_ZT_IS_CONV(wpv)) {
+ /*
+ * ZONE_RESET_ALL is a global operation that is allowed when the
+ * starting zone is conventional; the zone reset path itself skips
+ * conventional zones.
+ */
+ if (op != BLK_ZO_RESET || len != capacity) {
+ error_report("zone mgmt operation 0x%x is not allowed on "
+ "a conventional zone", op);
+ return -EIO;
+ }
+ }
+
+ switch (op) {
+ case BLK_ZO_OPEN:
+ ret = qcow2_open_zone(bs, index);
+ break;
+ case BLK_ZO_CLOSE:
+ ret = qcow2_close_zone(bs, index);
+ break;
+ case BLK_ZO_FINISH:
+ ret = qcow2_finish_zone(bs, index);
+ break;
+ case BLK_ZO_RESET:
+ ret = qcow2_reset_zone(bs, index, len);
+ break;
+ default:
+ error_report("Unsupported zone op: 0x%x", op);
+ ret = -ENOTSUP;
+ break;
+ }
+ return ret;
+}
+
+static int coroutine_fn GRAPH_RDLOCK
+qcow2_co_zone_append(BlockDriverState *bs, int64_t *offset, QEMUIOVector *qiov,
+ BdrvRequestFlags flags)
+{
+ assert(flags == 0);
+ int64_t capacity = bs->total_sectors << BDRV_SECTOR_BITS;
+ int64_t zone_size_mask = bs->bl.zone_size - 1;
+ int64_t iov_len = 0;
+ int64_t len = 0;
+
+ if (*offset >= capacity) {
+ error_report("*offset %" PRId64 " is equal to or greater than the"
+ "device capacity %" PRId64 "", *offset, capacity);
+ return -EINVAL;
+ }
+
+ /* offset + len should not pass the end of that zone starting from offset */
+ if (*offset & zone_size_mask) {
+ error_report("sector offset %" PRId64 " is not aligned to zone size "
+ "%" PRId64 "", *offset / 512, bs->bl.zone_size / 512);
+ return -EINVAL;
+ }
+
+ int64_t wg = bs->bl.write_granularity;
+ int64_t wg_mask = wg - 1;
+ for (int i = 0; i < qiov->niov; i++) {
+ iov_len = qiov->iov[i].iov_len;
+ if (iov_len & wg_mask) {
+ error_report("len of IOVector[%d] 0x%" PRIx64 " is not aligned to "
+ "block size 0x%" PRIx64 "", i, iov_len, wg);
+ return -EINVAL;
+ }
+ }
+ len = qiov->size;
+
+ if ((len >> BDRV_SECTOR_BITS) > bs->bl.max_append_sectors) {
+ error_report("len 0x%" PRIx64 " in sectors is greater than "
+ "max_append_sectors 0x%" PRIx32 "",
+ len >> BDRV_SECTOR_BITS, bs->bl.max_append_sectors);
+ return -EINVAL;
+ }
+
+ return qcow2_co_pwv_part(bs, offset, len, qiov, 0, true, 0);
+}
+
static int coroutine_fn GRAPH_RDLOCK
qcow2_co_copy_range_from(BlockDriverState *bs,
BdrvChild *src, int64_t src_offset,
@@ -6643,6 +7766,10 @@ BlockDriver bdrv_qcow2 = {
.bdrv_co_pwritev_compressed_part = qcow2_co_pwritev_compressed_part,
.bdrv_make_empty = qcow2_make_empty,
+ .bdrv_co_zone_report = qcow2_co_zone_report,
+ .bdrv_co_zone_mgmt = qcow2_co_zone_mgmt,
+ .bdrv_co_zone_append = qcow2_co_zone_append,
+
.bdrv_snapshot_create = qcow2_snapshot_create,
.bdrv_snapshot_goto = qcow2_snapshot_goto,
.bdrv_snapshot_delete = qcow2_snapshot_delete,
diff --git a/block/trace-events b/block/trace-events
index 950c82d4b8..30a3e303ca 100644
--- a/block/trace-events
+++ b/block/trace-events
@@ -76,6 +76,8 @@ qcow2_writev_data(void *co, uint64_t offset) "co %p offset 0x%" PRIx64
qcow2_pwrite_zeroes_start_req(void *co, int64_t offset, int64_t bytes) "co %p offset 0x%" PRIx64 " bytes %" PRId64
qcow2_pwrite_zeroes(void *co, int64_t offset, int64_t bytes) "co %p offset 0x%" PRIx64 " bytes %" PRId64
qcow2_skip_cow(void *co, uint64_t offset, int nb_clusters) "co %p offset 0x%" PRIx64 " nb_clusters %d"
+qcow2_wp_tracking(int index, uint64_t wp) "wps[%d]: 0x%" PRIx64
+qcow2_imp_open_zones(uint8_t op, int nrz) "nr_imp_open_zones after op 0x%x: %d"
# qcow2-cluster.c
qcow2_alloc_clusters_offset(void *co, uint64_t offset, int bytes) "co %p offset 0x%" PRIx64 " bytes %d"
--
2.53.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* Re: [PATCH v11 4/5] qcow2: add zoned emulation capability
2026-06-01 21:44 ` [PATCH v11 4/5] qcow2: add zoned emulation capability Sam Li
@ 2026-06-04 20:51 ` Stefan Hajnoczi
0 siblings, 0 replies; 13+ messages in thread
From: Stefan Hajnoczi @ 2026-06-04 20:51 UTC (permalink / raw)
To: Sam Li
Cc: qemu-devel, Markus Armbruster, qemu-block, Eric Blake,
Pierrick Bouvier, dmitry.fomichev, Hanna Reitz, hare,
Michael S. Tsirkin, Kevin Wolf, cassel, dlemoal
[-- Attachment #1: Type: text/plain, Size: 10650 bytes --]
On Mon, Jun 01, 2026 at 11:44:04PM +0200, Sam Li wrote:
> By adding zone operations and zoned metadata, the zoned emulation
> capability enables full emulation support of zoned device using
> a qcow2 file. The zoned device metadata includes zone type,
> zoned device state and write pointer (WP) of each zone, which is
> stored to an array of unsigned integers.
>
> WP accessor (qcow2_rw_wp_at) routes reads and writes of an 8-byte
> WP slot through the write pointer cache. The write pointer cache is
> written to disk after the qcow2 metadata is written, thus guaranteeing
> that the write pointer is updated after the corresponding data is
> written. Per-completion cache flush is deferred. The WP cluster reaches
> disk on the next flush.
>
> Each zone of a zoned device makes state transitions following
> the zone state machine. The zone state machine mainly describes
> five states, IMPLICIT OPEN, EXPLICIT OPEN, FULL, EMPTY and CLOSED.
> READ ONLY and OFFLINE states will generally be affected by device
> internal events. The operations on zones cause corresponding state
> changing.
>
> Zoned devices have limits on zone resources, which put constraints on
> write operations on zones. It is managed by active zone queues
> following LRU policy.
