Change etcd_host to etcd_address in QEMU driver, add etcd_host and etcd aliases for convenience

From now on, reads will return the server-side object version numbers
and writes and deletes will have an additional "version" parameter
which, if set to a non-zero value, will be atomically compared with
the current version of the object plus 1 and the modification will
fail if it doesn't match.

This feature opens the road to correct online flattening of snapshot
layers and other interesting things.

cinder-driver/libvirt-5.0-vitastor.diff

@@ -0,0 +1,609 @@
+commit bd283191b3e7a4c6d1c100d3d96e348a1ebffe55
+Author: Vitaliy Filippov <vitalif@yourcmc.ru>
+Date:   Sun Jun 27 12:52:40 2021 +0300
+
+    Add Vitastor support
+
+diff --git a/docs/schemas/domaincommon.rng b/docs/schemas/domaincommon.rng
+index aa50eac..082b4f8 100644
+--- a/docs/schemas/domaincommon.rng
++++ b/docs/schemas/domaincommon.rng
+@@ -1728,6 +1728,35 @@
+     </element>
+   </define>
+ 
++  <define name="diskSourceNetworkProtocolVitastor">
++    <element name="source">
++      <interleave>
++        <attribute name="protocol">
++          <value>vitastor</value>
++        </attribute>
++        <ref name="diskSourceCommon"/>
++        <optional>
++          <attribute name="name"/>
++        </optional>
++        <optional>
++          <attribute name="query"/>
++        </optional>
++        <zeroOrMore>
++          <ref name="diskSourceNetworkHost"/>
++        </zeroOrMore>
++        <optional>
++          <element name="config">
++            <attribute name="file">
++              <ref name="absFilePath"/>
++            </attribute>
++            <empty/>
++          </element>
++        </optional>
++        <empty/>
++      </interleave>
++    </element>
++  </define>
++
+   <define name="diskSourceNetworkProtocolISCSI">
+     <element name="source">
+       <attribute name="protocol">
+@@ -1851,6 +1880,7 @@
+       <ref name="diskSourceNetworkProtocolHTTP"/>
+       <ref name="diskSourceNetworkProtocolSimple"/>
+       <ref name="diskSourceNetworkProtocolVxHS"/>
++      <ref name="diskSourceNetworkProtocolVitastor"/>
+     </choice>
+   </define>
+ 
+diff --git a/include/libvirt/libvirt-storage.h b/include/libvirt/libvirt-storage.h
+index 4bf2b5f..dbc011b 100644
+--- a/include/libvirt/libvirt-storage.h
++++ b/include/libvirt/libvirt-storage.h
+@@ -240,6 +240,7 @@ typedef enum {
+     VIR_CONNECT_LIST_STORAGE_POOLS_GLUSTER       = 1 << 16,
+     VIR_CONNECT_LIST_STORAGE_POOLS_ZFS           = 1 << 17,
+     VIR_CONNECT_LIST_STORAGE_POOLS_VSTORAGE      = 1 << 18,
++    VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR      = 1 << 20,
+ } virConnectListAllStoragePoolsFlags;
+ 
+ int                     virConnectListAllStoragePools(virConnectPtr conn,
+diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c
+index 222bb8c..685d255 100644
+--- a/src/conf/domain_conf.c
++++ b/src/conf/domain_conf.c
+@@ -8653,6 +8653,10 @@ virDomainDiskSourceNetworkParse(xmlNodePtr node,
+         goto cleanup;
+     }
+ 
++    if (src->protocol == VIR_STORAGE_NET_PROTOCOL_VITASTOR) {
++        src->relPath = virXMLPropString(node, "query");
++    }
++
+     if ((haveTLS = virXMLPropString(node, "tls")) &&
+         (src->haveTLS = virTristateBoolTypeFromString(haveTLS)) <= 0) {
+         virReportError(VIR_ERR_XML_ERROR,
+@@ -23849,6 +23853,10 @@ virDomainDiskSourceFormatNetwork(virBufferPtr attrBuf,
+ 
+     virBufferEscapeString(attrBuf, " name='%s'", path ? path : src->path);
+ 
++    if (src->protocol == VIR_STORAGE_NET_PROTOCOL_VITASTOR && src->relPath != NULL) {
++        virBufferEscapeString(attrBuf, " query='%s'", src->relPath);
++    }
++
+     VIR_FREE(path);
+ 
+     if (src->haveTLS != VIR_TRISTATE_BOOL_ABSENT &&
+@@ -30930,6 +30938,7 @@ virDomainDiskTranslateSourcePool(virDomainDiskDefPtr def)
+ 
+     case VIR_STORAGE_POOL_MPATH:
+     case VIR_STORAGE_POOL_RBD:
++    case VIR_STORAGE_POOL_VITASTOR:
+     case VIR_STORAGE_POOL_SHEEPDOG:
+     case VIR_STORAGE_POOL_GLUSTER:
+     case VIR_STORAGE_POOL_LAST:
+diff --git a/src/conf/storage_conf.c b/src/conf/storage_conf.c
+index 55db7a9..7cbe937 100644
+--- a/src/conf/storage_conf.c
++++ b/src/conf/storage_conf.c
+@@ -58,7 +58,7 @@ VIR_ENUM_IMPL(virStoragePool,
+               "logical", "disk", "iscsi",
+               "iscsi-direct", "scsi", "mpath",
+               "rbd", "sheepdog", "gluster",
+-              "zfs", "vstorage")
++              "zfs", "vstorage", "vitastor")
+ 
+ VIR_ENUM_IMPL(virStoragePoolFormatFileSystem,
+               VIR_STORAGE_POOL_FS_LAST,
+@@ -232,6 +232,18 @@ static virStoragePoolTypeInfo poolTypeInfo[] = {
+           .formatToString = virStorageFileFormatTypeToString,
+       }
+     },
++    {.poolType = VIR_STORAGE_POOL_VITASTOR,
++     .poolOptions = {
++         .flags = (VIR_STORAGE_POOL_SOURCE_HOST |
++                   VIR_STORAGE_POOL_SOURCE_NETWORK |
++                   VIR_STORAGE_POOL_SOURCE_NAME),
++      },
++      .volOptions = {
++          .defaultFormat = VIR_STORAGE_FILE_RAW,
++          .formatFromString = virStorageVolumeFormatFromString,
++          .formatToString = virStorageFileFormatTypeToString,
++      }
++    },
+     {.poolType = VIR_STORAGE_POOL_SHEEPDOG,
+      .poolOptions = {
+          .flags = (VIR_STORAGE_POOL_SOURCE_HOST |
+@@ -434,6 +446,11 @@ virStoragePoolDefParseSource(xmlXPathContextPtr ctxt,
+                        _("element 'name' is mandatory for RBD pool"));
+         goto cleanup;
+     }
++    if (pool_type == VIR_STORAGE_POOL_VITASTOR && source->name == NULL) {
++        virReportError(VIR_ERR_XML_ERROR, "%s",
++                       _("element 'name' is mandatory for Vitastor pool"));
++        return -1;
++    }
+ 
+     if (options->formatFromString) {
+         char *format = virXPathString("string(./format/@type)", ctxt);
+@@ -1009,6 +1026,7 @@ virStoragePoolDefFormatBuf(virBufferPtr buf,
+     /* RBD, Sheepdog, Gluster and Iscsi-direct devices are not local block devs nor
+      * files, so they don't have a target */
+     if (def->type != VIR_STORAGE_POOL_RBD &&
++        def->type != VIR_STORAGE_POOL_VITASTOR &&
+         def->type != VIR_STORAGE_POOL_SHEEPDOG &&
+         def->type != VIR_STORAGE_POOL_GLUSTER &&
+         def->type != VIR_STORAGE_POOL_ISCSI_DIRECT) {
+diff --git a/src/conf/storage_conf.h b/src/conf/storage_conf.h
+index dc0aa2a..ed4983d 100644
+--- a/src/conf/storage_conf.h
++++ b/src/conf/storage_conf.h
+@@ -91,6 +91,7 @@ typedef enum {
+     VIR_STORAGE_POOL_GLUSTER,  /* Gluster device */
+     VIR_STORAGE_POOL_ZFS,      /* ZFS */
+     VIR_STORAGE_POOL_VSTORAGE, /* Virtuozzo Storage */
++    VIR_STORAGE_POOL_VITASTOR, /* Vitastor */
+ 
+     VIR_STORAGE_POOL_LAST,
+ } virStoragePoolType;
+@@ -422,6 +423,7 @@ VIR_ENUM_DECL(virStoragePartedFs)
+                  VIR_CONNECT_LIST_STORAGE_POOLS_SCSI     | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_MPATH    | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_RBD      | \
++                 VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_SHEEPDOG | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_GLUSTER  | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_ZFS      | \
+diff --git a/src/conf/virstorageobj.c b/src/conf/virstorageobj.c
+index 6ea6a97..3ba45b9 100644
+--- a/src/conf/virstorageobj.c
++++ b/src/conf/virstorageobj.c
+@@ -1478,6 +1478,7 @@ virStoragePoolObjSourceFindDuplicateCb(const void *payload,
+             return 1;
+         break;
+ 
++    case VIR_STORAGE_POOL_VITASTOR:
+     case VIR_STORAGE_POOL_RBD:
+     case VIR_STORAGE_POOL_LAST:
+         break;
+@@ -1971,6 +1972,8 @@ virStoragePoolObjMatch(virStoragePoolObjPtr obj,
+                (obj->def->type == VIR_STORAGE_POOL_MPATH))   ||
+               (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_RBD) &&
+                (obj->def->type == VIR_STORAGE_POOL_RBD))     ||
++              (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR) &&
++               (obj->def->type == VIR_STORAGE_POOL_VITASTOR)) ||
+               (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_SHEEPDOG) &&
+                (obj->def->type == VIR_STORAGE_POOL_SHEEPDOG)) ||
+               (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_GLUSTER) &&
+diff --git a/src/libvirt-storage.c b/src/libvirt-storage.c
+index 2ea3e94..d5d2273 100644
+--- a/src/libvirt-storage.c
++++ b/src/libvirt-storage.c
+@@ -92,6 +92,7 @@ virStoragePoolGetConnect(virStoragePoolPtr pool)
+  * VIR_CONNECT_LIST_STORAGE_POOLS_SCSI
+  * VIR_CONNECT_LIST_STORAGE_POOLS_MPATH
+  * VIR_CONNECT_LIST_STORAGE_POOLS_RBD
++ * VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR
+  * VIR_CONNECT_LIST_STORAGE_POOLS_SHEEPDOG
+  *
+  * Returns the number of storage pools found or -1 and sets @pools to
+diff --git a/src/libxl/libxl_conf.c b/src/libxl/libxl_conf.c
+index 73e988a..ab7bb81 100644
+--- a/src/libxl/libxl_conf.c
++++ b/src/libxl/libxl_conf.c
+@@ -905,6 +905,7 @@ libxlMakeNetworkDiskSrcStr(virStorageSourcePtr src,
+     case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+     case VIR_STORAGE_NET_PROTOCOL_SSH:
+     case VIR_STORAGE_NET_PROTOCOL_VXHS:
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+     case VIR_STORAGE_NET_PROTOCOL_LAST:
+     case VIR_STORAGE_NET_PROTOCOL_NONE:
+         virReportError(VIR_ERR_NO_SUPPORT,
+diff --git a/src/qemu/qemu_block.c b/src/qemu/qemu_block.c
+index cbf0aa4..096700d 100644
+--- a/src/qemu/qemu_block.c
++++ b/src/qemu/qemu_block.c
+@@ -959,6 +959,42 @@ qemuBlockStorageSourceGetRBDProps(virStorageSourcePtr src)
+ }
+ 
+ 
++static virJSONValuePtr
++qemuBlockStorageSourceGetVitastorProps(virStorageSource *src)
++{
++    virJSONValuePtr ret = NULL;
++    virStorageNetHostDefPtr host;
++    size_t i;
++    virBuffer buf = VIR_BUFFER_INITIALIZER;
++    char *etcd = NULL;
++
++    for (i = 0; i < src->nhosts; i++) {
++        host = src->hosts + i;
++        if ((virStorageNetHostTransport)host->transport != VIR_STORAGE_NET_HOST_TRANS_TCP) {
++            goto cleanup;
++        }
++        virBufferAsprintf(&buf, i > 0 ? ",%s:%u" : "%s:%u", host->name, host->port);
++    }
++    if (src->nhosts > 0) {
++        etcd = virBufferContentAndReset(&buf);
++    }
++
++    if (virJSONValueObjectCreate(&ret,
++                                 "s:driver", "vitastor",
++                                 "S:etcd_host", etcd,
++                                 "S:etcd_prefix", src->relPath,
++                                 "S:config_path", src->configFile,
++                                 "s:image", src->path,
++                                 NULL) < 0)
++        goto cleanup;
++
++cleanup:
++    VIR_FREE(etcd);
++    virBufferFreeAndReset(&buf);
++    return ret;
++}
++
++
+ static virJSONValuePtr
+ qemuBlockStorageSourceGetSheepdogProps(virStorageSourcePtr src)
+ {
+@@ -1174,6 +1210,11 @@ qemuBlockStorageSourceGetBackendProps(virStorageSourcePtr src,
+                 return NULL;
+             break;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++            if (!(fileprops = qemuBlockStorageSourceGetVitastorProps(src)))
++                return NULL;
++            break;
++
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+             if (!(fileprops = qemuBlockStorageSourceGetSheepdogProps(src)))
+                 return NULL;
+diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c
+index 822d5f8..e375cef 100644
+--- a/src/qemu/qemu_command.c
++++ b/src/qemu/qemu_command.c
+@@ -975,6 +975,43 @@ qemuBuildNetworkDriveStr(virStorageSourcePtr src,
+             ret = virBufferContentAndReset(&buf);
+             break;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++            if (strchr(src->path, ':')) {
++                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
++                               _("':' not allowed in Vitastor source volume name '%s'"),
++                               src->path);
++                return NULL;
++            }
++
++            virBufferStrcat(&buf, "vitastor:image=", src->path, NULL);
++
++            if (src->nhosts > 0) {
++                virBufferAddLit(&buf, ":etcd_host=");
++                for (i = 0; i < src->nhosts; i++) {
++                    if (i)
++                        virBufferAddLit(&buf, ",");
++
++                    /* assume host containing : is ipv6 */
++                    if (strchr(src->hosts[i].name, ':'))
++                        virBufferEscape(&buf, '\\', ":", "[%s]",
++                                        src->hosts[i].name);
++                    else
++                        virBufferAsprintf(&buf, "%s", src->hosts[i].name);
++
++                    if (src->hosts[i].port)
++                        virBufferAsprintf(&buf, "\\:%u", src->hosts[i].port);
++                }
++            }
++
++            if (src->configFile)
++                virBufferEscape(&buf, '\\', ":", ":config_path=%s", src->configFile);
++
++            if (src->relPath)
++                virBufferEscape(&buf, '\\', ":", ":etcd_prefix=%s", src->relPath);
++
++            ret = virBufferContentAndReset(&buf);
++            break;
++
+         case VIR_STORAGE_NET_PROTOCOL_VXHS:
+             virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
+                            _("VxHS protocol does not support URI syntax"));
+diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c
+index ec6b340..f399efa 100644
+--- a/src/qemu/qemu_domain.c
++++ b/src/qemu/qemu_domain.c
+@@ -10881,6 +10881,7 @@ qemuDomainPrepareStorageSourceTLS(virStorageSourcePtr src,
+         break;
+ 
+     case VIR_STORAGE_NET_PROTOCOL_RBD:
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+     case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+     case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+     case VIR_STORAGE_NET_PROTOCOL_ISCSI:
+diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c
+index 1d96170..2d24396 100644
+--- a/src/qemu/qemu_driver.c
++++ b/src/qemu/qemu_driver.c
+@@ -14687,6 +14687,7 @@ qemuDomainSnapshotPrepareDiskExternalInactive(virDomainSnapshotDiskDefPtr snapdi
+         case VIR_STORAGE_NET_PROTOCOL_TFTP:
+         case VIR_STORAGE_NET_PROTOCOL_SSH:
+         case VIR_STORAGE_NET_PROTOCOL_VXHS:
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_LAST:
+             virReportError(VIR_ERR_INTERNAL_ERROR,
+                            _("external inactive snapshots are not supported on "
+@@ -14764,6 +14765,7 @@ qemuDomainSnapshotPrepareDiskExternalActive(virDomainSnapshotDiskDefPtr snapdisk
+         case VIR_STORAGE_NET_PROTOCOL_TFTP:
+         case VIR_STORAGE_NET_PROTOCOL_SSH:
+         case VIR_STORAGE_NET_PROTOCOL_VXHS:
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_LAST:
+             virReportError(VIR_ERR_INTERNAL_ERROR,
+                            _("external active snapshots are not supported on "
+@@ -14887,6 +14889,7 @@ qemuDomainSnapshotPrepareDiskInternal(virDomainDiskDefPtr disk,
+         case VIR_STORAGE_NET_PROTOCOL_TFTP:
+         case VIR_STORAGE_NET_PROTOCOL_SSH:
+         case VIR_STORAGE_NET_PROTOCOL_VXHS:
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_LAST:
+             virReportError(VIR_ERR_INTERNAL_ERROR,
+                            _("internal inactive snapshots are not supported on "
+diff --git a/src/qemu/qemu_parse_command.c b/src/qemu/qemu_parse_command.c
+index c4650f0..551da41 100644
+--- a/src/qemu/qemu_parse_command.c
++++ b/src/qemu/qemu_parse_command.c
+@@ -2184,6 +2184,7 @@ qemuParseCommandLine(virFileCachePtr capsCache,
+                 case VIR_STORAGE_NET_PROTOCOL_TFTP:
+                 case VIR_STORAGE_NET_PROTOCOL_SSH:
+                 case VIR_STORAGE_NET_PROTOCOL_LAST:
++                case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+                 case VIR_STORAGE_NET_PROTOCOL_NONE:
+                     /* ignored for now */
+                     break;
+diff --git a/src/storage/storage_driver.c b/src/storage/storage_driver.c
+index 4a13e90..33301c7 100644
+--- a/src/storage/storage_driver.c
++++ b/src/storage/storage_driver.c
+@@ -1568,6 +1568,7 @@ storageVolLookupByPathCallback(virStoragePoolObjPtr obj,
+         case VIR_STORAGE_POOL_RBD:
+         case VIR_STORAGE_POOL_SHEEPDOG:
+         case VIR_STORAGE_POOL_ZFS:
++        case VIR_STORAGE_POOL_VITASTOR:
+         case VIR_STORAGE_POOL_LAST:
+             ignore_value(VIR_STRDUP(stable_path, data->path));
+             break;
+diff --git a/src/util/virstoragefile.c b/src/util/virstoragefile.c
+index bd4b027..b323cd6 100644
+--- a/src/util/virstoragefile.c
++++ b/src/util/virstoragefile.c
+@@ -84,7 +84,8 @@ VIR_ENUM_IMPL(virStorageNetProtocol, VIR_STORAGE_NET_PROTOCOL_LAST,
+               "ftps",
+               "tftp",
+               "ssh",
+-              "vxhs")
++              "vxhs",
++              "vitastor")
+ 
+ VIR_ENUM_IMPL(virStorageNetHostTransport, VIR_STORAGE_NET_HOST_TRANS_LAST,
+               "tcp",
+@@ -2839,6 +2840,83 @@ virStorageSourceParseRBDColonString(const char *rbdstr,
+ }
+ 
+ 
++static int
++virStorageSourceParseVitastorColonString(const char *colonstr,
++                                         virStorageSourcePtr src)
++{
++    char *p, *e, *next;
++    char *options = NULL;
++
++    /* optionally skip the "vitastor:" prefix if provided */
++    if (STRPREFIX(colonstr, "vitastor:"))
++        colonstr += strlen("vitastor:");
++
++    if (VIR_STRDUP(options, colonstr) < 0)
++        return -1;
++
++    p = options;
++    while (*p) {
++        /* find : delimiter or end of string */
++        for (e = p; *e && *e != ':'; ++e) {
++            if (*e == '\\') {
++                e++;
++                if (*e == '\0')
++                    break;
++            }
++        }
++        if (*e == '\0') {
++            next = e;    /* last kv pair */
++        } else {
++            next = e + 1;
++            *e = '\0';
++        }
++
++        if (STRPREFIX(p, "image=")) {
++            if (VIR_STRDUP(src->path, p + strlen("image=")) < 0)
++                return -1;
++        } else if (STRPREFIX(p, "etcd_prefix=")) {
++            if (VIR_STRDUP(src->relPath, p + strlen("etcd_prefix=")) < 0)
++                return -1;
++        } else if (STRPREFIX(p, "config_file=")) {
++            if (VIR_STRDUP(src->configFile, p + strlen("config_file=")) < 0)
++                return -1;
++        } else if (STRPREFIX(p, "etcd_host=")) {
++            char *h, *sep;
++
++            h = p + strlen("etcd_host=");
++            while (h < e) {
++                for (sep = h; sep < e; ++sep) {
++                    if (*sep == '\\' && (sep[1] == ',' ||
++                                         sep[1] == ';' ||
++                                         sep[1] == ' ')) {
++                        *sep = '\0';
++                        sep += 2;
++                        break;
++                    }
++                }
++
++                if (virStorageSourceRBDAddHost(src, h) < 0)
++                    goto error;
++
++                h = sep;
++            }
++        }
++
++        p = next;
++    }
++
++    if (!src->path) {
++        goto error;
++    }
++
++    return 0;
++
++error:
++    VIR_FREE(options);
++    return -1;
++}
++
++
+ static int
+ virStorageSourceParseNBDColonString(const char *nbdstr,
+                                     virStorageSourcePtr src)
+@@ -2942,6 +3020,11 @@ virStorageSourceParseBackingColon(virStorageSourcePtr src,
+             goto cleanup;
+         break;
+ 
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++        if (virStorageSourceParseVitastorColonString(path, src) < 0)
++            return -1;
++        break;
++
+     case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+     case VIR_STORAGE_NET_PROTOCOL_LAST:
+     case VIR_STORAGE_NET_PROTOCOL_NONE:
+@@ -3441,6 +3524,56 @@ virStorageSourceParseBackingJSONRBD(virStorageSourcePtr src,
+     return ret;
+ }
+ 
++static int
++virStorageSourceParseBackingJSONVitastor(virStorageSourcePtr src,
++                                         virJSONValuePtr json,
++                                         int opaque ATTRIBUTE_UNUSED)
++{
++    const char *filename;
++    const char *image = virJSONValueObjectGetString(json, "image");
++    const char *conf = virJSONValueObjectGetString(json, "config_path");
++    const char *etcd_prefix = virJSONValueObjectGetString(json, "etcd_prefix");
++    virJSONValuePtr servers = virJSONValueObjectGetArray(json, "server");
++    size_t nservers;
++    size_t i;
++
++    src->type = VIR_STORAGE_TYPE_NETWORK;
++    src->protocol = VIR_STORAGE_NET_PROTOCOL_VITASTOR;
++
++    /* legacy syntax passed via 'filename' option */
++    if ((filename = virJSONValueObjectGetString(json, "filename")))
++        return virStorageSourceParseVitastorColonString(filename, src);
++
++    if (!image) {
++        virReportError(VIR_ERR_INVALID_ARG, "%s",
++                       _("missing image name in Vitastor backing volume "
++                         "JSON specification"));
++        return -1;
++    }
++
++    if (VIR_STRDUP(src->path, image) < 0 ||
++        VIR_STRDUP(src->configFile, conf) < 0 ||
++        VIR_STRDUP(src->relPath, etcd_prefix) < 0)
++        return -1;
++
++    if (servers) {
++        nservers = virJSONValueArraySize(servers);
++
++        if (VIR_ALLOC_N(src->hosts, nservers) < 0)
++            return -1;
++
++        src->nhosts = nservers;
++
++        for (i = 0; i < nservers; i++) {
++            if (virStorageSourceParseBackingJSONInetSocketAddress(src->hosts + i,
++                                                                  virJSONValueArrayGet(servers, i)) < 0)
++                return -1;
++        }
++    }
++
++    return 0;
++}
++
+ static int
+ virStorageSourceParseBackingJSONRaw(virStorageSourcePtr src,
+                                     virJSONValuePtr json,
+@@ -3507,6 +3640,7 @@ static const struct virStorageSourceJSONDriverParser jsonParsers[] = {
+     {"sheepdog", virStorageSourceParseBackingJSONSheepdog, 0},
+     {"ssh", virStorageSourceParseBackingJSONSSH, 0},
+     {"rbd", virStorageSourceParseBackingJSONRBD, 0},
++    {"vitastor", virStorageSourceParseBackingJSONVitastor, 0},
+     {"raw", virStorageSourceParseBackingJSONRaw, 0},
+     {"vxhs", virStorageSourceParseBackingJSONVxHS, 0},
+ };
+@@ -4276,6 +4410,7 @@ virStorageSourceNetworkDefaultPort(virStorageNetProtocol protocol)
+         case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+             return 24007;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
+             /* we don't provide a default for RBD */
+             return 0;
+diff --git a/src/util/virstoragefile.h b/src/util/virstoragefile.h
+index 1d6161a..8d83bf3 100644
+--- a/src/util/virstoragefile.h
++++ b/src/util/virstoragefile.h
+@@ -134,6 +134,7 @@ typedef enum {
+     VIR_STORAGE_NET_PROTOCOL_TFTP,
+     VIR_STORAGE_NET_PROTOCOL_SSH,
+     VIR_STORAGE_NET_PROTOCOL_VXHS,
++    VIR_STORAGE_NET_PROTOCOL_VITASTOR,
+ 
+     VIR_STORAGE_NET_PROTOCOL_LAST
+ } virStorageNetProtocol;
+diff --git a/src/xenconfig/xen_xl.c b/src/xenconfig/xen_xl.c
+index accfc3a..a18f9c3 100644
+--- a/src/xenconfig/xen_xl.c
++++ b/src/xenconfig/xen_xl.c
+@@ -1535,6 +1535,7 @@ xenFormatXLDiskSrcNet(virStorageSourcePtr src)
+     case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+     case VIR_STORAGE_NET_PROTOCOL_SSH:
+     case VIR_STORAGE_NET_PROTOCOL_VXHS:
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+     case VIR_STORAGE_NET_PROTOCOL_LAST:
+     case VIR_STORAGE_NET_PROTOCOL_NONE:
+         virReportError(VIR_ERR_NO_SUPPORT,
+diff --git a/tools/virsh-pool.c b/tools/virsh-pool.c
+index 70ca39b..9caef51 100644
+--- a/tools/virsh-pool.c
++++ b/tools/virsh-pool.c
+@@ -1219,6 +1219,9 @@ cmdPoolList(vshControl *ctl, const vshCmd *cmd ATTRIBUTE_UNUSED)
+             case VIR_STORAGE_POOL_VSTORAGE:
+                 flags |= VIR_CONNECT_LIST_STORAGE_POOLS_VSTORAGE;
+                 break;
++            case VIR_STORAGE_POOL_VITASTOR:
++                flags |= VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR;
++                break;
+             case VIR_STORAGE_POOL_LAST:
+                 break;
+             }