>
> Signed-off-by: Sam Li <faithilikerun@gmail.com>
> ---
> block/qcow2-cache.c | 8 +
> block/qcow2-refcount.c | 7 +
> block/qcow2.c | 1137 +++++++++++++++++++++++++++++++++++++++-
> block/trace-events | 2 +
> 4 files changed, 1149 insertions(+), 5 deletions(-)
>
> diff --git a/block/qcow2-cache.c b/block/qcow2-cache.c
> index 23d9588b08..bdfb11ce88 100644
> --- a/block/qcow2-cache.c
> +++ b/block/qcow2-cache.c
> @@ -275,6 +275,14 @@ int qcow2_cache_set_dependency(BlockDriverState *bs, Qcow2Cache *c,
> {
> int ret;
>
> + /*
> + * If the dependency graph is unchanged, nothing to do. This avoids
> + * a synchronous flush on every call below.
> + */
> + if (c->depends == dependency) {
> + return 0;
> + }
This makes part of the expression below tautologous:
if (c->depends && (c->depends != dependency)) {
^^^^^^^^^^^^^^^^^^^^^^^^^^
ret = qcow2_cache_flush_dependency(bs, c);
if (ret < 0) {
return ret;
}
}
That sub-expression could be dropped, but it makes me worry that the
earlier if (dependency->depends) statement is needed even when
c->depends == dependency.
Kevin: Any thoughts on this?
> +
> if (dependency->depends) {
> ret = qcow2_cache_flush_dependency(bs, dependency);
> if (ret < 0) {
> diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
> index 6512cda407..f551726609 100644
> --- a/block/qcow2-refcount.c
> +++ b/block/qcow2-refcount.c
> @@ -1239,6 +1239,13 @@ int qcow2_write_caches(BlockDriverState *bs)
> }
> }
>
> + if (s->wp_cache) {
> + ret = qcow2_cache_write(bs, s->wp_cache);
> + if (ret < 0) {
> + return ret;
> + }
> + }
> +
> return 0;
> }
>
> diff --git a/block/qcow2.c b/block/qcow2.c
> index 29eec33e34..bdc8923b71 100644
> --- a/block/qcow2.c
> +++ b/block/qcow2.c
> @@ -195,6 +195,300 @@ qcow2_extract_crypto_opts(QemuOpts *opts, const char *fmt, Error **errp)
> return cryptoopts_qdict;
> }
>
> +#define QCOW2_ZT_IS_CONV(wp) (wp & 1ULL << 59)
> +#define QCOW2_GET_WP(wp) ((wp << 5) >> 5)
> +
> +/*
> + * To emulate a real zoned device, closed, empty and full states are
> + * preserved after a power cycle. The open states are in-memory and will
> + * be lost after closing the device. Read-only and offline states are
> + * device-internal events, which are not considered for simplicity.
> + */
> +static inline BlockZoneState qcow2_get_zone_state(BlockDriverState *bs,
> + uint32_t index)
I guess this function requires bs->wps->colock or s->lock, otherwise the
TAILQ accesses could race? Please check and document the locking
requirements.
I/O requests may be processed in multiple threads simultaneously.
s->lock protects qcow2 state.
> +{
> + BDRVQcow2State *s = bs->opaque;
> + Qcow2ZoneListEntry *zone_entry = &s->zone_list_entries[index];
> + uint64_t zone_wp = bs->wps->wp[index];
> + uint64_t zone_start;
> +
> + if (QCOW2_ZT_IS_CONV(zone_wp)) {
> + return BLK_ZS_NOT_WP;
> + }
> +
> + if (QTAILQ_IN_USE(zone_entry, exp_open_zone_entry)) {
> + return BLK_ZS_EOPEN;
> + }
> + if (QTAILQ_IN_USE(zone_entry, imp_open_zone_entry)) {
> + return BLK_ZS_IOPEN;
> + }
> +
> + zone_start = index * bs->bl.zone_size;
This is a uint32_t * uint32_t multiplication that can overflow. Avoid
that with:
zone_start = (uint64_t)index * bs->bl.zone_size;
> + if (zone_wp == zone_start) {
> + return BLK_ZS_EMPTY;
> + }
> + if (zone_wp >= zone_start + bs->bl.zone_capacity) {
> + return BLK_ZS_FULL;
> + }
> + if (zone_wp > zone_start) {
> + if (!QTAILQ_IN_USE(zone_entry, closed_zone_entry)) {
> + /*
> + * The number of closed zones is not always updated in time when
> + * the device is closed. However, it only matters when doing
> + * zone report. Refresh the count and list of closed zones to
> + * provide correct zone states for zone report.
> + */
> + QTAILQ_INSERT_HEAD(&s->closed_zones, zone_entry, closed_zone_entry);
> + s->nr_zones_closed++;
> + }
> + return BLK_ZS_CLOSED;
> + }
> + return BLK_ZS_NOT_WP;
> +}
> +
> +static void qcow2_rm_exp_open_zone(BDRVQcow2State *s,
> + uint32_t index)
Locking requirements here and in the functions that follow?
> +{
> + Qcow2ZoneListEntry *zone_entry = &s->zone_list_entries[index];
> +
> + QTAILQ_REMOVE(&s->exp_open_zones, zone_entry, exp_open_zone_entry);
> + s->nr_zones_exp_open--;
> +}
> +
> +static void qcow2_rm_imp_open_zone(BDRVQcow2State *s,
> + int32_t index)
> +{
> + Qcow2ZoneListEntry *zone_entry;
> + if (index < 0) {
> + /* Apply LRU when the index is not specified. */
> + zone_entry = QTAILQ_LAST(&s->imp_open_zones);
> + } else {
> + zone_entry = &s->zone_list_entries[index];
> + }
> +
> + QTAILQ_REMOVE(&s->imp_open_zones, zone_entry, imp_open_zone_entry);
> + s->nr_zones_imp_open--;
> +}
> +
> +static void qcow2_rm_open_zone(BDRVQcow2State *s,
> + uint32_t index)
> +{
> + Qcow2ZoneListEntry *zone_entry = &s->zone_list_entries[index];
> +
> + if (QTAILQ_IN_USE(zone_entry, exp_open_zone_entry)) {
> + qcow2_rm_exp_open_zone(s, index);
> + } else if (QTAILQ_IN_USE(zone_entry, imp_open_zone_entry)) {
> + qcow2_rm_imp_open_zone(s, index);
> + }
> +}
> +
> +static void qcow2_rm_closed_zone(BDRVQcow2State *s,
> + uint32_t index)
> +{
> + Qcow2ZoneListEntry *zone_entry = &s->zone_list_entries[index];
> +
> + QTAILQ_REMOVE(&s->closed_zones, zone_entry, closed_zone_entry);
> + s->nr_zones_closed--;
> +}
> +
> +static void qcow2_do_imp_open_zone(BDRVQcow2State *s,
> + uint32_t index,
> + BlockZoneState zs)
> +{
> + Qcow2ZoneListEntry *zone_entry = &s->zone_list_entries[index];
> +
> + switch (zs) {
> + case BLK_ZS_EMPTY:
> + break;
> + case BLK_ZS_CLOSED:
> + qcow2_rm_closed_zone(s, index);
> + break;
> + case BLK_ZS_IOPEN:
> + /*
> + * The LRU policy: update the zone that is most recently
> + * used to the head of the zone list
> + */
> + if (zone_entry == QTAILQ_FIRST(&s->imp_open_zones)) {
> + return;
> + }
> + QTAILQ_REMOVE(&s->imp_open_zones, zone_entry, imp_open_zone_entry);
> + s->nr_zones_imp_open--;
> + break;
> + default:
> + return;
> + }
> +
> + QTAILQ_INSERT_HEAD(&s->imp_open_zones, zone_entry, imp_open_zone_entry);
> + s->nr_zones_imp_open++;
> +}
> +
> +static void qcow2_do_exp_open_zone(BDRVQcow2State *s,
> + uint32_t index)
> +{
> + Qcow2ZoneListEntry *zone_entry = &s->zone_list_entries[index];
> +
> + QTAILQ_INSERT_HEAD(&s->exp_open_zones, zone_entry, exp_open_zone_entry);
> + s->nr_zones_exp_open++;
> +}
> +
> +/*
> + * The list of zones is managed using an LRU policy: the last
> + * zone of the list is always the one that was least recently used
> + * for writing and is chosen as the zone to close to be able to
> + * implicitly open another zone.
> + *
> + * We can only close the open zones. The index is not specified
> + * when it is less than 0.
> + */
> +static void qcow2_do_close_zone(BlockDriverState *bs,
> + int32_t index,
> + BlockZoneState zs)
> +{
> + BDRVQcow2State *s = bs->opaque;
> + Qcow2ZoneListEntry *zone_entry;
> +
> + if (index >= 0) {
> + zone_entry = &s->zone_list_entries[index];
> + } else {
> + /* before removal of the last implicitly open zone */
> + zone_entry = QTAILQ_LAST(&s->imp_open_zones);
There is an assumption that zone_entry is no NULL when zs ==
BLK_ZS_IOPEN? I think that make sense and it means NULL dereferences
cannot happen, but I wanted to check. You could add an assert(zone_entry
!= NULL) here to make that explicit.
> + }
> +
> + if (zs == BLK_ZS_IOPEN) {
> + qcow2_rm_imp_open_zone(s, index);
> + goto close_zone;
> + }
> +
> + if (index >= 0 && zs == BLK_ZS_EOPEN) {
> + qcow2_rm_exp_open_zone(s, index);
> + /*
> + * The zone state changes when the zone is removed from the list of
> + * open zones (explicitly open -> empty). The closed zone list is
> + * refreshed during get_zone_state().
> + */
> + qcow2_get_zone_state(bs, index);
> + }
> + return;
> +
> +close_zone:
> + QTAILQ_INSERT_HEAD(&s->closed_zones, zone_entry, closed_zone_entry);
> + s->nr_zones_closed++;
Is the goto and label necessary? Maybe move this inside the if
statement instead to simplify the function.
> +}
I've reviewed up to here for now.
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]
^ permalink raw reply [flat|nested] 13+ messages in thread
* [PATCH v11 5/5] iotests: test the zoned format feature for qcow2 file
2026-06-01 21:44 [PATCH v11 0/5] Add full zoned storage emulation to the qcow2 driver Sam Li
` (3 preceding siblings ...)
2026-06-01 21:44 ` [PATCH v11 4/5] qcow2: add zoned emulation capability Sam Li
@ 2026-06-01 21:44 ` Sam Li
2026-06-02 20:06 ` [PATCH v11 0/5] Add full zoned storage emulation to the qcow2 driver Stefan Hajnoczi
5 siblings, 0 replies; 13+ messages in thread
From: Sam Li @ 2026-06-01 21:44 UTC (permalink / raw)
To: qemu-devel
Cc: Markus Armbruster, qemu-block, Eric Blake, Stefan Hajnoczi,
Pierrick Bouvier, dmitry.fomichev, Hanna Reitz, hare,
Michael S. Tsirkin, Kevin Wolf, cassel, dlemoal, Sam Li
The zoned format feature can be tested by:
$ tests/qemu-iotests/check -qcow2 zoned-qcow2
Signed-off-by: Sam Li <faithilikerun@gmail.com>
---
tests/qemu-iotests/tests/zoned-qcow2 | 209 +++++++++++++++++++++++
tests/qemu-iotests/tests/zoned-qcow2.out | 191 +++++++++++++++++++++
2 files changed, 400 insertions(+)
create mode 100755 tests/qemu-iotests/tests/zoned-qcow2
create mode 100644 tests/qemu-iotests/tests/zoned-qcow2.out
diff --git a/tests/qemu-iotests/tests/zoned-qcow2 b/tests/qemu-iotests/tests/zoned-qcow2
new file mode 100755
index 0000000000..d37100c8ab
--- /dev/null
+++ b/tests/qemu-iotests/tests/zoned-qcow2
@@ -0,0 +1,209 @@
+#!/usr/bin/env bash
+#
+# SPDX-License-Identifier: GPL-2.0-or-later
+#
+# Test zone management operations for qcow2 file.
+#
+
+seq="$(basename $0)"
+echo "QA output created by $seq"
+status=1 # failure is the default!
+
+file_name="zbc.qcow2"
+_cleanup()
+{
+ _cleanup_test_img
+ _rm_test_img "$file_name"
+}
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+# get standard environment, filters and checks
+. ../common.rc
+. ../common.filter
+. ../common.qemu
+
+# This test only runs on Linux hosts with qcow2 image files.