cinder-driver/libvirt-7.0-vitastor.diff

@@ -0,0 +1,657 @@
+commit 41cdfe8317d98f70aadedfdbb381effed2641bdd
+Author: Vitaliy Filippov <vitalif@yourcmc.ru>
+Date:   Fri Jul 9 01:31:57 2021 +0300
+
+    Add Vitastor support
+
+diff --git a/docs/schemas/domaincommon.rng b/docs/schemas/domaincommon.rng
+index 7dc419b..875433b 100644
+--- a/docs/schemas/domaincommon.rng
++++ b/docs/schemas/domaincommon.rng
+@@ -1827,6 +1827,35 @@
+     </element>
+   </define>
+ 
++  <define name="diskSourceNetworkProtocolVitastor">
++    <element name="source">
++      <interleave>
++        <attribute name="protocol">
++          <value>vitastor</value>
++        </attribute>
++        <ref name="diskSourceCommon"/>
++        <optional>
++          <attribute name="name"/>
++        </optional>
++        <optional>
++          <attribute name="query"/>
++        </optional>
++        <zeroOrMore>
++          <ref name="diskSourceNetworkHost"/>
++        </zeroOrMore>
++        <optional>
++          <element name="config">
++            <attribute name="file">
++              <ref name="absFilePath"/>
++            </attribute>
++            <empty/>
++          </element>
++        </optional>
++        <empty/>
++      </interleave>
++    </element>
++  </define>
++
+   <define name="diskSourceNetworkProtocolISCSI">
+     <element name="source">
+       <attribute name="protocol">
+@@ -2083,6 +2112,7 @@
+       <ref name="diskSourceNetworkProtocolSimple"/>
+       <ref name="diskSourceNetworkProtocolVxHS"/>
+       <ref name="diskSourceNetworkProtocolNFS"/>
++      <ref name="diskSourceNetworkProtocolVitastor"/>
+     </choice>
+   </define>
+ 
+diff --git a/include/libvirt/libvirt-storage.h b/include/libvirt/libvirt-storage.h
+index 089e1e0..d7e7ef4 100644
+--- a/include/libvirt/libvirt-storage.h
++++ b/include/libvirt/libvirt-storage.h
+@@ -245,6 +245,7 @@ typedef enum {
+     VIR_CONNECT_LIST_STORAGE_POOLS_ZFS           = 1 << 17,
+     VIR_CONNECT_LIST_STORAGE_POOLS_VSTORAGE      = 1 << 18,
+     VIR_CONNECT_LIST_STORAGE_POOLS_ISCSI_DIRECT  = 1 << 19,
++    VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR      = 1 << 20,
+ } virConnectListAllStoragePoolsFlags;
+ 
+ int                     virConnectListAllStoragePools(virConnectPtr conn,
+diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c
+index 01b7187..c6e9702 100644
+--- a/src/conf/domain_conf.c
++++ b/src/conf/domain_conf.c
+@@ -8261,7 +8261,8 @@ virDomainDiskSourceNetworkParse(xmlNodePtr node,
+     src->configFile = virXPathString("string(./config/@file)", ctxt);
+ 
+     if (src->protocol == VIR_STORAGE_NET_PROTOCOL_HTTP ||
+-        src->protocol == VIR_STORAGE_NET_PROTOCOL_HTTPS)
++        src->protocol == VIR_STORAGE_NET_PROTOCOL_HTTPS ||
++        src->protocol == VIR_STORAGE_NET_PROTOCOL_VITASTOR)
+         src->query = virXMLPropString(node, "query");
+ 
+     if (virDomainStorageNetworkParseHosts(node, ctxt, &src->hosts, &src->nhosts) < 0)
+@@ -31392,6 +31393,7 @@ virDomainStorageSourceTranslateSourcePool(virStorageSourcePtr src,
+ 
+     case VIR_STORAGE_POOL_MPATH:
+     case VIR_STORAGE_POOL_RBD:
++    case VIR_STORAGE_POOL_VITASTOR:
+     case VIR_STORAGE_POOL_SHEEPDOG:
+     case VIR_STORAGE_POOL_GLUSTER:
+     case VIR_STORAGE_POOL_LAST:
+diff --git a/src/conf/storage_conf.c b/src/conf/storage_conf.c
+index 0c50529..fe97574 100644
+--- a/src/conf/storage_conf.c
++++ b/src/conf/storage_conf.c
+@@ -60,7 +60,7 @@ VIR_ENUM_IMPL(virStoragePool,
+               "logical", "disk", "iscsi",
+               "iscsi-direct", "scsi", "mpath",
+               "rbd", "sheepdog", "gluster",
+-              "zfs", "vstorage",
++              "zfs", "vstorage", "vitastor",
+ );
+ 
+ VIR_ENUM_IMPL(virStoragePoolFormatFileSystem,
+@@ -249,6 +249,18 @@ static virStoragePoolTypeInfo poolTypeInfo[] = {
+           .formatToString = virStorageFileFormatTypeToString,
+       }
+     },
++    {.poolType = VIR_STORAGE_POOL_VITASTOR,
++     .poolOptions = {
++         .flags = (VIR_STORAGE_POOL_SOURCE_HOST |
++                   VIR_STORAGE_POOL_SOURCE_NETWORK |
++                   VIR_STORAGE_POOL_SOURCE_NAME),
++      },
++      .volOptions = {
++          .defaultFormat = VIR_STORAGE_FILE_RAW,
++          .formatFromString = virStorageVolumeFormatFromString,
++          .formatToString = virStorageFileFormatTypeToString,
++      }
++    },
+     {.poolType = VIR_STORAGE_POOL_SHEEPDOG,
+      .poolOptions = {
+          .flags = (VIR_STORAGE_POOL_SOURCE_HOST |
+@@ -551,6 +563,11 @@ virStoragePoolDefParseSource(xmlXPathContextPtr ctxt,
+                        _("element 'name' is mandatory for RBD pool"));
+         goto cleanup;
+     }
++    if (pool_type == VIR_STORAGE_POOL_VITASTOR && source->name == NULL) {
++        virReportError(VIR_ERR_XML_ERROR, "%s",
++                       _("element 'name' is mandatory for Vitastor pool"));
++        return -1;
++    }
+ 
+     if (options->formatFromString) {
+         g_autofree char *format = NULL;
+@@ -1217,6 +1234,7 @@ virStoragePoolDefFormatBuf(virBufferPtr buf,
+     /* RBD, Sheepdog, Gluster and Iscsi-direct devices are not local block devs nor
+      * files, so they don't have a target */
+     if (def->type != VIR_STORAGE_POOL_RBD &&
++        def->type != VIR_STORAGE_POOL_VITASTOR &&
+         def->type != VIR_STORAGE_POOL_SHEEPDOG &&
+         def->type != VIR_STORAGE_POOL_GLUSTER &&
+         def->type != VIR_STORAGE_POOL_ISCSI_DIRECT) {
+diff --git a/src/conf/storage_conf.h b/src/conf/storage_conf.h
+index ffd406e..8868a05 100644
+--- a/src/conf/storage_conf.h
++++ b/src/conf/storage_conf.h
+@@ -110,6 +110,7 @@ typedef enum {
+     VIR_STORAGE_POOL_GLUSTER,  /* Gluster device */
+     VIR_STORAGE_POOL_ZFS,      /* ZFS */
+     VIR_STORAGE_POOL_VSTORAGE, /* Virtuozzo Storage */
++    VIR_STORAGE_POOL_VITASTOR, /* Vitastor */
+ 
+     VIR_STORAGE_POOL_LAST,
+ } virStoragePoolType;
+@@ -474,6 +475,7 @@ VIR_ENUM_DECL(virStoragePartedFs);
+                  VIR_CONNECT_LIST_STORAGE_POOLS_SCSI     | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_MPATH    | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_RBD      | \
++                 VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_SHEEPDOG | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_GLUSTER  | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_ZFS      | \
+diff --git a/src/conf/virstorageobj.c b/src/conf/virstorageobj.c
+index 9fe8b3f..bf595b0 100644
+--- a/src/conf/virstorageobj.c
++++ b/src/conf/virstorageobj.c
+@@ -1491,6 +1491,7 @@ virStoragePoolObjSourceFindDuplicateCb(const void *payload,
+             return 1;
+         break;
+ 
++    case VIR_STORAGE_POOL_VITASTOR:
+     case VIR_STORAGE_POOL_RBD:
+     case VIR_STORAGE_POOL_LAST:
+         break;
+@@ -1990,6 +1991,8 @@ virStoragePoolObjMatch(virStoragePoolObjPtr obj,
+                (obj->def->type == VIR_STORAGE_POOL_MPATH))   ||
+               (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_RBD) &&
+                (obj->def->type == VIR_STORAGE_POOL_RBD))     ||
++              (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR) &&
++               (obj->def->type == VIR_STORAGE_POOL_VITASTOR)) ||
+               (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_SHEEPDOG) &&
+                (obj->def->type == VIR_STORAGE_POOL_SHEEPDOG)) ||
+               (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_GLUSTER) &&
+diff --git a/src/libvirt-storage.c b/src/libvirt-storage.c
+index 2a7cdca..f756be1 100644
+--- a/src/libvirt-storage.c
++++ b/src/libvirt-storage.c
+@@ -92,6 +92,7 @@ virStoragePoolGetConnect(virStoragePoolPtr pool)
+  * VIR_CONNECT_LIST_STORAGE_POOLS_SCSI
+  * VIR_CONNECT_LIST_STORAGE_POOLS_MPATH
+  * VIR_CONNECT_LIST_STORAGE_POOLS_RBD
++ * VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR
+  * VIR_CONNECT_LIST_STORAGE_POOLS_SHEEPDOG
+  * VIR_CONNECT_LIST_STORAGE_POOLS_GLUSTER
+  * VIR_CONNECT_LIST_STORAGE_POOLS_ZFS
+diff --git a/src/libxl/libxl_conf.c b/src/libxl/libxl_conf.c
+index 6a8ae27..a735bc6 100644
+--- a/src/libxl/libxl_conf.c
++++ b/src/libxl/libxl_conf.c
+@@ -942,6 +942,7 @@ libxlMakeNetworkDiskSrcStr(virStorageSourcePtr src,
+     case VIR_STORAGE_NET_PROTOCOL_SSH:
+     case VIR_STORAGE_NET_PROTOCOL_VXHS:
+     case VIR_STORAGE_NET_PROTOCOL_NFS:
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+     case VIR_STORAGE_NET_PROTOCOL_LAST:
+     case VIR_STORAGE_NET_PROTOCOL_NONE:
+         virReportError(VIR_ERR_NO_SUPPORT,
+diff --git a/src/libxl/xen_xl.c b/src/libxl/xen_xl.c
+index 17b93d0..c5a0084 100644
+--- a/src/libxl/xen_xl.c
++++ b/src/libxl/xen_xl.c
+@@ -1601,6 +1601,7 @@ xenFormatXLDiskSrcNet(virStorageSourcePtr src)
+     case VIR_STORAGE_NET_PROTOCOL_SSH:
+     case VIR_STORAGE_NET_PROTOCOL_VXHS:
+     case VIR_STORAGE_NET_PROTOCOL_NFS:
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+     case VIR_STORAGE_NET_PROTOCOL_LAST:
+     case VIR_STORAGE_NET_PROTOCOL_NONE:
+         virReportError(VIR_ERR_NO_SUPPORT,
+diff --git a/src/qemu/qemu_block.c b/src/qemu/qemu_block.c
+index f9c6da2..922dde5 100644
+--- a/src/qemu/qemu_block.c
++++ b/src/qemu/qemu_block.c
+@@ -938,6 +938,38 @@ qemuBlockStorageSourceGetRBDProps(virStorageSourcePtr src,
+ }
+ 
+ 
++static virJSONValuePtr
++qemuBlockStorageSourceGetVitastorProps(virStorageSource *src)
++{
++    virJSONValuePtr ret = NULL;
++    virStorageNetHostDefPtr host;
++    size_t i;
++    g_auto(virBuffer) buf = VIR_BUFFER_INITIALIZER;
++    g_autofree char *etcd = NULL;
++
++    for (i = 0; i < src->nhosts; i++) {
++        host = src->hosts + i;
++        if ((virStorageNetHostTransport)host->transport != VIR_STORAGE_NET_HOST_TRANS_TCP) {
++            return NULL;
++        }
++        virBufferAsprintf(&buf, i > 0 ? ",%s:%u" : "%s:%u", host->name, host->port);
++    }
++    if (src->nhosts > 0) {
++        etcd = virBufferContentAndReset(&buf);
++    }
++
++    if (virJSONValueObjectCreate(&ret,
++                                 "S:etcd_host", etcd,
++                                 "S:etcd_prefix", src->query,
++                                 "S:config_path", src->configFile,
++                                 "s:image", src->path,
++                                 NULL) < 0)
++        return NULL;
++
++    return ret;
++}
++
++
+ static virJSONValuePtr
+ qemuBlockStorageSourceGetSheepdogProps(virStorageSourcePtr src)
+ {
+@@ -1224,6 +1256,12 @@ qemuBlockStorageSourceGetBackendProps(virStorageSourcePtr src,
+                 return NULL;
+             break;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++            driver = "vitastor";
++            if (!(fileprops = qemuBlockStorageSourceGetVitastorProps(src)))
++                return NULL;
++            break;
++
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+             driver = "sheepdog";
+             if (!(fileprops = qemuBlockStorageSourceGetSheepdogProps(src)))
+@@ -2183,6 +2221,7 @@ qemuBlockGetBackingStoreString(virStorageSourcePtr src,
+ 
+             case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+             case VIR_STORAGE_NET_PROTOCOL_RBD:
++            case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+             case VIR_STORAGE_NET_PROTOCOL_VXHS:
+             case VIR_STORAGE_NET_PROTOCOL_NFS:
+             case VIR_STORAGE_NET_PROTOCOL_SSH:
+@@ -2560,6 +2599,12 @@ qemuBlockStorageSourceCreateGetStorageProps(virStorageSourcePtr src,
+                 return -1;
+             break;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++            driver = "vitastor";
++            if (!(location = qemuBlockStorageSourceGetVitastorProps(src)))
++                return -1;
++            break;
++
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+             driver = "sheepdog";
+             if (!(location = qemuBlockStorageSourceGetSheepdogProps(src)))
+diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c
+index 6f970a3..10b39ca 100644
+--- a/src/qemu/qemu_command.c
++++ b/src/qemu/qemu_command.c
+@@ -1034,6 +1034,43 @@ qemuBuildNetworkDriveStr(virStorageSourcePtr src,
+             ret = virBufferContentAndReset(&buf);
+             break;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++            if (strchr(src->path, ':')) {
++                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
++                               _("':' not allowed in Vitastor source volume name '%s'"),
++                               src->path);
++                return NULL;
++            }
++
++            virBufferStrcat(&buf, "vitastor:image=", src->path, NULL);
++
++            if (src->nhosts > 0) {
++                virBufferAddLit(&buf, ":etcd_host=");
++                for (i = 0; i < src->nhosts; i++) {
++                    if (i)
++                        virBufferAddLit(&buf, ",");
++