+_supported_fmt qcow2
+_supported_proto file
+_supported_os Linux
+
+echo
+echo "=== Initial image setup ==="
+echo
+
+$QEMU_IMG create -f qcow2 $file_name -o size=768M -o zone.size=64M -o \
+zone.capacity=64M -o zone.conventional_zones=0 -o zone.max_append_bytes=32M \
+-o zone.max_open_zones=6 -o zone.max_active_zones=8 -o zone.mode=host-managed
+
+IMG="--image-opts -n driver=qcow2,file.driver=file,file.filename=$file_name"
+QEMU_IO_OPTIONS=$QEMU_IO_OPTIONS_NO_FMT
+
+echo
+echo "=== Testing a qcow2 img with zoned format ==="
+echo
+echo "case 1: test zone operations one by one"
+
+echo "(1) report zones[0]:"
+$QEMU_IO $IMG -c "zrp 0 1"
+echo
+echo "report zones[0~9]:"
+$QEMU_IO $IMG -c "zrp 0 10"
+echo
+echo "report zones[-1]:" # zones[-1] dictates the last zone
+$QEMU_IO $IMG -c "zrp 0x2C000000 2" # 0x2C000000 / 512 = 0x160000
+echo
+echo
+echo "(2) open zones[0], zones[1], zones[-1] then close, finish, reset:"
+$QEMU_IO $IMG << EOF
+zo 0 0x4000000
+zrp 0 1
+zo 0x4000000 0x4000000
+zrp 0x4000000 1
+zo 0x2C000000 0x4000000
+zrp 0x2C000000 2
+zc 0 0x4000000
+zrp 0 1
+zc 0x4000000 0x4000000
+zrp 0x4000000 1
+zc 0x2C000000 0x4000000
+zrp 0x2C000000 2
+zf 0 0x4000000
+zrp 0 1
+zf 64M 64M
+zrp 0x4000000 2
+zf 0x2C000000 0x4000000
+zrp 0x2C000000 2
+zrs 0 0x4000000
+zrp 0 1
+zrs 0x4000000 0x4000000
+zrp 0x4000000 1
+zrs 0x2C000000 0x4000000
+zrp 0x2C000000 2
+EOF
+
+echo
+echo "(3) append write with (4k, 8k) data"
+$QEMU_IO $IMG -c "zrp 0 12" # the physical block size of the device is 4096
+echo "Append write zones[0], zones[1] twice"
+$QEMU_IO $IMG << EOF
+zap -p 0 0x1000 0x2000
+zrp 0 1
+zap -p 0 0x1000 0x2000
+zrp 0 1
+zap -p 0x4000000 0x1000 0x2000
+zrp 0x4000000 1
+zap -p 0x4000000 0x1000 0x2000
+zrp 0x4000000 1
+EOF
+
+echo
+echo "Reset all:"
+$QEMU_IO $IMG -c "zrp 0 12" -c "zrs 0 768M" -c "zrp 0 12"
+echo
+echo
+
+echo "case 2: test a sets of ops that works or not"
+echo "(1) append write (4k, 4k) and then write to full"
+$QEMU_IO $IMG << EOF
+zap -p 0 0x1000 0x1000
+zrp 0 1
+zap -p 0 0x1000 0x1ffd000
+zap -p 0 0x1000000 0x1000000
+zrp 0 1
+EOF
+
+echo "Reset zones[0]:"
+$QEMU_IO $IMG -c "zrs 0 64M" -c "zrp 0 1"
+
+echo "(2) write in zones[0], zones[3], zones[8], and then reset all"
+$QEMU_IO $IMG << EOF
+zap -p 0 0x1000 0x1000
+zap -p 0xc000000 0x1000 0x1000
+zap -p 0x20000000 0x1000 0x1000
+zrp 0 12
+zrs 0 768M
+zrp 0 12
+EOF
+
+echo "case 3: test zone resource management"
+echo "(1) write in zones[0], zones[1], zones[2] and then close it"
+$QEMU_IO $IMG << EOF
+zap -p 0 0x1000 0x1000
+zap -p 0x4000000 0x1000 0x1000
+zap -p 0x8000000 0x1000 0x1000
+zrp 0 12
+zc 0 64M
+zc 0x4000000 64M
+zc 0x8000000 64M
+zrp 0 12
+EOF
+
+echo "(2) reset all after 3(1)"
+$QEMU_IO $IMG << EOF
+zrs 0 768M
+zrp 0 12
+EOF
+
+echo
+echo "case 4: WP cache crash consistency under concurrent appends"
+echo "(1) concurrent writes to the same sequential zone (zone 5 @ 320M)"
+# Three concurrent aio_writes at WP, WP+4K, WP+8K. Data writes are
+# permitted to complete out of order
+$QEMU_IO $IMG <<EOF
+aio_write -q -P 0xa1 0x14000000 4k
+aio_write -q -P 0xa2 0x14001000 4k
+aio_write -q -P 0xa3 0x14002000 4k
+aio_flush
+zrp 0x14000000 1
+read -q -P 0xa1 0x14000000 4k
+read -q -P 0xa2 0x14001000 4k
+read -q -P 0xa3 0x14002000 4k
+EOF
+
+echo "(2) concurrent writes to different sequential zones (zones 6, 7, 9)"
+# Spread across three zones. No per-zone interaction; each zone's WP
+# must advance independently.
+$QEMU_IO $IMG <<EOF
+aio_write -q -P 0xb1 0x18000000 4k
+aio_write -q -P 0xb2 0x1C000000 4k
+aio_write -q -P 0xb3 0x24000000 4k
+aio_flush
+zrp 0x18000000 1
+zrp 0x1C000000 1
+zrp 0x24000000 1
+read -q -P 0xb1 0x18000000 4k
+read -q -P 0xb2 0x1C000000 4k
+read -q -P 0xb3 0x24000000 4k
+EOF
+
+echo "(3) reset zones with no in-flight writes (zones 5, 6)"
+# Reset zones 5 and 6 (which we wrote in (1) and (2)). WP should drop
+# back to zone_start; data reads must return zero (unallocated
+# post-reset).
+$QEMU_IO $IMG -c "zrs 0x14000000 64M" -c "zrs 0x18000000 64M" \
+ -c "zrp 0x14000000 1" -c "zrp 0x18000000 1" \
+ -c "read -q -P 0 0x14000000 4k" \
+ -c "read -q -P 0 0x18000000 4k"
+
+echo "(4) BLKRESETALL after concurrent writes"
+# Reset everything first, refill three zones with concurrent writes,
+# then RESETALL. Each zone's WP must return to its start.
+$QEMU_IO $IMG <<EOF
+zrs 0 768M
+aio_write -q -P 0xc1 0x14000000 4k
+aio_write -q -P 0xc2 0x18000000 4k
+aio_write -q -P 0xc3 0x1C000000 4k
+aio_flush
+zrp 0x14000000 1
+zrp 0x18000000 1
+zrp 0x1C000000 1
+zrs 0 768M
+zrp 0x14000000 1
+zrp 0x18000000 1
+zrp 0x1C000000 1
+EOF
+
+# success, all done
+echo "*** done"
+rm -f $seq.full
+status=0
diff --git a/tests/qemu-iotests/tests/zoned-qcow2.out b/tests/qemu-iotests/tests/zoned-qcow2.out
new file mode 100644
index 0000000000..2e2c65b6c0
--- /dev/null
+++ b/tests/qemu-iotests/tests/zoned-qcow2.out
@@ -0,0 +1,191 @@
+QA output created by zoned-qcow2
+
+=== Initial image setup ===
+
+Formatting 'zbc.qcow2', fmt=qcow2 cluster_size=65536 extended_l2=off compression_type=zlib zone.mode=host-managed zone.size=67108864 zone.capacity=67108864 zone.conventional_zones=0 zone.max_append_bytes=33554432 zone.max_active_zones=8 zone.max_open_zones=6 size=805306368 lazy_refcounts=off refcount_bits=16
+
+=== Testing a qcow2 img with zoned format ===
+
+case 1: test zone operations one by one
+(1) report zones[0]:
+start: 0x0, len 0x20000, cap 0x20000, wptr 0x0, zcond:1, [type: 2]
+
+report zones[0~9]:
+start: 0x0, len 0x20000, cap 0x20000, wptr 0x0, zcond:1, [type: 2]
+start: 0x20000, len 0x20000, cap 0x20000, wptr 0x20000, zcond:1, [type: 2]
+start: 0x40000, len 0x20000, cap 0x20000, wptr 0x40000, zcond:1, [type: 2]
+start: 0x60000, len 0x20000, cap 0x20000, wptr 0x60000, zcond:1, [type: 2]
+start: 0x80000, len 0x20000, cap 0x20000, wptr 0x80000, zcond:1, [type: 2]
+start: 0xa0000, len 0x20000, cap 0x20000, wptr 0xa0000, zcond:1, [type: 2]
+start: 0xc0000, len 0x20000, cap 0x20000, wptr 0xc0000, zcond:1, [type: 2]
+start: 0xe0000, len 0x20000, cap 0x20000, wptr 0xe0000, zcond:1, [type: 2]
+start: 0x100000, len 0x20000, cap 0x20000, wptr 0x100000, zcond:1, [type: 2]
+start: 0x120000, len 0x20000, cap 0x20000, wptr 0x120000, zcond:1, [type: 2]
+
+report zones[-1]:
+start: 0x160000, len 0x20000, cap 0x20000, wptr 0x160000, zcond:1, [type: 2]
+
+
+(2) open zones[0], zones[1], zones[-1] then close, finish, reset:
+qemu-io> qemu-io> start: 0x0, len 0x20000, cap 0x20000, wptr 0x0, zcond:3, [type: 2]
+qemu-io> qemu-io> start: 0x20000, len 0x20000, cap 0x20000, wptr 0x20000, zcond:3, [type: 2]
+qemu-io> qemu-io> start: 0x160000, len 0x20000, cap 0x20000, wptr 0x160000, zcond:3, [type: 2]
+qemu-io> qemu-io> start: 0x0, len 0x20000, cap 0x20000, wptr 0x0, zcond:1, [type: 2]
+qemu-io> qemu-io> start: 0x20000, len 0x20000, cap 0x20000, wptr 0x20000, zcond:1, [type: 2]
+qemu-io> qemu-io> start: 0x160000, len 0x20000, cap 0x20000, wptr 0x160000, zcond:1, [type: 2]
+qemu-io> qemu-io> start: 0x0, len 0x20000, cap 0x20000, wptr 0x20000, zcond:14, [type: 2]
+qemu-io> qemu-io> start: 0x20000, len 0x20000, cap 0x20000, wptr 0x40000, zcond:14, [type: 2]
+start: 0x40000, len 0x20000, cap 0x20000, wptr 0x40000, zcond:1, [type: 2]
+qemu-io> qemu-io> start: 0x160000, len 0x20000, cap 0x20000, wptr 0x180000, zcond:14, [type: 2]
+qemu-io> qemu-io> start: 0x0, len 0x20000, cap 0x20000, wptr 0x0, zcond:1, [type: 2]
+qemu-io> qemu-io> start: 0x20000, len 0x20000, cap 0x20000, wptr 0x20000, zcond:1, [type: 2]
+qemu-io> qemu-io> start: 0x160000, len 0x20000, cap 0x20000, wptr 0x160000, zcond:1, [type: 2]
+qemu-io>
+(3) append write with (4k, 8k) data
+start: 0x0, len 0x20000, cap 0x20000, wptr 0x0, zcond:1, [type: 2]
+start: 0x20000, len 0x20000, cap 0x20000, wptr 0x20000, zcond:1, [type: 2]
+start: 0x40000, len 0x20000, cap 0x20000, wptr 0x40000, zcond:1, [type: 2]
+start: 0x60000, len 0x20000, cap 0x20000, wptr 0x60000, zcond:1, [type: 2]
+start: 0x80000, len 0x20000, cap 0x20000, wptr 0x80000, zcond:1, [type: 2]
+start: 0xa0000, len 0x20000, cap 0x20000, wptr 0xa0000, zcond:1, [type: 2]
+start: 0xc0000, len 0x20000, cap 0x20000, wptr 0xc0000, zcond:1, [type: 2]
+start: 0xe0000, len 0x20000, cap 0x20000, wptr 0xe0000, zcond:1, [type: 2]
+start: 0x100000, len 0x20000, cap 0x20000, wptr 0x100000, zcond:1, [type: 2]
+start: 0x120000, len 0x20000, cap 0x20000, wptr 0x120000, zcond:1, [type: 2]
+start: 0x140000, len 0x20000, cap 0x20000, wptr 0x140000, zcond:1, [type: 2]
+start: 0x160000, len 0x20000, cap 0x20000, wptr 0x160000, zcond:1, [type: 2]
+Append write zones[0], zones[1] twice
+qemu-io> After zap done, the append sector is 0x0
+qemu-io> start: 0x0, len 0x20000, cap 0x20000, wptr 0x18, zcond:2, [type: 2]
+qemu-io> After zap done, the append sector is 0x18
+qemu-io> start: 0x0, len 0x20000, cap 0x20000, wptr 0x30, zcond:2, [type: 2]
+qemu-io> After zap done, the append sector is 0x20000
+qemu-io> start: 0x20000, len 0x20000, cap 0x20000, wptr 0x20018, zcond:2, [type: 2]
+qemu-io> After zap done, the append sector is 0x20018
+qemu-io> start: 0x20000, len 0x20000, cap 0x20000, wptr 0x20030, zcond:2, [type: 2]
+qemu-io>
+Reset all:
+start: 0x0, len 0x20000, cap 0x20000, wptr 0x30, zcond:4, [type: 2]
+start: 0x20000, len 0x20000, cap 0x20000, wptr 0x20030, zcond:4, [type: 2]
+start: 0x40000, len 0x20000, cap 0x20000, wptr 0x40000, zcond:1, [type: 2]
+start: 0x60000, len 0x20000, cap 0x20000, wptr 0x60000, zcond:1, [type: 2]
+start: 0x80000, len 0x20000, cap 0x20000, wptr 0x80000, zcond:1, [type: 2]