++                    /* assume host containing : is ipv6 */
++                    if (strchr(src->hosts[i].name, ':'))
++                        virBufferEscape(&buf, '\\', ":", "[%s]",
++                                        src->hosts[i].name);
++                    else
++                        virBufferAsprintf(&buf, "%s", src->hosts[i].name);
++
++                    if (src->hosts[i].port)
++                        virBufferAsprintf(&buf, "\\:%u", src->hosts[i].port);
++                }
++            }
++
++            if (src->configFile)
++                virBufferEscape(&buf, '\\', ":", ":config_path=%s", src->configFile);
++
++            if (src->query)
++                virBufferEscape(&buf, '\\', ":", ":etcd_prefix=%s", src->query);
++
++            ret = virBufferContentAndReset(&buf);
++            break;
++
+         case VIR_STORAGE_NET_PROTOCOL_VXHS:
+             virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
+                            _("VxHS protocol does not support URI syntax"));
+diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c
+index 0765dc7..4cff344 100644
+--- a/src/qemu/qemu_domain.c
++++ b/src/qemu/qemu_domain.c
+@@ -4610,7 +4610,8 @@ qemuDomainValidateStorageSource(virStorageSourcePtr src,
+     if (src->query &&
+         (actualType != VIR_STORAGE_TYPE_NETWORK ||
+          (src->protocol != VIR_STORAGE_NET_PROTOCOL_HTTPS &&
+-          src->protocol != VIR_STORAGE_NET_PROTOCOL_HTTP))) {
++          src->protocol != VIR_STORAGE_NET_PROTOCOL_HTTP &&
++          src->protocol != VIR_STORAGE_NET_PROTOCOL_VITASTOR))) {
+         virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
+                        _("query is supported only with HTTP(S) protocols"));
+         return -1;
+@@ -9704,6 +9705,7 @@ qemuDomainPrepareStorageSourceTLS(virStorageSourcePtr src,
+         break;
+ 
+     case VIR_STORAGE_NET_PROTOCOL_RBD:
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+     case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+     case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+     case VIR_STORAGE_NET_PROTOCOL_ISCSI:
+diff --git a/src/qemu/qemu_snapshot.c b/src/qemu/qemu_snapshot.c
+index ee333c3..674aa58 100644
+--- a/src/qemu/qemu_snapshot.c
++++ b/src/qemu/qemu_snapshot.c
+@@ -403,6 +403,7 @@ qemuSnapshotPrepareDiskExternalInactive(virDomainSnapshotDiskDefPtr snapdisk,
+         case VIR_STORAGE_NET_PROTOCOL_NONE:
+         case VIR_STORAGE_NET_PROTOCOL_NBD:
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+         case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+         case VIR_STORAGE_NET_PROTOCOL_ISCSI:
+@@ -493,6 +494,7 @@ qemuSnapshotPrepareDiskExternalActive(virDomainObjPtr vm,
+         case VIR_STORAGE_NET_PROTOCOL_NONE:
+         case VIR_STORAGE_NET_PROTOCOL_NBD:
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+         case VIR_STORAGE_NET_PROTOCOL_ISCSI:
+         case VIR_STORAGE_NET_PROTOCOL_HTTP:
+@@ -623,6 +625,7 @@ qemuSnapshotPrepareDiskInternal(virDomainDiskDefPtr disk,
+         case VIR_STORAGE_NET_PROTOCOL_NONE:
+         case VIR_STORAGE_NET_PROTOCOL_NBD:
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+         case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+         case VIR_STORAGE_NET_PROTOCOL_ISCSI:
+diff --git a/src/storage/storage_driver.c b/src/storage/storage_driver.c
+index 16bc53a..1e5d820 100644
+--- a/src/storage/storage_driver.c
++++ b/src/storage/storage_driver.c
+@@ -1645,6 +1645,7 @@ storageVolLookupByPathCallback(virStoragePoolObjPtr obj,
+ 
+         case VIR_STORAGE_POOL_GLUSTER:
+         case VIR_STORAGE_POOL_RBD:
++        case VIR_STORAGE_POOL_VITASTOR:
+         case VIR_STORAGE_POOL_SHEEPDOG:
+         case VIR_STORAGE_POOL_ZFS:
+         case VIR_STORAGE_POOL_LAST:
+diff --git a/src/test/test_driver.c b/src/test/test_driver.c
+index 29c4c86..a27ad94 100644
+--- a/src/test/test_driver.c
++++ b/src/test/test_driver.c
+@@ -7096,6 +7096,7 @@ testStorageVolumeTypeForPool(int pooltype)
+     case VIR_STORAGE_POOL_ISCSI_DIRECT:
+     case VIR_STORAGE_POOL_GLUSTER:
+     case VIR_STORAGE_POOL_RBD:
++    case VIR_STORAGE_POOL_VITASTOR:
+         return VIR_STORAGE_VOL_NETWORK;
+     case VIR_STORAGE_POOL_LOGICAL:
+     case VIR_STORAGE_POOL_DISK:
+diff --git a/src/util/virstoragefile.c b/src/util/virstoragefile.c
+index 0d3c2af..36e3afc 100644
+--- a/src/util/virstoragefile.c
++++ b/src/util/virstoragefile.c
+@@ -91,6 +91,7 @@ VIR_ENUM_IMPL(virStorageNetProtocol,
+               "ssh",
+               "vxhs",
+               "nfs",
++              "vitastor",
+ );
+ 
+ VIR_ENUM_IMPL(virStorageNetHostTransport,
+@@ -2880,6 +2881,75 @@ virStorageSourceParseRBDColonString(const char *rbdstr,
+ }
+ 
+ 
++static int
++virStorageSourceParseVitastorColonString(const char *colonstr,
++                                         virStorageSourcePtr src)
++{
++    char *p, *e, *next;
++    g_autofree char *options = NULL;
++
++    /* optionally skip the "vitastor:" prefix if provided */
++    if (STRPREFIX(colonstr, "vitastor:"))
++        colonstr += strlen("vitastor:");
++
++    options = g_strdup(colonstr);
++
++    p = options;
++    while (*p) {
++        /* find : delimiter or end of string */
++        for (e = p; *e && *e != ':'; ++e) {
++            if (*e == '\\') {
++                e++;
++                if (*e == '\0')
++                    break;
++            }
++        }
++        if (*e == '\0') {
++            next = e;    /* last kv pair */
++        } else {
++            next = e + 1;
++            *e = '\0';
++        }
++
++        if (STRPREFIX(p, "image=")) {
++            src->path = g_strdup(p + strlen("image="));
++        } else if (STRPREFIX(p, "etcd_prefix=")) {
++            src->query = g_strdup(p + strlen("etcd_prefix="));
++        } else if (STRPREFIX(p, "config_file=")) {
++            src->configFile = g_strdup(p + strlen("config_file="));
++        } else if (STRPREFIX(p, "etcd_host=")) {
++            char *h, *sep;
++
++            h = p + strlen("etcd_host=");
++            while (h < e) {
++                for (sep = h; sep < e; ++sep) {
++                    if (*sep == '\\' && (sep[1] == ',' ||
++                                         sep[1] == ';' ||
++                                         sep[1] == ' ')) {
++                        *sep = '\0';
++                        sep += 2;
++                        break;
++                    }
++                }
++
++                if (virStorageSourceRBDAddHost(src, h) < 0)
++                    return -1;
++
++                h = sep;
++            }
++        }
++
++        p = next;
++    }
++
++    if (!src->path) {
++        return -1;
++    }
++
++    return 0;
++}
++
++
+ static int
+ virStorageSourceParseNBDColonString(const char *nbdstr,
+                                     virStorageSourcePtr src)
+@@ -2992,6 +3062,11 @@ virStorageSourceParseBackingColon(virStorageSourcePtr src,
+             return -1;
+         break;
+ 
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++        if (virStorageSourceParseVitastorColonString(path, src) < 0)
++            return -1;
++        break;
++
+     case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+     case VIR_STORAGE_NET_PROTOCOL_LAST:
+     case VIR_STORAGE_NET_PROTOCOL_NONE:
+@@ -3581,6 +3656,54 @@ virStorageSourceParseBackingJSONRBD(virStorageSourcePtr src,
+     return 0;
+ }
+ 
++static int
++virStorageSourceParseBackingJSONVitastor(virStorageSourcePtr src,
++                                         virJSONValuePtr json,
++                                         const char *jsonstr G_GNUC_UNUSED,
++                                         int opaque G_GNUC_UNUSED)
++{
++    const char *filename;
++    const char *image = virJSONValueObjectGetString(json, "image");
++    const char *conf = virJSONValueObjectGetString(json, "config_path");
++    const char *etcd_prefix = virJSONValueObjectGetString(json, "etcd_prefix");
++    virJSONValuePtr servers = virJSONValueObjectGetArray(json, "server");
++    size_t nservers;
++    size_t i;
++
++    src->type = VIR_STORAGE_TYPE_NETWORK;
++    src->protocol = VIR_STORAGE_NET_PROTOCOL_VITASTOR;
++
++    /* legacy syntax passed via 'filename' option */
++    if ((filename = virJSONValueObjectGetString(json, "filename")))
++        return virStorageSourceParseVitastorColonString(filename, src);
++
++    if (!image) {
++        virReportError(VIR_ERR_INVALID_ARG, "%s",
++                       _("missing image name in Vitastor backing volume "
++                         "JSON specification"));
++        return -1;
++    }
++
++    src->path = g_strdup(image);
++    src->configFile = g_strdup(conf);
++    src->query = g_strdup(etcd_prefix);
++
++    if (servers) {
++        nservers = virJSONValueArraySize(servers);
++
++        src->hosts = g_new0(virStorageNetHostDef, nservers);
++        src->nhosts = nservers;
++
++        for (i = 0; i < nservers; i++) {
++            if (virStorageSourceParseBackingJSONInetSocketAddress(src->hosts + i,
++                                                                  virJSONValueArrayGet(servers, i)) < 0)
++                return -1;
++        }
++    }
++
++    return 0;
++}
++
+ static int
+ virStorageSourceParseBackingJSONRaw(virStorageSourcePtr src,
+                                     virJSONValuePtr json,
+@@ -3759,6 +3882,7 @@ static const struct virStorageSourceJSONDriverParser jsonParsers[] = {
+     {"sheepdog", false, virStorageSourceParseBackingJSONSheepdog, 0},
+     {"ssh", false, virStorageSourceParseBackingJSONSSH, 0},
+     {"rbd", false, virStorageSourceParseBackingJSONRBD, 0},
++    {"vitastor", false, virStorageSourceParseBackingJSONVitastor, 0},
+     {"raw", true, virStorageSourceParseBackingJSONRaw, 0},
+     {"nfs", false, virStorageSourceParseBackingJSONNFS, 0},
+     {"vxhs", false, virStorageSourceParseBackingJSONVxHS, 0},
+@@ -4503,6 +4627,7 @@ virStorageSourceNetworkDefaultPort(virStorageNetProtocol protocol)
+         case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+             return 24007;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
+             /* we don't provide a default for RBD */
+             return 0;
+diff --git a/src/util/virstoragefile.h b/src/util/virstoragefile.h
+index 5689c39..3eb4e3c 100644
+--- a/src/util/virstoragefile.h
++++ b/src/util/virstoragefile.h
+@@ -136,6 +136,7 @@ typedef enum {
+     VIR_STORAGE_NET_PROTOCOL_SSH,
+     VIR_STORAGE_NET_PROTOCOL_VXHS,
+     VIR_STORAGE_NET_PROTOCOL_NFS,
++    VIR_STORAGE_NET_PROTOCOL_VITASTOR,
+ 
+     VIR_STORAGE_NET_PROTOCOL_LAST
+ } virStorageNetProtocol;
+diff --git a/tests/storagepoolcapsschemadata/poolcaps-fs.xml b/tests/storagepoolcapsschemadata/poolcaps-fs.xml
+index eee75af..8bd0a57 100644
+--- a/tests/storagepoolcapsschemadata/poolcaps-fs.xml
++++ b/tests/storagepoolcapsschemadata/poolcaps-fs.xml
+@@ -204,4 +204,11 @@
+       </enum>
+     </volOptions>
+   </pool>
++  <pool type='vitastor' supported='no'>
++    <volOptions>
++      <defaultFormat type='raw'/>
++      <enum name='targetFormatType'>
++      </enum>
++    </volOptions>
++  </pool>
+ </storagepoolCapabilities>
+diff --git a/tests/storagepoolcapsschemadata/poolcaps-full.xml b/tests/storagepoolcapsschemadata/poolcaps-full.xml
+index 805950a..852df0d 100644
+--- a/tests/storagepoolcapsschemadata/poolcaps-full.xml
++++ b/tests/storagepoolcapsschemadata/poolcaps-full.xml
+@@ -204,4 +204,11 @@
+       </enum>
+     </volOptions>
+   </pool>
++  <pool type='vitastor' supported='yes'>
++    <volOptions>
++      <defaultFormat type='raw'/>
++      <enum name='targetFormatType'>
++      </enum>
++    </volOptions>
++  </pool>
+ </storagepoolCapabilities>
+diff --git a/tests/storagepoolxml2argvtest.c b/tests/storagepoolxml2argvtest.c
+index 967d1f2..1e8ff7a 100644
+--- a/tests/storagepoolxml2argvtest.c
++++ b/tests/storagepoolxml2argvtest.c
+@@ -68,6 +68,7 @@ testCompareXMLToArgvFiles(bool shouldFail,
+     case VIR_STORAGE_POOL_GLUSTER:
+     case VIR_STORAGE_POOL_ZFS:
+     case VIR_STORAGE_POOL_VSTORAGE:
++    case VIR_STORAGE_POOL_VITASTOR:
+     case VIR_STORAGE_POOL_LAST:
+     default:
+         VIR_TEST_DEBUG("pool type '%s' has no xml2argv test", defTypeStr);
+diff --git a/tools/virsh-pool.c b/tools/virsh-pool.c
+index 7835fa6..8841fcf 100644
+--- a/tools/virsh-pool.c
++++ b/tools/virsh-pool.c
+@@ -1237,6 +1237,9 @@ cmdPoolList(vshControl *ctl, const vshCmd *cmd G_GNUC_UNUSED)
+             case VIR_STORAGE_POOL_VSTORAGE:
+                 flags |= VIR_CONNECT_LIST_STORAGE_POOLS_VSTORAGE;
+                 break;
++            case VIR_STORAGE_POOL_VITASTOR:
++                flags |= VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR;
++                break;
+             case VIR_STORAGE_POOL_LAST:
+                 break;
+             }