+start: 0xa0000, len 0x20000, cap 0x20000, wptr 0xa0000, zcond:1, [type: 2]
+start: 0xc0000, len 0x20000, cap 0x20000, wptr 0xc0000, zcond:1, [type: 2]
+start: 0xe0000, len 0x20000, cap 0x20000, wptr 0xe0000, zcond:1, [type: 2]
+start: 0x100000, len 0x20000, cap 0x20000, wptr 0x100000, zcond:1, [type: 2]
+start: 0x120000, len 0x20000, cap 0x20000, wptr 0x120000, zcond:1, [type: 2]
+start: 0x140000, len 0x20000, cap 0x20000, wptr 0x140000, zcond:1, [type: 2]
+start: 0x160000, len 0x20000, cap 0x20000, wptr 0x160000, zcond:1, [type: 2]
+start: 0x0, len 0x20000, cap 0x20000, wptr 0x0, zcond:1, [type: 2]
+start: 0x20000, len 0x20000, cap 0x20000, wptr 0x20000, zcond:1, [type: 2]
+start: 0x40000, len 0x20000, cap 0x20000, wptr 0x40000, zcond:1, [type: 2]
+start: 0x60000, len 0x20000, cap 0x20000, wptr 0x60000, zcond:1, [type: 2]
+start: 0x80000, len 0x20000, cap 0x20000, wptr 0x80000, zcond:1, [type: 2]
+start: 0xa0000, len 0x20000, cap 0x20000, wptr 0xa0000, zcond:1, [type: 2]
+start: 0xc0000, len 0x20000, cap 0x20000, wptr 0xc0000, zcond:1, [type: 2]
+start: 0xe0000, len 0x20000, cap 0x20000, wptr 0xe0000, zcond:1, [type: 2]
+start: 0x100000, len 0x20000, cap 0x20000, wptr 0x100000, zcond:1, [type: 2]
+start: 0x120000, len 0x20000, cap 0x20000, wptr 0x120000, zcond:1, [type: 2]
+start: 0x140000, len 0x20000, cap 0x20000, wptr 0x140000, zcond:1, [type: 2]
+start: 0x160000, len 0x20000, cap 0x20000, wptr 0x160000, zcond:1, [type: 2]
+
+
+case 2: test a sets of ops that works or not
+(1) append write (4k, 4k) and then write to full
+qemu-io> After zap done, the append sector is 0x0
+qemu-io> start: 0x0, len 0x20000, cap 0x20000, wptr 0x10, zcond:2, [type: 2]
+qemu-io> After zap done, the append sector is 0x10
+qemu-io> After zap done, the append sector is 0x10000
+qemu-io> start: 0x0, len 0x20000, cap 0x20000, wptr 0x20000, zcond:14, [type: 2]
+qemu-io> Reset zones[0]:
+start: 0x0, len 0x20000, cap 0x20000, wptr 0x0, zcond:1, [type: 2]
+(2) write in zones[0], zones[3], zones[8], and then reset all
+qemu-io> After zap done, the append sector is 0x0
+qemu-io> After zap done, the append sector is 0x60000
+qemu-io> After zap done, the append sector is 0x100000
+qemu-io> start: 0x0, len 0x20000, cap 0x20000, wptr 0x10, zcond:2, [type: 2]
+start: 0x20000, len 0x20000, cap 0x20000, wptr 0x20000, zcond:1, [type: 2]
+start: 0x40000, len 0x20000, cap 0x20000, wptr 0x40000, zcond:1, [type: 2]
+start: 0x60000, len 0x20000, cap 0x20000, wptr 0x60010, zcond:2, [type: 2]
+start: 0x80000, len 0x20000, cap 0x20000, wptr 0x80000, zcond:1, [type: 2]
+start: 0xa0000, len 0x20000, cap 0x20000, wptr 0xa0000, zcond:1, [type: 2]
+start: 0xc0000, len 0x20000, cap 0x20000, wptr 0xc0000, zcond:1, [type: 2]
+start: 0xe0000, len 0x20000, cap 0x20000, wptr 0xe0000, zcond:1, [type: 2]
+start: 0x100000, len 0x20000, cap 0x20000, wptr 0x100010, zcond:2, [type: 2]
+start: 0x120000, len 0x20000, cap 0x20000, wptr 0x120000, zcond:1, [type: 2]
+start: 0x140000, len 0x20000, cap 0x20000, wptr 0x140000, zcond:1, [type: 2]
+start: 0x160000, len 0x20000, cap 0x20000, wptr 0x160000, zcond:1, [type: 2]
+qemu-io> qemu-io> start: 0x0, len 0x20000, cap 0x20000, wptr 0x0, zcond:1, [type: 2]
+start: 0x20000, len 0x20000, cap 0x20000, wptr 0x20000, zcond:1, [type: 2]
+start: 0x40000, len 0x20000, cap 0x20000, wptr 0x40000, zcond:1, [type: 2]
+start: 0x60000, len 0x20000, cap 0x20000, wptr 0x60000, zcond:1, [type: 2]
+start: 0x80000, len 0x20000, cap 0x20000, wptr 0x80000, zcond:1, [type: 2]
+start: 0xa0000, len 0x20000, cap 0x20000, wptr 0xa0000, zcond:1, [type: 2]
+start: 0xc0000, len 0x20000, cap 0x20000, wptr 0xc0000, zcond:1, [type: 2]
+start: 0xe0000, len 0x20000, cap 0x20000, wptr 0xe0000, zcond:1, [type: 2]
+start: 0x100000, len 0x20000, cap 0x20000, wptr 0x100000, zcond:1, [type: 2]
+start: 0x120000, len 0x20000, cap 0x20000, wptr 0x120000, zcond:1, [type: 2]
+start: 0x140000, len 0x20000, cap 0x20000, wptr 0x140000, zcond:1, [type: 2]
+start: 0x160000, len 0x20000, cap 0x20000, wptr 0x160000, zcond:1, [type: 2]
+qemu-io> case 3: test zone resource management
+(1) write in zones[0], zones[1], zones[2] and then close it
+qemu-io> After zap done, the append sector is 0x0
+qemu-io> After zap done, the append sector is 0x20000
+qemu-io> After zap done, the append sector is 0x40000
+qemu-io> start: 0x0, len 0x20000, cap 0x20000, wptr 0x10, zcond:2, [type: 2]
+start: 0x20000, len 0x20000, cap 0x20000, wptr 0x20010, zcond:2, [type: 2]
+start: 0x40000, len 0x20000, cap 0x20000, wptr 0x40010, zcond:2, [type: 2]
+start: 0x60000, len 0x20000, cap 0x20000, wptr 0x60000, zcond:1, [type: 2]
+start: 0x80000, len 0x20000, cap 0x20000, wptr 0x80000, zcond:1, [type: 2]
+start: 0xa0000, len 0x20000, cap 0x20000, wptr 0xa0000, zcond:1, [type: 2]
+start: 0xc0000, len 0x20000, cap 0x20000, wptr 0xc0000, zcond:1, [type: 2]
+start: 0xe0000, len 0x20000, cap 0x20000, wptr 0xe0000, zcond:1, [type: 2]
+start: 0x100000, len 0x20000, cap 0x20000, wptr 0x100000, zcond:1, [type: 2]
+start: 0x120000, len 0x20000, cap 0x20000, wptr 0x120000, zcond:1, [type: 2]
+start: 0x140000, len 0x20000, cap 0x20000, wptr 0x140000, zcond:1, [type: 2]
+start: 0x160000, len 0x20000, cap 0x20000, wptr 0x160000, zcond:1, [type: 2]
+qemu-io> qemu-io> qemu-io> qemu-io> start: 0x0, len 0x20000, cap 0x20000, wptr 0x10, zcond:4, [type: 2]
+start: 0x20000, len 0x20000, cap 0x20000, wptr 0x20010, zcond:4, [type: 2]
+start: 0x40000, len 0x20000, cap 0x20000, wptr 0x40010, zcond:4, [type: 2]
+start: 0x60000, len 0x20000, cap 0x20000, wptr 0x60000, zcond:1, [type: 2]
+start: 0x80000, len 0x20000, cap 0x20000, wptr 0x80000, zcond:1, [type: 2]
+start: 0xa0000, len 0x20000, cap 0x20000, wptr 0xa0000, zcond:1, [type: 2]
+start: 0xc0000, len 0x20000, cap 0x20000, wptr 0xc0000, zcond:1, [type: 2]
+start: 0xe0000, len 0x20000, cap 0x20000, wptr 0xe0000, zcond:1, [type: 2]
+start: 0x100000, len 0x20000, cap 0x20000, wptr 0x100000, zcond:1, [type: 2]
+start: 0x120000, len 0x20000, cap 0x20000, wptr 0x120000, zcond:1, [type: 2]
+start: 0x140000, len 0x20000, cap 0x20000, wptr 0x140000, zcond:1, [type: 2]
+start: 0x160000, len 0x20000, cap 0x20000, wptr 0x160000, zcond:1, [type: 2]
+qemu-io> (2) reset all after 3(1)
+qemu-io> qemu-io> start: 0x0, len 0x20000, cap 0x20000, wptr 0x0, zcond:1, [type: 2]
+start: 0x20000, len 0x20000, cap 0x20000, wptr 0x20000, zcond:1, [type: 2]
+start: 0x40000, len 0x20000, cap 0x20000, wptr 0x40000, zcond:1, [type: 2]
+start: 0x60000, len 0x20000, cap 0x20000, wptr 0x60000, zcond:1, [type: 2]
+start: 0x80000, len 0x20000, cap 0x20000, wptr 0x80000, zcond:1, [type: 2]
+start: 0xa0000, len 0x20000, cap 0x20000, wptr 0xa0000, zcond:1, [type: 2]
+start: 0xc0000, len 0x20000, cap 0x20000, wptr 0xc0000, zcond:1, [type: 2]
+start: 0xe0000, len 0x20000, cap 0x20000, wptr 0xe0000, zcond:1, [type: 2]
+start: 0x100000, len 0x20000, cap 0x20000, wptr 0x100000, zcond:1, [type: 2]
+start: 0x120000, len 0x20000, cap 0x20000, wptr 0x120000, zcond:1, [type: 2]
+start: 0x140000, len 0x20000, cap 0x20000, wptr 0x140000, zcond:1, [type: 2]
+start: 0x160000, len 0x20000, cap 0x20000, wptr 0x160000, zcond:1, [type: 2]
+qemu-io>
+case 4: WP cache crash consistency under concurrent appends
+(1) concurrent writes to the same sequential zone (zone 5 @ 320M)
+qemu-io> qemu-io> qemu-io> qemu-io> qemu-io> start: 0xa0000, len 0x20000, cap 0x20000, wptr 0xa0018, zcond:2, [type: 2]
+qemu-io> qemu-io> qemu-io> qemu-io> (2) concurrent writes to different sequential zones (zones 6, 7, 9)
+qemu-io> qemu-io> qemu-io> qemu-io> qemu-io> start: 0xc0000, len 0x20000, cap 0x20000, wptr 0xc0008, zcond:2, [type: 2]
+qemu-io> start: 0xe0000, len 0x20000, cap 0x20000, wptr 0xe0008, zcond:2, [type: 2]
+qemu-io> start: 0x120000, len 0x20000, cap 0x20000, wptr 0x120008, zcond:2, [type: 2]
+qemu-io> qemu-io> qemu-io> qemu-io> (3) reset zones with no in-flight writes (zones 5, 6)
+start: 0xa0000, len 0x20000, cap 0x20000, wptr 0xa0000, zcond:1, [type: 2]
+start: 0xc0000, len 0x20000, cap 0x20000, wptr 0xc0000, zcond:1, [type: 2]
+(4) BLKRESETALL after concurrent writes
+qemu-io> qemu-io> qemu-io> qemu-io> qemu-io> qemu-io> start: 0xa0000, len 0x20000, cap 0x20000, wptr 0xa0008, zcond:2, [type: 2]
+qemu-io> start: 0xc0000, len 0x20000, cap 0x20000, wptr 0xc0008, zcond:2, [type: 2]
+qemu-io> start: 0xe0000, len 0x20000, cap 0x20000, wptr 0xe0008, zcond:2, [type: 2]
+qemu-io> qemu-io> start: 0xa0000, len 0x20000, cap 0x20000, wptr 0xa0000, zcond:1, [type: 2]
+qemu-io> start: 0xc0000, len 0x20000, cap 0x20000, wptr 0xc0000, zcond:1, [type: 2]
+qemu-io> start: 0xe0000, len 0x20000, cap 0x20000, wptr 0xe0000, zcond:1, [type: 2]
+qemu-io> *** done
--
2.53.0
^ permalink raw reply related [flat|nested] 13+ messages in thread* Re: [PATCH v11 0/5] Add full zoned storage emulation to the qcow2 driver
2026-06-01 21:44 [PATCH v11 0/5] Add full zoned storage emulation to the qcow2 driver Sam Li
` (4 preceding siblings ...)
2026-06-01 21:44 ` [PATCH v11 5/5] iotests: test the zoned format feature for qcow2 file Sam Li
@ 2026-06-02 20:06 ` Stefan Hajnoczi
2026-06-02 20:07 ` Stefan Hajnoczi
2026-06-02 20:10 ` Sam Li
5 siblings, 2 replies; 13+ messages in thread
From: Stefan Hajnoczi @ 2026-06-02 20:06 UTC (permalink / raw)
To: Sam Li
Cc: qemu-devel, Markus Armbruster, qemu-block, Eric Blake,
Pierrick Bouvier, dmitry.fomichev, Hanna Reitz, hare,
Michael S. Tsirkin, Kevin Wolf, cassel, dlemoal
[-- Attachment #1: Type: text/plain, Size: 2090 bytes --]
On Mon, Jun 01, 2026 at 11:44:00PM +0200, Sam Li wrote:
> This patch series add a new extension - zoned format - to the
> qcow2 driver, allowing full zoned storage emulation on a qcow2
> image file. A user can attach such an image to a guest and have
> it appear as a host-managed zoned block device.
>
> The zoned format is opt-in through a new qcow2 header extension
> that pins the zone geometry. Behind the extension is a dedicated
> zoned metadata region that stores one 8-byte write pointer (WP)
> per zone. The extension is gated by an incompatible bit, so an
> older qcow2 implementation cannot accidentally open the image.
>
> Each write pointer is routed through the write pointer cache,
> a Qcow2Cache object. The write pointer cache is written to disk
> after the qcow2 metadata is written, thus guaranteeing that
> the write pointer is updated after the corresponding data is
> written.