cinder-driver/libvirt-7.5-vitastor.diff

@@ -0,0 +1,661 @@
+commit c6e1958a1b4974828e8e5852beb252ce6594e670
+Author: Vitaliy Filippov <vitalif@yourcmc.ru>
+Date:   Mon Jun 28 01:20:19 2021 +0300
+
+    Add Vitastor support
+
+diff --git a/docs/schemas/domaincommon.rng b/docs/schemas/domaincommon.rng
+index 5ea14b6..a9df168 100644
+--- a/docs/schemas/domaincommon.rng
++++ b/docs/schemas/domaincommon.rng
+@@ -1859,6 +1859,35 @@
+     </element>
+   </define>
+ 
++  <define name="diskSourceNetworkProtocolVitastor">
++    <element name="source">
++      <interleave>
++        <attribute name="protocol">
++          <value>vitastor</value>
++        </attribute>
++        <ref name="diskSourceCommon"/>
++        <optional>
++          <attribute name="name"/>
++        </optional>
++        <optional>
++          <attribute name="query"/>
++        </optional>
++        <zeroOrMore>
++          <ref name="diskSourceNetworkHost"/>
++        </zeroOrMore>
++        <optional>
++          <element name="config">
++            <attribute name="file">
++              <ref name="absFilePath"/>
++            </attribute>
++            <empty/>
++          </element>
++        </optional>
++        <empty/>
++      </interleave>
++    </element>
++  </define>
++
+   <define name="diskSourceNetworkProtocolISCSI">
+     <element name="source">
+       <attribute name="protocol">
+@@ -2115,6 +2144,7 @@
+       <ref name="diskSourceNetworkProtocolSimple"/>
+       <ref name="diskSourceNetworkProtocolVxHS"/>
+       <ref name="diskSourceNetworkProtocolNFS"/>
++      <ref name="diskSourceNetworkProtocolVitastor"/>
+     </choice>
+   </define>
+ 
+diff --git a/include/libvirt/libvirt-storage.h b/include/libvirt/libvirt-storage.h
+index 089e1e0..d7e7ef4 100644
+--- a/include/libvirt/libvirt-storage.h
++++ b/include/libvirt/libvirt-storage.h
+@@ -245,6 +245,7 @@ typedef enum {
+     VIR_CONNECT_LIST_STORAGE_POOLS_ZFS           = 1 << 17,
+     VIR_CONNECT_LIST_STORAGE_POOLS_VSTORAGE      = 1 << 18,
+     VIR_CONNECT_LIST_STORAGE_POOLS_ISCSI_DIRECT  = 1 << 19,
++    VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR      = 1 << 20,
+ } virConnectListAllStoragePoolsFlags;
+ 
+ int                     virConnectListAllStoragePools(virConnectPtr conn,
+diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c
+index d78f846..f7222e3 100644
+--- a/src/conf/domain_conf.c
++++ b/src/conf/domain_conf.c
+@@ -8251,7 +8251,8 @@ virDomainDiskSourceNetworkParse(xmlNodePtr node,
+     src->configFile = virXPathString("string(./config/@file)", ctxt);
+ 
+     if (src->protocol == VIR_STORAGE_NET_PROTOCOL_HTTP ||
+-        src->protocol == VIR_STORAGE_NET_PROTOCOL_HTTPS)
++        src->protocol == VIR_STORAGE_NET_PROTOCOL_HTTPS ||
++        src->protocol == VIR_STORAGE_NET_PROTOCOL_VITASTOR)
+         src->query = virXMLPropString(node, "query");
+ 
+     if (virDomainStorageNetworkParseHosts(node, ctxt, &src->hosts, &src->nhosts) < 0)
+@@ -30775,6 +30776,7 @@ virDomainStorageSourceTranslateSourcePool(virStorageSource *src,
+ 
+     case VIR_STORAGE_POOL_MPATH:
+     case VIR_STORAGE_POOL_RBD:
++    case VIR_STORAGE_POOL_VITASTOR:
+     case VIR_STORAGE_POOL_SHEEPDOG:
+     case VIR_STORAGE_POOL_GLUSTER:
+     case VIR_STORAGE_POOL_LAST:
+diff --git a/src/conf/storage_conf.c b/src/conf/storage_conf.c
+index 2aa9a3d..166ca1f 100644
+--- a/src/conf/storage_conf.c
++++ b/src/conf/storage_conf.c
+@@ -60,7 +60,7 @@ VIR_ENUM_IMPL(virStoragePool,
+               "logical", "disk", "iscsi",
+               "iscsi-direct", "scsi", "mpath",
+               "rbd", "sheepdog", "gluster",
+-              "zfs", "vstorage",
++              "zfs", "vstorage", "vitastor",
+ );
+ 
+ VIR_ENUM_IMPL(virStoragePoolFormatFileSystem,
+@@ -246,6 +246,18 @@ static virStoragePoolTypeInfo poolTypeInfo[] = {
+           .formatToString = virStorageFileFormatTypeToString,
+       }
+     },
++    {.poolType = VIR_STORAGE_POOL_VITASTOR,
++     .poolOptions = {
++         .flags = (VIR_STORAGE_POOL_SOURCE_HOST |
++                   VIR_STORAGE_POOL_SOURCE_NETWORK |
++                   VIR_STORAGE_POOL_SOURCE_NAME),
++      },
++      .volOptions = {
++          .defaultFormat = VIR_STORAGE_FILE_RAW,
++          .formatFromString = virStorageVolumeFormatFromString,
++          .formatToString = virStorageFileFormatTypeToString,
++      }
++    },
+     {.poolType = VIR_STORAGE_POOL_SHEEPDOG,
+      .poolOptions = {
+          .flags = (VIR_STORAGE_POOL_SOURCE_HOST |
+@@ -546,6 +558,11 @@ virStoragePoolDefParseSource(xmlXPathContextPtr ctxt,
+                        _("element 'name' is mandatory for RBD pool"));
+         return -1;
+     }
++    if (pool_type == VIR_STORAGE_POOL_VITASTOR && source->name == NULL) {
++        virReportError(VIR_ERR_XML_ERROR, "%s",
++                       _("element 'name' is mandatory for Vitastor pool"));
++        return -1;
++    }
+ 
+     if (options->formatFromString) {
+         g_autofree char *format = NULL;
+@@ -1182,6 +1199,7 @@ virStoragePoolDefFormatBuf(virBuffer *buf,
+     /* RBD, Sheepdog, Gluster and Iscsi-direct devices are not local block devs nor
+      * files, so they don't have a target */
+     if (def->type != VIR_STORAGE_POOL_RBD &&
++        def->type != VIR_STORAGE_POOL_VITASTOR &&
+         def->type != VIR_STORAGE_POOL_SHEEPDOG &&
+         def->type != VIR_STORAGE_POOL_GLUSTER &&
+         def->type != VIR_STORAGE_POOL_ISCSI_DIRECT) {
+diff --git a/src/conf/storage_conf.h b/src/conf/storage_conf.h
+index 76efaac..928149a 100644
+--- a/src/conf/storage_conf.h
++++ b/src/conf/storage_conf.h
+@@ -106,6 +106,7 @@ typedef enum {
+     VIR_STORAGE_POOL_GLUSTER,  /* Gluster device */
+     VIR_STORAGE_POOL_ZFS,      /* ZFS */
+     VIR_STORAGE_POOL_VSTORAGE, /* Virtuozzo Storage */
++    VIR_STORAGE_POOL_VITASTOR, /* Vitastor */
+ 
+     VIR_STORAGE_POOL_LAST,
+ } virStoragePoolType;
+@@ -465,6 +466,7 @@ VIR_ENUM_DECL(virStoragePartedFs);
+                  VIR_CONNECT_LIST_STORAGE_POOLS_SCSI     | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_MPATH    | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_RBD      | \
++                 VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_SHEEPDOG | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_GLUSTER  | \
+                  VIR_CONNECT_LIST_STORAGE_POOLS_ZFS      | \
+diff --git a/src/conf/storage_source_conf.c b/src/conf/storage_source_conf.c
+index 5ca06fa..05ded49 100644
+--- a/src/conf/storage_source_conf.c
++++ b/src/conf/storage_source_conf.c
+@@ -85,6 +85,7 @@ VIR_ENUM_IMPL(virStorageNetProtocol,
+               "ssh",
+               "vxhs",
+               "nfs",
++              "vitastor",
+ );
+ 
+ 
+@@ -1262,6 +1263,7 @@ virStorageSourceNetworkDefaultPort(virStorageNetProtocol protocol)
+         case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+             return 24007;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
+             /* we don't provide a default for RBD */
+             return 0;
+diff --git a/src/conf/storage_source_conf.h b/src/conf/storage_source_conf.h
+index 389c7b5..dbf02e3 100644
+--- a/src/conf/storage_source_conf.h
++++ b/src/conf/storage_source_conf.h
+@@ -127,6 +127,7 @@ typedef enum {
+     VIR_STORAGE_NET_PROTOCOL_SSH,
+     VIR_STORAGE_NET_PROTOCOL_VXHS,
+     VIR_STORAGE_NET_PROTOCOL_NFS,
++    VIR_STORAGE_NET_PROTOCOL_VITASTOR,
+ 
+     VIR_STORAGE_NET_PROTOCOL_LAST
+ } virStorageNetProtocol;
+diff --git a/src/conf/virstorageobj.c b/src/conf/virstorageobj.c
+index 24957d6..4520a73 100644
+--- a/src/conf/virstorageobj.c
++++ b/src/conf/virstorageobj.c
+@@ -1487,6 +1487,7 @@ virStoragePoolObjSourceFindDuplicateCb(const void *payload,
+             return 1;
+         break;
+ 
++    case VIR_STORAGE_POOL_VITASTOR:
+     case VIR_STORAGE_POOL_RBD:
+     case VIR_STORAGE_POOL_LAST:
+         break;
+@@ -1986,6 +1987,8 @@ virStoragePoolObjMatch(virStoragePoolObj *obj,
+                (obj->def->type == VIR_STORAGE_POOL_MPATH))   ||
+               (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_RBD) &&
+                (obj->def->type == VIR_STORAGE_POOL_RBD))     ||
++              (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR) &&
++               (obj->def->type == VIR_STORAGE_POOL_VITASTOR)) ||
+               (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_SHEEPDOG) &&
+                (obj->def->type == VIR_STORAGE_POOL_SHEEPDOG)) ||
+               (MATCH(VIR_CONNECT_LIST_STORAGE_POOLS_GLUSTER) &&
+diff --git a/src/libvirt-storage.c b/src/libvirt-storage.c
+index 2a7cdca..f756be1 100644
+--- a/src/libvirt-storage.c
++++ b/src/libvirt-storage.c
+@@ -92,6 +92,7 @@ virStoragePoolGetConnect(virStoragePoolPtr pool)
+  * VIR_CONNECT_LIST_STORAGE_POOLS_SCSI
+  * VIR_CONNECT_LIST_STORAGE_POOLS_MPATH
+  * VIR_CONNECT_LIST_STORAGE_POOLS_RBD
++ * VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR
+  * VIR_CONNECT_LIST_STORAGE_POOLS_SHEEPDOG
+  * VIR_CONNECT_LIST_STORAGE_POOLS_GLUSTER
+  * VIR_CONNECT_LIST_STORAGE_POOLS_ZFS
+diff --git a/src/libxl/libxl_conf.c b/src/libxl/libxl_conf.c
+index 56cb9ab..dfb31b9 100644
+--- a/src/libxl/libxl_conf.c
++++ b/src/libxl/libxl_conf.c
+@@ -972,6 +972,7 @@ libxlMakeNetworkDiskSrcStr(virStorageSource *src,
+     case VIR_STORAGE_NET_PROTOCOL_SSH:
+     case VIR_STORAGE_NET_PROTOCOL_VXHS:
+     case VIR_STORAGE_NET_PROTOCOL_NFS:
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+     case VIR_STORAGE_NET_PROTOCOL_LAST:
+     case VIR_STORAGE_NET_PROTOCOL_NONE:
+         virReportError(VIR_ERR_NO_SUPPORT,
+diff --git a/src/libxl/xen_xl.c b/src/libxl/xen_xl.c
+index c0905b0..c172378 100644
+--- a/src/libxl/xen_xl.c
++++ b/src/libxl/xen_xl.c
+@@ -1540,6 +1540,7 @@ xenFormatXLDiskSrcNet(virStorageSource *src)
+     case VIR_STORAGE_NET_PROTOCOL_SSH:
+     case VIR_STORAGE_NET_PROTOCOL_VXHS:
+     case VIR_STORAGE_NET_PROTOCOL_NFS:
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+     case VIR_STORAGE_NET_PROTOCOL_LAST:
+     case VIR_STORAGE_NET_PROTOCOL_NONE:
+         virReportError(VIR_ERR_NO_SUPPORT,
+diff --git a/src/qemu/qemu_block.c b/src/qemu/qemu_block.c
+index 6627d04..c33f428 100644
+--- a/src/qemu/qemu_block.c
++++ b/src/qemu/qemu_block.c
+@@ -928,6 +928,38 @@ qemuBlockStorageSourceGetRBDProps(virStorageSource *src,
+ }
+ 
+ 
++static virJSONValue *
++qemuBlockStorageSourceGetVitastorProps(virStorageSource *src)
++{
++    virJSONValuePtr ret = NULL;
++    virStorageNetHostDefPtr host;
++    size_t i;
++    g_auto(virBuffer) buf = VIR_BUFFER_INITIALIZER;
++    g_autofree char *etcd = NULL;
++
++    for (i = 0; i < src->nhosts; i++) {
++        host = src->hosts + i;
++        if ((virStorageNetHostTransport)host->transport != VIR_STORAGE_NET_HOST_TRANS_TCP) {
++            return NULL;
++        }
++        virBufferAsprintf(&buf, i > 0 ? ",%s:%u" : "%s:%u", host->name, host->port);
++    }
++    if (src->nhosts > 0) {
++        etcd = virBufferContentAndReset(&buf);
++    }
++
++    if (virJSONValueObjectCreate(&ret,
++                                 "S:etcd_host", etcd,
++                                 "S:etcd_prefix", src->query,
++                                 "S:config_path", src->configFile,
++                                 "s:image", src->path,
++                                 NULL) < 0)
++        return NULL;
++
++    return ret;
++}
++
++
+ static virJSONValue *
+ qemuBlockStorageSourceGetSheepdogProps(virStorageSource *src)
+ {
+@@ -1218,6 +1250,12 @@ qemuBlockStorageSourceGetBackendProps(virStorageSource *src,
+                 return NULL;
+             break;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++            driver = "vitastor";
++            if (!(fileprops = qemuBlockStorageSourceGetVitastorProps(src)))
++                return NULL;
++            break;
++
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+             driver = "sheepdog";
+             if (!(fileprops = qemuBlockStorageSourceGetSheepdogProps(src)))
+@@ -2231,6 +2269,7 @@ qemuBlockGetBackingStoreString(virStorageSource *src,
+ 
+             case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+             case VIR_STORAGE_NET_PROTOCOL_RBD:
++            case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+             case VIR_STORAGE_NET_PROTOCOL_VXHS:
+             case VIR_STORAGE_NET_PROTOCOL_NFS:
+             case VIR_STORAGE_NET_PROTOCOL_SSH:
+@@ -2608,6 +2647,12 @@ qemuBlockStorageSourceCreateGetStorageProps(virStorageSource *src,
+                 return -1;
+             break;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++            driver = "vitastor";
++            if (!(location = qemuBlockStorageSourceGetVitastorProps(src)))
++                return -1;
++            break;
++
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+             driver = "sheepdog";
+             if (!(location = qemuBlockStorageSourceGetSheepdogProps(src)))
+diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c
+index ea51369..8258632 100644
+--- a/src/qemu/qemu_command.c
++++ b/src/qemu/qemu_command.c
+@@ -1074,6 +1074,43 @@ qemuBuildNetworkDriveStr(virStorageSource *src,
+             ret = virBufferContentAndReset(&buf);
+             break;
+ 
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++            if (strchr(src->path, ':')) {
++                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
++                               _("':' not allowed in Vitastor source volume name '%s'"),
++                               src->path);
++                return NULL;
++            }
++
++            virBufferStrcat(&buf, "vitastor:image=", src->path, NULL);
++
++            if (src->nhosts > 0) {
++                virBufferAddLit(&buf, ":etcd_host=");
++                for (i = 0; i < src->nhosts; i++) {
++                    if (i)
++                        virBufferAddLit(&buf, ",");
++
++                    /* assume host containing : is ipv6 */
++                    if (strchr(src->hosts[i].name, ':'))
++                        virBufferEscape(&buf, '\\', ":", "[%s]",
++                                        src->hosts[i].name);
++                    else
++                        virBufferAsprintf(&buf, "%s", src->hosts[i].name);
++
++                    if (src->hosts[i].port)
++                        virBufferAsprintf(&buf, "\\:%u", src->hosts[i].port);
++                }
++            }
++
++            if (src->configFile)
++                virBufferEscape(&buf, '\\', ":", ":config_path=%s", src->configFile);
++
++            if (src->query)
++                virBufferEscape(&buf, '\\', ":", ":etcd_prefix=%s", src->query);
++
++            ret = virBufferContentAndReset(&buf);
++            break;
++
+         case VIR_STORAGE_NET_PROTOCOL_VXHS:
+             virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
+                            _("VxHS protocol does not support URI syntax"));
+diff --git a/src/qemu/qemu_domain.c b/src/qemu/qemu_domain.c
+index fc60e15..5ab410d 100644
+--- a/src/qemu/qemu_domain.c
++++ b/src/qemu/qemu_domain.c
+@@ -4829,7 +4829,8 @@ qemuDomainValidateStorageSource(virStorageSource *src,
+     if (src->query &&
+         (actualType != VIR_STORAGE_TYPE_NETWORK ||
+          (src->protocol != VIR_STORAGE_NET_PROTOCOL_HTTPS &&
+-          src->protocol != VIR_STORAGE_NET_PROTOCOL_HTTP))) {
++          src->protocol != VIR_STORAGE_NET_PROTOCOL_HTTP &&
++          src->protocol != VIR_STORAGE_NET_PROTOCOL_VITASTOR))) {
+         virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
+                        _("query is supported only with HTTP(S) protocols"));
+         return -1;
+@@ -10027,6 +10028,7 @@ qemuDomainPrepareStorageSourceTLS(virStorageSource *src,
+         break;
+ 
+     case VIR_STORAGE_NET_PROTOCOL_RBD:
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+     case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+     case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+     case VIR_STORAGE_NET_PROTOCOL_ISCSI:
+diff --git a/src/qemu/qemu_snapshot.c b/src/qemu/qemu_snapshot.c
+index 4e74ddd..14e5f2e 100644
+--- a/src/qemu/qemu_snapshot.c
++++ b/src/qemu/qemu_snapshot.c
+@@ -402,6 +402,7 @@ qemuSnapshotPrepareDiskExternalInactive(virDomainSnapshotDiskDef *snapdisk,
+         case VIR_STORAGE_NET_PROTOCOL_NONE:
+         case VIR_STORAGE_NET_PROTOCOL_NBD:
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+         case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+         case VIR_STORAGE_NET_PROTOCOL_ISCSI:
+@@ -494,6 +495,7 @@ qemuSnapshotPrepareDiskExternalActive(virDomainObj *vm,
+         case VIR_STORAGE_NET_PROTOCOL_NONE:
+         case VIR_STORAGE_NET_PROTOCOL_NBD:
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+         case VIR_STORAGE_NET_PROTOCOL_ISCSI:
+         case VIR_STORAGE_NET_PROTOCOL_HTTP:
+@@ -647,6 +649,7 @@ qemuSnapshotPrepareDiskInternal(virDomainDiskDef *disk,
+         case VIR_STORAGE_NET_PROTOCOL_NONE:
+         case VIR_STORAGE_NET_PROTOCOL_NBD:
+         case VIR_STORAGE_NET_PROTOCOL_RBD:
++        case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
+         case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+         case VIR_STORAGE_NET_PROTOCOL_GLUSTER:
+         case VIR_STORAGE_NET_PROTOCOL_ISCSI:
+diff --git a/src/storage/storage_driver.c b/src/storage/storage_driver.c
+index c2ff4b8..70d0689 100644
+--- a/src/storage/storage_driver.c
++++ b/src/storage/storage_driver.c
+@@ -1644,6 +1644,7 @@ storageVolLookupByPathCallback(virStoragePoolObj *obj,
+ 
+         case VIR_STORAGE_POOL_GLUSTER:
+         case VIR_STORAGE_POOL_RBD:
++        case VIR_STORAGE_POOL_VITASTOR:
+         case VIR_STORAGE_POOL_SHEEPDOG:
+         case VIR_STORAGE_POOL_ZFS:
+         case VIR_STORAGE_POOL_LAST:
+diff --git a/src/storage_file/storage_source_backingstore.c b/src/storage_file/storage_source_backingstore.c
+index e48ae72..d7a9b72 100644
+--- a/src/storage_file/storage_source_backingstore.c
++++ b/src/storage_file/storage_source_backingstore.c
+@@ -284,6 +284,75 @@ virStorageSourceParseRBDColonString(const char *rbdstr,
+ }
+ 
+ 
++static int
++virStorageSourceParseVitastorColonString(const char *colonstr,
++                                         virStorageSource *src)
++{
++    char *p, *e, *next;
++    g_autofree char *options = NULL;
++
++    /* optionally skip the "vitastor:" prefix if provided */
++    if (STRPREFIX(colonstr, "vitastor:"))
++        colonstr += strlen("vitastor:");
++
++    options = g_strdup(colonstr);
++
++    p = options;
++    while (*p) {
++        /* find : delimiter or end of string */
++        for (e = p; *e && *e != ':'; ++e) {
++            if (*e == '\\') {
++                e++;
++                if (*e == '\0')
++                    break;
++            }
++        }
++        if (*e == '\0') {
++            next = e;    /* last kv pair */
++        } else {
++            next = e + 1;
++            *e = '\0';
++        }
++
++        if (STRPREFIX(p, "image=")) {
++            src->path = g_strdup(p + strlen("image="));
++        } else if (STRPREFIX(p, "etcd_prefix=")) {
++            src->query = g_strdup(p + strlen("etcd_prefix="));
++        } else if (STRPREFIX(p, "config_file=")) {
++            src->configFile = g_strdup(p + strlen("config_file="));
++        } else if (STRPREFIX(p, "etcd_host=")) {
++            char *h, *sep;
++
++            h = p + strlen("etcd_host=");
++            while (h < e) {
++                for (sep = h; sep < e; ++sep) {
++                    if (*sep == '\\' && (sep[1] == ',' ||
++                                         sep[1] == ';' ||
++                                         sep[1] == ' ')) {
++                        *sep = '\0';
++                        sep += 2;
++                        break;
++                    }
++                }
++
++                if (virStorageSourceRBDAddHost(src, h) < 0)
++                    return -1;
++
++                h = sep;
++            }
++        }
++
++        p = next;
++    }
++
++    if (!src->path) {
++        return -1;
++    }
++
++    return 0;
++}
++
++
+ static int
+ virStorageSourceParseNBDColonString(const char *nbdstr,
+                                     virStorageSource *src)
+@@ -396,6 +465,11 @@ virStorageSourceParseBackingColon(virStorageSource *src,
+             return -1;
+         break;
+ 
++    case VIR_STORAGE_NET_PROTOCOL_VITASTOR:
++        if (virStorageSourceParseVitastorColonString(path, src) < 0)
++            return -1;
++        break;
++
+     case VIR_STORAGE_NET_PROTOCOL_SHEEPDOG:
+     case VIR_STORAGE_NET_PROTOCOL_LAST:
+     case VIR_STORAGE_NET_PROTOCOL_NONE:
+@@ -984,6 +1058,54 @@ virStorageSourceParseBackingJSONRBD(virStorageSource *src,
+     return 0;
+ }
+ 
++static int
++virStorageSourceParseBackingJSONVitastor(virStorageSource *src,
++                                         virJSONValue *json,
++                                         const char *jsonstr G_GNUC_UNUSED,
++                                         int opaque G_GNUC_UNUSED)
++{
++    const char *filename;
++    const char *image = virJSONValueObjectGetString(json, "image");
++    const char *conf = virJSONValueObjectGetString(json, "config_path");
++    const char *etcd_prefix = virJSONValueObjectGetString(json, "etcd_prefix");
++    virJSONValue *servers = virJSONValueObjectGetArray(json, "server");
++    size_t nservers;
++    size_t i;
++
++    src->type = VIR_STORAGE_TYPE_NETWORK;
++    src->protocol = VIR_STORAGE_NET_PROTOCOL_VITASTOR;
++
++    /* legacy syntax passed via 'filename' option */
++    if ((filename = virJSONValueObjectGetString(json, "filename")))
++        return virStorageSourceParseVitastorColonString(filename, src);
++
++    if (!image) {
++        virReportError(VIR_ERR_INVALID_ARG, "%s",
++                       _("missing image name in Vitastor backing volume "
++                         "JSON specification"));
++        return -1;
++    }
++
++    src->path = g_strdup(image);
++    src->configFile = g_strdup(conf);
++    src->query = g_strdup(etcd_prefix);
++
++    if (servers) {
++        nservers = virJSONValueArraySize(servers);
++
++        src->hosts = g_new0(virStorageNetHostDef, nservers);
++        src->nhosts = nservers;
++
++        for (i = 0; i < nservers; i++) {
++            if (virStorageSourceParseBackingJSONInetSocketAddress(src->hosts + i,
++                                                                  virJSONValueArrayGet(servers, i)) < 0)
++                return -1;
++        }
++    }
++
++    return 0;
++}
++
+ static int
+ virStorageSourceParseBackingJSONRaw(virStorageSource *src,
+                                     virJSONValue *json,
+@@ -1162,6 +1284,7 @@ static const struct virStorageSourceJSONDriverParser jsonParsers[] = {
+     {"sheepdog", false, virStorageSourceParseBackingJSONSheepdog, 0},
+     {"ssh", false, virStorageSourceParseBackingJSONSSH, 0},
+     {"rbd", false, virStorageSourceParseBackingJSONRBD, 0},
++    {"vitastor", false, virStorageSourceParseBackingJSONVitastor, 0},
+     {"raw", true, virStorageSourceParseBackingJSONRaw, 0},
+     {"nfs", false, virStorageSourceParseBackingJSONNFS, 0},
+     {"vxhs", false, virStorageSourceParseBackingJSONVxHS, 0},
+diff --git a/src/test/test_driver.c b/src/test/test_driver.c
+index ef0ddab..2173dc3 100644
+--- a/src/test/test_driver.c
++++ b/src/test/test_driver.c
+@@ -7131,6 +7131,7 @@ testStorageVolumeTypeForPool(int pooltype)
+     case VIR_STORAGE_POOL_ISCSI_DIRECT:
+     case VIR_STORAGE_POOL_GLUSTER:
+     case VIR_STORAGE_POOL_RBD:
++    case VIR_STORAGE_POOL_VITASTOR:
+         return VIR_STORAGE_VOL_NETWORK;
+     case VIR_STORAGE_POOL_LOGICAL:
+     case VIR_STORAGE_POOL_DISK:
+diff --git a/tests/storagepoolcapsschemadata/poolcaps-fs.xml b/tests/storagepoolcapsschemadata/poolcaps-fs.xml
+index eee75af..8bd0a57 100644
+--- a/tests/storagepoolcapsschemadata/poolcaps-fs.xml
++++ b/tests/storagepoolcapsschemadata/poolcaps-fs.xml
+@@ -204,4 +204,11 @@
+       </enum>
+     </volOptions>
+   </pool>
++  <pool type='vitastor' supported='no'>
++    <volOptions>
++      <defaultFormat type='raw'/>
++      <enum name='targetFormatType'>
++      </enum>
++    </volOptions>
++  </pool>
+ </storagepoolCapabilities>
+diff --git a/tests/storagepoolcapsschemadata/poolcaps-full.xml b/tests/storagepoolcapsschemadata/poolcaps-full.xml
+index 805950a..852df0d 100644
+--- a/tests/storagepoolcapsschemadata/poolcaps-full.xml
++++ b/tests/storagepoolcapsschemadata/poolcaps-full.xml
+@@ -204,4 +204,11 @@
+       </enum>
+     </volOptions>
+   </pool>
++  <pool type='vitastor' supported='yes'>
++    <volOptions>
++      <defaultFormat type='raw'/>
++      <enum name='targetFormatType'>
++      </enum>
++    </volOptions>
++  </pool>
+ </storagepoolCapabilities>
+diff --git a/tests/storagepoolxml2argvtest.c b/tests/storagepoolxml2argvtest.c
+index 449b745..7f95cc8 100644
+--- a/tests/storagepoolxml2argvtest.c
++++ b/tests/storagepoolxml2argvtest.c
+@@ -68,6 +68,7 @@ testCompareXMLToArgvFiles(bool shouldFail,
+     case VIR_STORAGE_POOL_GLUSTER:
+     case VIR_STORAGE_POOL_ZFS:
+     case VIR_STORAGE_POOL_VSTORAGE:
++    case VIR_STORAGE_POOL_VITASTOR:
+     case VIR_STORAGE_POOL_LAST:
+     default:
+         VIR_TEST_DEBUG("pool type '%s' has no xml2argv test", defTypeStr);
+diff --git a/tools/virsh-pool.c b/tools/virsh-pool.c
+index 18f3839..c8e1436 100644
+--- a/tools/virsh-pool.c
++++ b/tools/virsh-pool.c
+@@ -1231,6 +1231,9 @@ cmdPoolList(vshControl *ctl, const vshCmd *cmd G_GNUC_UNUSED)
+             case VIR_STORAGE_POOL_VSTORAGE:
+                 flags |= VIR_CONNECT_LIST_STORAGE_POOLS_VSTORAGE;
+                 break;
++            case VIR_STORAGE_POOL_VITASTOR:
++                flags |= VIR_CONNECT_LIST_STORAGE_POOLS_VITASTOR;
++                break;
+             case VIR_STORAGE_POOL_LAST:
+                 break;
+             }