>
> Zone states are in memory. Read-only and offline states are
> device-internal events, which are not modelled in qcow2
> emulation for simplicity. The other zone states
> (closed, empty, full) can be inferred from write poiner
> values, presistent across QEMU reboots. The open states are
> kept in memory using open zone lists.
>
> To create a qcow2 file with the zoned format:
>
> qemu-img create -f qcow2 zbc.qcow2 \
> -o size=768M \
> -o zone.size=64M \
> -o zone.capacity=64M \
> -o zone.conventional_zones=0 \
> -o zone.max_append_bytes=4096 \
> -o zone.max_open_zones=6 \
> -o zone.max_active_zones=8 \
> -o zone.mode=host-managed
>
> Then attach it to a guest via the QEMU command line:
> -blockdev node-name=drive1,driver=qcow2,\
> file.driver=file,file.filename=zbc.qcow2 \
> -device virtio-blk-pci,drive=drive1 \
Please run `make check-block-qcow` and ensure the tests pass. In some
cases the test reference output may be outdated and you can simply copy
the <test>.out.bad file over the <test>.out reference file.
Thanks,
Stefan
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: [PATCH v11 0/5] Add full zoned storage emulation to the qcow2 driver
2026-06-02 20:06 ` [PATCH v11 0/5] Add full zoned storage emulation to the qcow2 driver Stefan Hajnoczi
@ 2026-06-02 20:07 ` Stefan Hajnoczi
2026-06-02 20:10 ` Sam Li
1 sibling, 0 replies; 13+ messages in thread
From: Stefan Hajnoczi @ 2026-06-02 20:07 UTC (permalink / raw)
To: Sam Li
Cc: qemu-devel, Markus Armbruster, qemu-block, Eric Blake,
Pierrick Bouvier, dmitry.fomichev, Hanna Reitz, hare,
Michael S. Tsirkin, Kevin Wolf, cassel, dlemoal
[-- Attachment #1: Type: text/plain, Size: 2133 bytes --]
On Tue, Jun 02, 2026 at 04:06:38PM -0400, Stefan Hajnoczi wrote:
> On Mon, Jun 01, 2026 at 11:44:00PM +0200, Sam Li wrote:
> > This patch series add a new extension - zoned format - to the
> > qcow2 driver, allowing full zoned storage emulation on a qcow2
> > image file. A user can attach such an image to a guest and have
> > it appear as a host-managed zoned block device.
> >
> > The zoned format is opt-in through a new qcow2 header extension
> > that pins the zone geometry. Behind the extension is a dedicated
> > zoned metadata region that stores one 8-byte write pointer (WP)
> > per zone. The extension is gated by an incompatible bit, so an
> > older qcow2 implementation cannot accidentally open the image.
> >
> > Each write pointer is routed through the write pointer cache,
> > a Qcow2Cache object. The write pointer cache is written to disk
> > after the qcow2 metadata is written, thus guaranteeing that
> > the write pointer is updated after the corresponding data is
> > written.
> >
> > Zone states are in memory. Read-only and offline states are
> > device-internal events, which are not modelled in qcow2
> > emulation for simplicity. The other zone states
> > (closed, empty, full) can be inferred from write poiner
> > values, presistent across QEMU reboots. The open states are
> > kept in memory using open zone lists.
> >
> > To create a qcow2 file with the zoned format:
> >
> > qemu-img create -f qcow2 zbc.qcow2 \
> > -o size=768M \
> > -o zone.size=64M \
> > -o zone.capacity=64M \
> > -o zone.conventional_zones=0 \
> > -o zone.max_append_bytes=4096 \
> > -o zone.max_open_zones=6 \
> > -o zone.max_active_zones=8 \
> > -o zone.mode=host-managed
> >
> > Then attach it to a guest via the QEMU command line:
> > -blockdev node-name=drive1,driver=qcow2,\
> > file.driver=file,file.filename=zbc.qcow2 \
> > -device virtio-blk-pci,drive=drive1 \
>
> Please run `make check-block-qcow` and ensure the tests pass. In some
That should be `make check-block-qcow2`.
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH v11 0/5] Add full zoned storage emulation to the qcow2 driver
2026-06-02 20:06 ` [PATCH v11 0/5] Add full zoned storage emulation to the qcow2 driver Stefan Hajnoczi
2026-06-02 20:07 ` Stefan Hajnoczi
@ 2026-06-02 20:10 ` Sam Li
1 sibling, 0 replies; 13+ messages in thread
From: Sam Li @ 2026-06-02 20:10 UTC (permalink / raw)
To: Stefan Hajnoczi
Cc: qemu-devel, Markus Armbruster, qemu-block, Eric Blake,
Pierrick Bouvier, dmitry.fomichev, Hanna Reitz, hare,
Michael S. Tsirkin, Kevin Wolf, cassel, dlemoal
On Tue, Jun 2, 2026 at 10:06 PM Stefan Hajnoczi <stefanha@redhat.com> wrote:
>
> On Mon, Jun 01, 2026 at 11:44:00PM +0200, Sam Li wrote:
> > This patch series add a new extension - zoned format - to the
> > qcow2 driver, allowing full zoned storage emulation on a qcow2
> > image file. A user can attach such an image to a guest and have
> > it appear as a host-managed zoned block device.
> >
> > The zoned format is opt-in through a new qcow2 header extension
> > that pins the zone geometry. Behind the extension is a dedicated
> > zoned metadata region that stores one 8-byte write pointer (WP)
> > per zone. The extension is gated by an incompatible bit, so an
> > older qcow2 implementation cannot accidentally open the image.
> >
> > Each write pointer is routed through the write pointer cache,
> > a Qcow2Cache object. The write pointer cache is written to disk
> > after the qcow2 metadata is written, thus guaranteeing that
> > the write pointer is updated after the corresponding data is
> > written.
> >
> > Zone states are in memory. Read-only and offline states are
> > device-internal events, which are not modelled in qcow2
> > emulation for simplicity. The other zone states
> > (closed, empty, full) can be inferred from write poiner
> > values, presistent across QEMU reboots. The open states are
> > kept in memory using open zone lists.
> >
> > To create a qcow2 file with the zoned format:
> >
> > qemu-img create -f qcow2 zbc.qcow2 \
> > -o size=768M \
> > -o zone.size=64M \
> > -o zone.capacity=64M \
> > -o zone.conventional_zones=0 \
> > -o zone.max_append_bytes=4096 \
> > -o zone.max_open_zones=6 \
> > -o zone.max_active_zones=8 \
> > -o zone.mode=host-managed
> >
> > Then attach it to a guest via the QEMU command line:
> > -blockdev node-name=drive1,driver=qcow2,\
> > file.driver=file,file.filename=zbc.qcow2 \
> > -device virtio-blk-pci,drive=drive1 \
>
> Please run `make check-block-qcow` and ensure the tests pass. In some
> cases the test reference output may be outdated and you can simply copy
> the <test>.out.bad file over the <test>.out reference file.
Thanks, I will check this.
Sam
^ permalink raw reply [flat|nested] 13+ messages in thread