cinder-driver/libvirt-example.xml

@@ -0,0 +1,32 @@
+<!-- Example libvirt VM configuration with Vitastor disk -->
+<domain type='kvm'>
+  <name>debian9</name>
+  <uuid>96f277fb-fd9c-49da-bf21-a5cfd54eb162</uuid>
+  <memory unit="KiB">524288</memory>
+  <currentMemory>524288</currentMemory>
+  <vcpu>1</vcpu>
+  <os>
+    <type arch='x86_64'>hvm</type>
+    <boot dev='hd' />
+  </os>
+  <devices>
+    <emulator>/usr/bin/qemu-system-x86_64</emulator>
+    <disk type='network' device='disk'>
+      <target dev='vda' bus='virtio' />
+      <driver name='qemu' type='raw' />
+      <!-- name is Vitastor image name -->
+      <!-- config (optional) is the path to Vitastor's configuration file -->
+      <!-- query (optional) is Vitastor's etcd_prefix -->
+      <source protocol='vitastor' name='debian9' query='/vitastor' config='/etc/vitastor/vitastor.conf'>
+        <!-- hosts = etcd addresses -->
+        <host name='192.168.7.2' port='2379' />
+      </source>
+      <!-- required because Vitastor only supports 4k physical sectors -->
+      <blockio physical_block_size="4096" logical_block_size="512" />
+    </disk>
+    <interface type='network'>
+      <source network='default' />
+    </interface>
+    <graphics type='vnc' port='-1' />
+  </devices>
+</domain>

mon/lp-optimizer.js

@@ -244,6 +244,7 @@ async function optimize_change({ prev_pgs: prev_int_pgs, osd_tree, pg_size = 3,
     {
         return null;
     }
+    // FIXME: use parity_chunks with parity_space instead of pg_minsize
     const pg_effsize = Math.min(pg_minsize, Object.keys(osd_tree).length)
         + Math.max(0, Math.min(pg_size, Object.keys(osd_tree).length) - pg_minsize) * parity_space;
     const pg_count = prev_int_pgs.length;

mon/mon.js

@@ -36,13 +36,15 @@ const etcd_allow = new RegExp('^'+[
     'history/last_clean_pgs',
     'inode/stats/[1-9]\\d*/[1-9]\\d*',
     'stats',
+    'index/image/.*',
+    'index/maxid/[1-9]\\d*',
 ].join('$|^')+'$');
 
 const etcd_tree = {
     config: {
         /* global: {
             // WARNING: NOT ALL OF THESE ARE ACTUALLY CONFIGURABLE HERE
-            // THIS IS JUST A POOR'S MAN CONFIG DOCUMENTATION
+            // THIS IS JUST A POOR MAN'S CONFIG DOCUMENTATION
             // etcd connection
             config_path: "/etc/vitastor/vitastor.conf",
             etcd_address: "10.0.115.10:2379/v3",
@@ -257,14 +259,26 @@ const etcd_tree = {
     },
     inode: {
         stats: {
-            /* <inode_t>: {
-                raw_used: uint64_t, // raw used bytes on OSDs
-                read: { count: uint64_t, usec: uint64_t, bytes: uint64_t },
-                write: { count: uint64_t, usec: uint64_t, bytes: uint64_t },
-                delete: { count: uint64_t, usec: uint64_t, bytes: uint64_t },
+            /* <pool_id>: {
+                <inode_t>: {
+                    raw_used: uint64_t, // raw used bytes on OSDs
+                    read: { count: uint64_t, usec: uint64_t, bytes: uint64_t },
+                    write: { count: uint64_t, usec: uint64_t, bytes: uint64_t },
+                    delete: { count: uint64_t, usec: uint64_t, bytes: uint64_t },
+                },
             }, */
         },
     },
+    pool: {
+        stats: {
+            /* <pool_id>: {
+                used_raw_tb: float, // used raw space in the pool
+                total_raw_tb: float, // maximum amount of space in the pool
+                raw_to_usable: float, // raw to usable ratio
+                space_efficiency: float, // 0..1
+            } */
+        },
+    },
     stats: {
         /* op_stats: {
             <string>: { count: uint64_t, usec: uint64_t, bytes: uint64_t },
@@ -287,6 +301,17 @@ const etcd_tree = {
     history: {
         last_clean_pgs: {},
     },
+    index: {
+        image: {
+            /* <name>: {
+                id: uint64_t,
+                pool_id: uint64_t,
+            }, */
+        },
+        maxid: {
+            /* <pool_id>: uint64_t, */
+        },
+    },
 };
 
 // FIXME Split into several files
@@ -361,6 +386,11 @@ class Mon
         {
             this.config.mon_stats_timeout = 100;
         }
+        this.config.mon_stats_interval = Number(this.config.mon_stats_interval) || 5000;
+        if (this.config.mon_stats_interval < 100)
+        {
+            this.config.mon_stats_interval = 100;
+        }
         // After this number of seconds, a dead OSD will be removed from PG distribution
         this.config.osd_out_time = Number(this.config.osd_out_time) || 0;
         if (!this.config.osd_out_time)
@@ -1025,6 +1055,17 @@ class Mon
                     } });
                 }
                 LPOptimizer.print_change_stats(optimize_result);
+                const pg_effsize = Math.min(pool_cfg.pg_size, Object.keys(pool_tree).length);
+                this.state.pool.stats[pool_id] = {
+                    used_raw_tb: (this.state.pool.stats[pool_id]||{}).used_raw_tb || 0,
+                    total_raw_tb: optimize_result.space,
+                    raw_to_usable: pg_effsize / (pool_cfg.pg_size - (pool_cfg.parity_chunks||0)),
+                    space_efficiency: optimize_result.space/(optimize_result.total_space||1),
+                };
+                etcd_request.success.push({ requestPut: {
+                    key: b64(this.etcd_prefix+'/pool/stats/'+pool_id),
+                    value: b64(JSON.stringify(this.state.pool.stats[pool_id])),
+                } });
                 this.save_new_pgs_txn(etcd_request, pool_id, up_osds, real_prev_pgs, optimize_result.int_pgs, pg_history);
             }
             this.state.config.pgs.hash = tree_hash;
@@ -1131,7 +1172,7 @@ class Mon
         }, this.config.mon_change_timeout || 1000);
     }
 
-    sum_stats()
+    sum_op_stats()
     {
         const op_stats = {}, subop_stats = {}, recovery_stats = {};
         for (const osd in this.state.osd.stats)
@@ -1192,18 +1233,31 @@ class Mon
             write: { count: 0n, usec: 0n, bytes: 0n },
             delete: { count: 0n, usec: 0n, bytes: 0n },
         });
+        for (const pool_id in this.state.config.pools)
+        {
+            this.state.pool.stats[pool_id] = this.state.pool.stats[pool_id] || {};
+            this.state.pool.stats[pool_id].used_raw_tb = 0n;
+        }
         for (const osd_num in this.state.osd.space)
         {
             for (const pool_id in this.state.osd.space[osd_num])
             {
+                this.state.pool.stats[pool_id] = this.state.pool.stats[pool_id] || { used_raw_tb: 0n };
                 inode_stats[pool_id] = inode_stats[pool_id] || {};
                 for (const inode_num in this.state.osd.space[osd_num][pool_id])
                 {
+                    const u = BigInt(this.state.osd.space[osd_num][pool_id][inode_num]||0);
                     inode_stats[pool_id][inode_num] = inode_stats[pool_id][inode_num] || inode_stub();
-                    inode_stats[pool_id][inode_num].raw_used += BigInt(this.state.osd.space[osd_num][pool_id][inode_num]||0);
+                    inode_stats[pool_id][inode_num].raw_used += u;
+                    this.state.pool.stats[pool_id].used_raw_tb += u;
                 }
             }
         }
+        for (const pool_id in this.state.config.pools)
+        {
+            const used = this.state.pool.stats[pool_id].used_raw_tb;
+            this.state.pool.stats[pool_id].used_raw_tb = Number(used)/1024/1024/1024/1024;
+        }
         for (const osd_num in this.state.osd.inodestats)
         {
             const ist = this.state.osd.inodestats[osd_num];
@@ -1275,7 +1329,7 @@ class Mon
     async update_total_stats()
     {
         const txn = [];
-        const stats = this.sum_stats();
+        const stats = this.sum_op_stats();
         const object_counts = this.sum_object_counts();
         const inode_stats = this.sum_inode_stats();
         this.fix_stat_overflows(stats, (this.prev_stats = this.prev_stats || {}));
@@ -1294,6 +1348,13 @@ class Mon
                 } });
             }
         }
+        for (const pool_id in this.state.pool.stats)
+        {
+            txn.push({ requestPut: {
+                key: b64(this.etcd_prefix+'/pool/stats/'+pool_id),
+                value: b64(JSON.stringify(this.state.pool.stats[pool_id])),
+            } });
+        }
         if (txn.length)
         {
             await this.etcd_call('/kv/txn', { success: txn }, this.config.etcd_mon_timeout, 0);
@@ -1307,11 +1368,17 @@ class Mon
             clearTimeout(this.stats_timer);
             this.stats_timer = null;
         }
+        let sleep = (this.stats_update_next||0) - Date.now();
+        if (sleep < this.config.mon_stats_timeout)
+        {
+            sleep = this.config.mon_stats_timeout;
+        }
         this.stats_timer = setTimeout(() =>
         {
             this.stats_timer = null;
+            this.stats_update_next = Date.now() + this.config.mon_stats_interval;
             this.update_total_stats().catch(console.error);
-        }, this.config.mon_stats_timeout || 1000);
+        }, sleep);
     }
 
     parse_kv(kv)

qemu-3.1-vitastor.patch

@@ -11,7 +11,7 @@ Index: qemu-3.1+dfsg/qapi/block-core.json
              'host_cdrom', 'host_device', 'http', 'https', 'iscsi', 'luks',
              'nbd', 'nfs', 'null-aio', 'null-co', 'nvme', 'parallels', 'qcow',
              'qcow2', 'qed', 'quorum', 'raw', 'rbd', 'replication', 'sheepdog',
-@@ -3367,6 +3367,24 @@
+@@ -3367,6 +3367,28 @@
              '*tag': 'str' } }
  
  ##
@@ -19,17 +19,21 @@ Index: qemu-3.1+dfsg/qapi/block-core.json
 +#
 +# Driver specific block device options for vitastor
 +#
++# @image:       Image name
 +# @inode:       Inode number
 +# @pool:        Pool ID
 +# @size:        Desired image size in bytes
-+# @etcd_host:   etcd connection address
++# @config_path: Path to Vitastor configuration
++# @etcd_address: etcd connection address(es)
 +# @etcd_prefix: etcd key/value prefix
 +##
 +{ 'struct': 'BlockdevOptionsVitastor',
-+  'data': { 'inode': 'uint64',
-+            'pool': 'uint64',
-+            'size': 'uint64',
-+            'etcd_host': 'str',
++  'data': { '*inode': 'uint64',
++            '*pool': 'uint64',
++            '*size': 'uint64',
++            '*image': 'str',
++            '*config_path': 'str',
++            '*etcd_address': 'str',
 +            '*etcd_prefix': 'str' } }
 +
 +##

qemu-4.2-vitastor.patch

@@ -11,7 +11,7 @@ Index: qemu/qapi/block-core.json
              'ssh', 'throttle', 'vdi', 'vhdx', 'vmdk', 'vpc', 'vvfat', 'vxhs' ] }
  
  ##
-@@ -3725,6 +3725,24 @@
+@@ -3725,6 +3725,28 @@
              '*tag': 'str' } }
  
  ##
@@ -19,17 +19,21 @@ Index: qemu/qapi/block-core.json
 +#
 +# Driver specific block device options for vitastor
 +#
++# @image:       Image name
 +# @inode:       Inode number
 +# @pool:        Pool ID
 +# @size:        Desired image size in bytes
-+# @etcd_host:   etcd connection address
++# @config_path: Path to Vitastor configuration
++# @etcd_address: etcd connection address(es)
 +# @etcd_prefix: etcd key/value prefix
 +##
 +{ 'struct': 'BlockdevOptionsVitastor',
-+  'data': { 'inode': 'uint64',
-+            'pool': 'uint64',
-+            'size': 'uint64',
-+            'etcd_host': 'str',
++  'data': { '*inode': 'uint64',
++            '*pool': 'uint64',
++            '*size': 'uint64',
++            '*image': 'str',
++            '*config_path': 'str',
++            '*etcd_address': 'str',
 +            '*etcd_prefix': 'str' } }
 +
 +##

qemu-5.0-vitastor.patch

@@ -11,7 +11,7 @@ Index: qemu/qapi/block-core.json
              'ssh', 'throttle', 'vdi', 'vhdx', 'vmdk', 'vpc', 'vvfat', 'vxhs' ] }
  
  ##
-@@ -3635,6 +3635,24 @@
+@@ -3635,6 +3635,28 @@
              '*tag': 'str' } }
  
  ##
@@ -19,17 +19,21 @@ Index: qemu/qapi/block-core.json
 +#
 +# Driver specific block device options for vitastor
 +#
++# @image:       Image name
 +# @inode:       Inode number
 +# @pool:        Pool ID
 +# @size:        Desired image size in bytes
-+# @etcd_host:   etcd connection address
++# @config_path: Path to Vitastor configuration
++# @etcd_address: etcd connection address(es)
 +# @etcd_prefix: etcd key/value prefix
 +##
 +{ 'struct': 'BlockdevOptionsVitastor',
-+  'data': { 'inode': 'uint64',
-+            'pool': 'uint64',
-+            'size': 'uint64',
-+            'etcd_host': 'str',
++  'data': { '*inode': 'uint64',
++            '*pool': 'uint64',
++            '*size': 'uint64',
++            '*image': 'str',
++            '*config_path': 'str',
++            '*etcd_address': 'str',
 +            '*etcd_prefix': 'str' } }
 +
 +##

qemu-5.1-vitastor.patch

@@ -11,7 +11,7 @@ Index: qemu-5.1+dfsg/qapi/block-core.json
              'ssh', 'throttle', 'vdi', 'vhdx', 'vmdk', 'vpc', 'vvfat' ] }
  
  ##
-@@ -3644,6 +3644,24 @@
+@@ -3644,6 +3644,28 @@
              '*tag': 'str' } }
  
  ##
@@ -19,17 +19,21 @@ Index: qemu-5.1+dfsg/qapi/block-core.json
 +#
 +# Driver specific block device options for vitastor
 +#
++# @image:       Image name
 +# @inode:       Inode number
 +# @pool:        Pool ID
 +# @size:        Desired image size in bytes
-+# @etcd_host:   etcd connection address
++# @config_path: Path to Vitastor configuration
++# @etcd_address: etcd connection address(es)
 +# @etcd_prefix: etcd key/value prefix
 +##
 +{ 'struct': 'BlockdevOptionsVitastor',
-+  'data': { 'inode': 'uint64',
-+            'pool': 'uint64',
-+            'size': 'uint64',
-+            'etcd_host': 'str',
++  'data': { '*inode': 'uint64',
++            '*pool': 'uint64',
++            '*size': 'uint64',
++            '*image': 'str',
++            '*config_path': 'str',
++            '*etcd_address': 'str',
 +            '*etcd_prefix': 'str' } }
 +
 +##

rpm/vitastor-el7.spec

@@ -64,6 +64,7 @@ cp -r mon %buildroot/usr/lib/vitastor/mon
 %_libdir/libfio_vitastor_sec.so
 %_libdir/libvitastor_blk.so*
 %_libdir/libvitastor_client.so*
+%_includedir/vitastor_c.h
 /usr/lib/vitastor
 
 

rpm/vitastor-el8.spec

@@ -61,6 +61,7 @@ cp -r mon %buildroot/usr/lib/vitastor
 %_libdir/libfio_vitastor_sec.so
 %_libdir/libvitastor_blk.so*
 %_libdir/libvitastor_client.so*
+%_includedir/vitastor_c.h
 /usr/lib/vitastor
 
 

src/CMakeLists.txt

@@ -4,6 +4,8 @@ project(vitastor)
 
 include(GNUInstallDirs)
 
+set(WITH_QEMU true CACHE BOOL "Build QEMU driver")
+set(WITH_FIO true CACHE BOOL "Build FIO driver")
 set(QEMU_PLUGINDIR qemu CACHE STRING "QEMU plugin directory suffix (qemu-kvm on RHEL)")
 set(WITH_ASAN false CACHE BOOL "Build with AddressSanitizer")
 if("${CMAKE_INSTALL_PREFIX}" MATCHES "^/usr/local/?$")
@@ -36,7 +38,9 @@ string(REGEX REPLACE "([\\/\\-]D) *NDEBUG" "" CMAKE_C_FLAGS_RELWITHDEBINFO "${CM
 
 find_package(PkgConfig)
 pkg_check_modules(LIBURING REQUIRED liburing)
-pkg_check_modules(GLIB REQUIRED glib-2.0)
+if (${WITH_QEMU})
+	pkg_check_modules(GLIB REQUIRED glib-2.0)
+endif (${WITH_QEMU})
 pkg_check_modules(IBVERBS libibverbs)
 if (IBVERBS_LIBRARIES)
 	add_definitions(-DWITH_RDMA)
@@ -62,14 +66,16 @@ target_link_libraries(vitastor_blk
 )
 set_target_properties(vitastor_blk PROPERTIES VERSION ${VERSION} SOVERSION 0)
 
-# libfio_vitastor_blk.so
-add_library(fio_vitastor_blk SHARED
-	fio_engine.cpp
-	../json11/json11.cpp
-)
-target_link_libraries(fio_vitastor_blk
-	vitastor_blk
-)
+if (${WITH_FIO})
+	# libfio_vitastor_blk.so
+	add_library(fio_vitastor_blk SHARED
+		fio_engine.cpp
+		../json11/json11.cpp
+	)
+	target_link_libraries(fio_vitastor_blk
+		vitastor_blk
+	)
+endif (${WITH_FIO})
 
 # libvitastor_common.a
 set(MSGR_RDMA "")
@@ -96,19 +102,23 @@ target_link_libraries(vitastor-osd
 	${IBVERBS_LIBRARIES}
 )
 
-# libfio_vitastor_sec.so
-add_library(fio_vitastor_sec SHARED
-	fio_sec_osd.cpp
-	rw_blocking.cpp
-)
-target_link_libraries(fio_vitastor_sec
-	tcmalloc_minimal
-)
+if (${WITH_FIO})
+	# libfio_vitastor_sec.so
+	add_library(fio_vitastor_sec SHARED
+		fio_sec_osd.cpp
+		rw_blocking.cpp
+	)
+	target_link_libraries(fio_vitastor_sec
+		tcmalloc_minimal
+	)
+endif (${WITH_FIO})
 
 # libvitastor_client.so
 add_library(vitastor_client SHARED
 	cluster_client.cpp
+	vitastor_c.cpp
 )
+set_target_properties(vitastor_client PROPERTIES PUBLIC_HEADER "vitastor_c.h")
 target_link_libraries(vitastor_client
 	vitastor_common
 	tcmalloc_minimal
@@ -117,13 +127,15 @@ target_link_libraries(vitastor_client
 )
 set_target_properties(vitastor_client PROPERTIES VERSION ${VERSION} SOVERSION 0)
 
-# libfio_vitastor.so
-add_library(fio_vitastor SHARED
-	fio_cluster.cpp
-)
-target_link_libraries(fio_vitastor
-	vitastor_client
-)
+if (${WITH_FIO})
+	# libfio_vitastor.so
+	add_library(fio_vitastor SHARED
+		fio_cluster.cpp
+	)
+	target_link_libraries(fio_vitastor
+		vitastor_client
+	)
+endif (${WITH_FIO})
 
 # vitastor-nbd
 add_executable(vitastor-nbd
@@ -146,27 +158,24 @@ add_executable(vitastor-dump-journal
 	dump_journal.cpp crc32c.c
 )
 
-# qemu_driver.so
-add_library(qemu_proxy STATIC qemu_proxy.cpp)
-target_compile_options(qemu_proxy PUBLIC -fPIC)
-target_include_directories(qemu_proxy PUBLIC
-	../qemu/b/qemu
-	../qemu/include
-	${GLIB_INCLUDE_DIRS}
-)
-target_link_libraries(qemu_proxy
-	vitastor_client
-)
-add_library(qemu_vitastor SHARED
-	qemu_driver.c
-)
-target_link_libraries(qemu_vitastor
-	qemu_proxy
-)
-set_target_properties(qemu_vitastor PROPERTIES
-	PREFIX ""
-	OUTPUT_NAME "block-vitastor"
-)
+if (${WITH_QEMU})
+	# qemu_driver.so
+	add_library(qemu_vitastor SHARED
+		qemu_driver.c
+	)
+	target_include_directories(qemu_vitastor PUBLIC
+		../qemu/b/qemu
+		../qemu/include
+		${GLIB_INCLUDE_DIRS}
+	)
+	target_link_libraries(qemu_vitastor
+		vitastor_client
+	)
+	set_target_properties(qemu_vitastor PROPERTIES
+		PREFIX ""
+		OUTPUT_NAME "block-vitastor"
+	)
+endif (${WITH_QEMU})
 
 ### Test stubs
 
@@ -200,6 +209,14 @@ target_link_libraries(osd_peering_pg_test tcmalloc_minimal)
 # test_allocator
 add_executable(test_allocator test_allocator.cpp allocator.cpp)
 
+# test_cas
+add_executable(test_cas
+	test_cas.cpp
+)
+target_link_libraries(test_cas
+	vitastor_client
+)
+
 # test_cluster_client
 add_executable(test_cluster_client
 	test_cluster_client.cpp
@@ -218,5 +235,14 @@ target_include_directories(test_cluster_client PUBLIC ${CMAKE_SOURCE_DIR}/src/mo
 ### Install
 
 install(TARGETS vitastor-osd vitastor-dump-journal vitastor-nbd vitastor-rm RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
-install(TARGETS fio_vitastor fio_vitastor_blk fio_vitastor_sec vitastor_blk vitastor_client LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR})
-install(TARGETS qemu_vitastor LIBRARY DESTINATION /usr/${CMAKE_INSTALL_LIBDIR}/${QEMU_PLUGINDIR})
+install(
+	TARGETS vitastor_blk vitastor_client
+	LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
+	PUBLIC_HEADER DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}
+)
+if (${WITH_FIO})
+	install(TARGETS fio_vitastor fio_vitastor_blk fio_vitastor_sec LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR})
+endif (${WITH_FIO})
+if (${WITH_QEMU})
+	install(TARGETS qemu_vitastor LIBRARY DESTINATION /usr/${CMAKE_INSTALL_LIBDIR}/${QEMU_PLUGINDIR})
+endif (${WITH_QEMU})

src/cluster_client.cpp

@@ -633,6 +633,13 @@ resume_1:
     // Slice the operation into parts
     slice_rw(op);
     op->needs_reslice = false;
+    if (op->opcode == OSD_OP_WRITE && op->version && op->parts.size() > 1)
+    {
+        // Atomic writes to multiple stripes are unsupported
+        op->retval = -EINVAL;
+        erase_op(op);
+        return 1;
+    }
 resume_2:
     // Send unsent parts, if they're not subject to change
     op->state = 3;
@@ -688,13 +695,16 @@ resume_3:
             // Check parent inode
             auto ino_it = st_cli.inode_config.find(op->cur_inode);
             while (ino_it != st_cli.inode_config.end() && ino_it->second.parent_id &&
-                INODE_POOL(ino_it->second.parent_id) == INODE_POOL(op->cur_inode))
+                INODE_POOL(ino_it->second.parent_id) == INODE_POOL(op->cur_inode) &&
+                // Check for loops
+                ino_it->second.parent_id != op->inode)
             {
                 // Skip parents from the same pool
                 ino_it = st_cli.inode_config.find(ino_it->second.parent_id);
             }
             if (ino_it != st_cli.inode_config.end() &&
-                ino_it->second.parent_id)
+                ino_it->second.parent_id &&
+                ino_it->second.parent_id != op->inode)
             {
                 // Continue reading from the parent inode
                 op->cur_inode = ino_it->second.parent_id;
@@ -922,6 +932,7 @@ bool cluster_client_t::try_send(cluster_op_t *op, int i)
                     .offset = part->offset,
                     .len = part->len,
                     .meta_revision = meta_rev,
+                    .version = op->opcode == OSD_OP_WRITE ? op->version : 0,
                 } },
                 .bitmap = op->opcode == OSD_OP_WRITE ? NULL : op->part_bitmaps + pg_bitmap_size*i,
                 .bitmap_len = (unsigned)(op->opcode == OSD_OP_WRITE ? 0 : pg_bitmap_size),
@@ -1072,10 +1083,6 @@ void cluster_client_t::handle_op_part(cluster_op_part_t *part)
     if (part->op.reply.hdr.retval != expected)
     {
         // Operation failed, retry
-        fprintf(
-            stderr, "%s operation failed on OSD %lu: retval=%ld (expected %d), dropping connection\n",
-            osd_op_names[part->op.req.hdr.opcode], part->osd_num, part->op.reply.hdr.retval, expected
-        );
         if (part->op.reply.hdr.retval == -EPIPE)
         {
             // Mark op->up_wait = true before stopping the client
@@ -1094,7 +1101,14 @@ void cluster_client_t::handle_op_part(cluster_op_part_t *part)
             // Don't overwrite other errors with -EPIPE
             op->retval = part->op.reply.hdr.retval;
         }
-        msgr.stop_client(part->op.peer_fd);
+        if (op->retval != -EINTR && op->retval != -EIO)
+        {
+            fprintf(
+                stderr, "%s operation failed on OSD %lu: retval=%ld (expected %d), dropping connection\n",
+                osd_op_names[part->op.req.hdr.opcode], part->osd_num, part->op.reply.hdr.retval, expected
+            );
+            msgr.stop_client(part->op.peer_fd);
+        }
         part->flags |= PART_ERROR;
     }
     else
@@ -1106,6 +1120,7 @@ void cluster_client_t::handle_op_part(cluster_op_part_t *part)
         if (op->opcode == OSD_OP_READ)
         {
             copy_part_bitmap(op, part);
+            op->version = op->parts.size() == 1 ? part->op.reply.rw.version : 0;
         }
     }
     if (op->inflight_count == 0)

src/cluster_client.h

@@ -31,6 +31,9 @@ struct cluster_op_t
     uint64_t inode;
     uint64_t offset;
     uint64_t len;
+    // for reads and writes within a single object (stripe),
+    // reads can return current version and writes can use "CAS" semantics
+    uint64_t version = 0;
     int retval;
     osd_op_buf_list_t iov;
     std::function<void(cluster_op_t*)> callback;

src/fio_cluster.cpp

@@ -25,20 +25,17 @@
 
 #include <vector>
 
-#include "epoll_manager.h"
-#include "cluster_client.h"
+#include "vitastor_c.h"
 #include "fio_headers.h"
 
 struct sec_data
 {
-    ring_loop_t *ringloop = NULL;
-    epoll_manager_t *epmgr = NULL;
-    cluster_client_t *cli = NULL;
-    inode_watch_t *watch = NULL;
+    vitastor_c *cli = NULL;
+    void *watch = NULL;
     bool last_sync = false;
     /* The list of completed io_u structs. */
     std::vector<io_u*> completed;
-    uint64_t op_n = 0, inflight = 0;
+    uint64_t inflight = 0;
     bool trace = false;
 };
 
@@ -189,6 +186,12 @@ static struct fio_option options[] = {
     },
 };
 
+static void watch_callback(void *opaque, long watch)
+{
+    struct sec_data *bsd = (struct sec_data*)opaque;
+    bsd->watch = (void*)watch;
+}
+
 static int sec_setup(struct thread_data *td)
 {
     sec_options *o = (sec_options*)td->eo;
@@ -209,27 +212,6 @@ static int sec_setup(struct thread_data *td)
         td->o.open_files++;
     }
 
-    json11::Json::object cfg;
-    if (o->config_path)
-        cfg["config_path"] = std::string(o->config_path);
-    if (o->etcd_host)
-        cfg["etcd_address"] = std::string(o->etcd_host);
-    if (o->etcd_prefix)
-        cfg["etcd_prefix"] = std::string(o->etcd_prefix);
-    if (o->rdma_device)
-        cfg["rdma_device"] = std::string(o->rdma_device);
-    if (o->rdma_port_num)
-        cfg["rdma_port_num"] = o->rdma_port_num;
-    if (o->rdma_gid_index)
-        cfg["rdma_gid_index"] = o->rdma_gid_index;
-    if (o->rdma_mtu)
-        cfg["rdma_mtu"] = o->rdma_mtu;
-    if (o->cluster_log)
-        cfg["log_level"] = o->cluster_log;
-    if (o->use_rdma != -1)
-        cfg["use_rdma"] = o->use_rdma;
-    json11::Json cfg_json(cfg);
-
     if (!o->image)
     {
         if (!(o->inode & ((1l << (64-POOL_ID_BITS)) - 1)))
@@ -251,20 +233,20 @@ static int sec_setup(struct thread_data *td)
     {
         o->inode = 0;
     }
-    bsd->ringloop = new ring_loop_t(512);
-    bsd->epmgr = new epoll_manager_t(bsd->ringloop);
-    bsd->cli = new cluster_client_t(bsd->ringloop, bsd->epmgr->tfd, cfg_json);
+    bsd->cli = vitastor_c_create_uring(o->config_path, o->etcd_host, o->etcd_prefix,
+        o->use_rdma, o->rdma_device, o->rdma_port_num, o->rdma_gid_index, o->rdma_mtu, o->cluster_log);
     if (o->image)
     {
-        while (!bsd->cli->is_ready())
+        bsd->watch = NULL;
+        vitastor_c_watch_inode(bsd->cli, o->image, watch_callback, bsd);
+        while (true)
         {
-            bsd->ringloop->loop();
-            if (bsd->cli->is_ready())
+            vitastor_c_uring_handle_events(bsd->cli);
+            if (bsd->watch)
                 break;
-            bsd->ringloop->wait();
+            vitastor_c_uring_wait_events(bsd->cli);
         }
-        bsd->watch = bsd->cli->st_cli.watch_inode(std::string(o->image));
-        td->files[0]->real_file_size = bsd->watch->cfg.size;
+        td->files[0]->real_file_size = vitastor_c_inode_get_size(bsd->watch);
     }
 
     bsd->trace = o->trace ? true : false;
@@ -279,11 +261,9 @@ static void sec_cleanup(struct thread_data *td)
     {
         if (bsd->watch)
         {
-            bsd->cli->st_cli.close_watch(bsd->watch);
+            vitastor_c_close_watch(bsd->cli, bsd->watch);
         }
-        delete bsd->cli;
-        delete bsd->epmgr;
-        delete bsd->ringloop;
+        vitastor_c_destroy(bsd->cli);
         delete bsd;
     }
 }
@@ -294,12 +274,31 @@ static int sec_init(struct thread_data *td)
     return 0;
 }
 
+static void io_callback(void *opaque, long retval)
+{
+    struct io_u *io = (struct io_u*)opaque;
+    io->error = retval < 0 ? -retval : 0;
+    sec_data *bsd = (sec_data*)io->engine_data;
+    bsd->inflight--;
+    bsd->completed.push_back(io);
+    if (bsd->trace)
+    {
+        printf("--- %s 0x%lx retval=%ld\n", io->ddir == DDIR_READ ? "READ" :
+            (io->ddir == DDIR_WRITE ? "WRITE" : "SYNC"), (uint64_t)io, retval);
+    }
+}
+
+static void read_callback(void *opaque, long retval, uint64_t version)
+{
+    io_callback(opaque, retval);
+}
+
 /* Begin read or write request. */
 static enum fio_q_status sec_queue(struct thread_data *td, struct io_u *io)
 {
     sec_options *opt = (sec_options*)td->eo;
     sec_data *bsd = (sec_data*)td->io_ops_data;
-    int n = bsd->op_n;
+    struct iovec iov;
 
     fio_ro_check(td, io);
     if (io->ddir == DDIR_SYNC && bsd->last_sync)
@@ -308,32 +307,29 @@ static enum fio_q_status sec_queue(struct thread_data *td, struct io_u *io)
     }
 
     io->engine_data = bsd;
-    cluster_op_t *op = new cluster_op_t;
+    io->error = 0;
+    bsd->inflight++;
 
-    op->inode = opt->image ? bsd->watch->cfg.num : opt->inode;
+    uint64_t inode = opt->image ? vitastor_c_inode_get_num(bsd->watch) : opt->inode;
     switch (io->ddir)
     {
     case DDIR_READ:
-        op->opcode = OSD_OP_READ;
-        op->offset = io->offset;
-        op->len = io->xfer_buflen;
-        op->iov.push_back(io->xfer_buf, io->xfer_buflen);
+        iov = { .iov_base = io->xfer_buf, .iov_len = io->xfer_buflen };
+        vitastor_c_read(bsd->cli, inode, io->offset, io->xfer_buflen, &iov, 1, read_callback, io);
         bsd->last_sync = false;
         break;
     case DDIR_WRITE:
-        if (opt->image && bsd->watch->cfg.readonly)
+        if (opt->image && vitastor_c_inode_get_readonly(bsd->watch))
         {
             io->error = EROFS;
             return FIO_Q_COMPLETED;
         }
-        op->opcode = OSD_OP_WRITE;
-        op->offset = io->offset;
-        op->len = io->xfer_buflen;
-        op->iov.push_back(io->xfer_buf, io->xfer_buflen);
+        iov = { .iov_base = io->xfer_buf, .iov_len = io->xfer_buflen };
+        vitastor_c_write(bsd->cli, inode, io->offset, io->xfer_buflen, 0, &iov, 1, io_callback, io);
         bsd->last_sync = false;
         break;
     case DDIR_SYNC:
-        op->opcode = OSD_OP_SYNC;
+        vitastor_c_sync(bsd->cli, io_callback, io);
         bsd->last_sync = true;
         break;
     default:
@@ -341,39 +337,20 @@ static enum fio_q_status sec_queue(struct thread_data *td, struct io_u *io)
         return FIO_Q_COMPLETED;
     }
 
-    op->callback = [io, n](cluster_op_t *op)
-    {
-        io->error = op->retval < 0 ? -op->retval : 0;
-        sec_data *bsd = (sec_data*)io->engine_data;
-        bsd->inflight--;
-        bsd->completed.push_back(io);
-        if (bsd->trace)
-        {
-            printf("--- %s n=%d retval=%d\n", io->ddir == DDIR_READ ? "READ" :
-                (io->ddir == DDIR_WRITE ? "WRITE" : "SYNC"), n, op->retval);
-        }
-        delete op;
-    };
-
     if (opt->trace)
     {
         if (io->ddir == DDIR_SYNC)
         {
-            printf("+++ SYNC # %d\n", n);
+            printf("+++ SYNC 0x%lx\n", (uint64_t)io);
         }
         else
         {
-            printf("+++ %s # %d 0x%llx+%llx\n",
+            printf("+++ %s 0x%lx 0x%llx+%llx\n",
                 io->ddir == DDIR_READ ? "READ" : "WRITE",
-                n, io->offset, io->xfer_buflen);
+                (uint64_t)io, io->offset, io->xfer_buflen);
         }
     }
 
-    io->error = 0;
-    bsd->inflight++;
-    bsd->op_n++;
-    bsd->cli->execute(op);
-
     if (io->error != 0)
         return FIO_Q_COMPLETED;
     return FIO_Q_QUEUED;
@@ -384,10 +361,10 @@ static int sec_getevents(struct thread_data *td, unsigned int min, unsigned int
     sec_data *bsd = (sec_data*)td->io_ops_data;
     while (true)
     {
-        bsd->ringloop->loop();
+        vitastor_c_uring_handle_events(bsd->cli);
         if (bsd->completed.size() >= min)
             break;
-        bsd->ringloop->wait();
+        vitastor_c_uring_wait_events(bsd->cli);
     }
     return bsd->completed.size();
 }

src/messenger.cpp

@@ -261,7 +261,7 @@ void osd_messenger_t::try_connect_peer_addr(osd_num_t peer_osd, const char *peer
         {
             osd_num_t peer_osd = clients.at(peer_fd)->osd_num;
             stop_client(peer_fd, true);
-            on_connect_peer(peer_osd, -EIO);
+            on_connect_peer(peer_osd, -EPIPE);
             return;
         });
     }

src/osd_ops.h

@@ -191,6 +191,9 @@ struct __attribute__((__packed__)) osd_op_rw_t
     uint32_t flags;
     // inode metadata revision
     uint64_t meta_revision;
+    // object version for atomic "CAS" (compare-and-set) writes
+    // writes and deletes fail with -EINTR if object version differs from (version-1)
+    uint64_t version;
 };
 
 struct __attribute__((__packed__)) osd_reply_rw_t
@@ -199,6 +202,8 @@ struct __attribute__((__packed__)) osd_reply_rw_t
     // for reads: bitmap length
     uint32_t bitmap_len;
     uint32_t pad0;
+    // for reads: object version
+    uint64_t version;
 };
 
 // sync to the primary OSD

src/osd_primary.cpp

@@ -67,7 +67,9 @@ bool osd_t::prepare_primary_rw(osd_op_t *cur_op)
         }
         // Find parents from the same pool. Optimized reads only work within pools
         while (inode_it != st_cli.inode_config.end() && inode_it->second.parent_id &&
-            INODE_POOL(inode_it->second.parent_id) == pg_it->second.pool_id)
+            INODE_POOL(inode_it->second.parent_id) == pg_it->second.pool_id &&
+            // Check for loops
+            inode_it->second.parent_id != cur_op->req.rw.inode)
         {
             chain_size++;
             inode_it = st_cli.inode_config.find(inode_it->second.parent_id);
@@ -123,7 +125,10 @@ bool osd_t::prepare_primary_rw(osd_op_t *cur_op)
         int chain_num = 0;
         op_data->read_chain[chain_num++] = cur_op->req.rw.inode;
         auto inode_it = st_cli.inode_config.find(cur_op->req.rw.inode);
-        while (inode_it != st_cli.inode_config.end() && inode_it->second.parent_id)
+        while (inode_it != st_cli.inode_config.end() && inode_it->second.parent_id &&
+            INODE_POOL(inode_it->second.parent_id) == pg_it->second.pool_id &&
+            // Check for loops
+            inode_it->second.parent_id != cur_op->req.rw.inode)
         {
             op_data->read_chain[chain_num++] = inode_it->second.parent_id;
             inode_it = st_cli.inode_config.find(inode_it->second.parent_id);
@@ -222,6 +227,7 @@ resume_2:
         finish_op(cur_op, op_data->epipe > 0 ? -EPIPE : -EIO);
         return;
     }
+    cur_op->reply.rw.version = op_data->fact_ver;
     cur_op->reply.rw.bitmap_len = op_data->pg_data_size * clean_entry_bitmap_size;
     if (op_data->degraded)
     {
@@ -343,6 +349,12 @@ resume_3:
         pg_cancel_write_queue(pg, cur_op, op_data->oid, op_data->epipe > 0 ? -EPIPE : -EIO);
         return;
     }
+    // Check CAS version
+    if (cur_op->req.rw.version && op_data->fact_ver != (cur_op->req.rw.version-1))
+    {
+        cur_op->reply.hdr.retval = -EINTR;
+        goto continue_others;
+    }
     // Save version override for parallel reads
     pg.ver_override[op_data->oid] = op_data->fact_ver;
     // Submit deletes
@@ -370,6 +382,8 @@ resume_5:
         free_object_state(pg, &op_data->object_state);
     }
     pg.total_count--;
+    cur_op->reply.hdr.retval = 0;
+continue_others:
     osd_op_t *next_op = NULL;
     auto next_it = pg.write_queue.find(op_data->oid);
     if (next_it != pg.write_queue.end() && next_it->second == cur_op)
@@ -378,7 +392,7 @@ resume_5:
         if (next_it != pg.write_queue.end() && next_it->first == op_data->oid)
             next_op = next_it->second;
     }
-    finish_op(cur_op, cur_op->req.rw.len);
+    finish_op(cur_op, cur_op->reply.hdr.retval);
     if (next_op)
     {
         // Continue next write to the same object

src/osd_primary_chain.cpp

@@ -65,7 +65,10 @@ int osd_t::read_bitmaps(osd_op_t *cur_op, pg_t & pg, int base_state)
             auto vo_it = pg.ver_override.find(cur_oid);
             auto read_version = (vo_it != pg.ver_override.end() ? vo_it->second : UINT64_MAX);
             // Read bitmap synchronously from the local database
-            bs->read_bitmap(cur_oid, read_version, op_data->snapshot_bitmaps + chain_num*clean_entry_bitmap_size, NULL);
+            bs->read_bitmap(
+                cur_oid, read_version, op_data->snapshot_bitmaps + chain_num*clean_entry_bitmap_size,
+                !chain_num ? &cur_op->reply.rw.version : NULL
+            );
         }
     }
     else
@@ -228,7 +231,10 @@ int osd_t::submit_bitmap_subops(osd_op_t *cur_op, pg_t & pg)
                 // Read bitmap synchronously from the local database
                 for (int j = prev; j <= i; j++)
                 {
-                    bs->read_bitmap((*bitmap_requests)[j].oid, (*bitmap_requests)[j].version, (*bitmap_requests)[j].bmp_buf, NULL);
+                    bs->read_bitmap(
+                        (*bitmap_requests)[j].oid, (*bitmap_requests)[j].version, (*bitmap_requests)[j].bmp_buf,
+                        (*bitmap_requests)[j].oid.inode == cur_op->req.rw.inode ? &cur_op->reply.rw.version : NULL
+                    );
                 }
             }
             else
@@ -264,6 +270,10 @@ int osd_t::submit_bitmap_subops(osd_op_t *cur_op, pg_t & pg)
                         for (int j = prev; j <= i; j++)
                         {
                             memcpy((*bitmap_requests)[j].bmp_buf, cur_buf, clean_entry_bitmap_size);
+                            if ((*bitmap_requests)[j].oid.inode == cur_op->req.rw.inode)
+                            {
+                                memcpy(&cur_op->reply.rw.version, cur_buf-8, 8);
+                            }
                             cur_buf += 8 + clean_entry_bitmap_size;
                         }
                     }

src/osd_primary_write.cpp

@@ -96,6 +96,12 @@ resume_3:
         pg_cancel_write_queue(pg, cur_op, op_data->oid, op_data->epipe > 0 ? -EPIPE : -EIO);
         return;
     }
+    // Check CAS version
+    if (cur_op->req.rw.version && op_data->fact_ver != (cur_op->req.rw.version-1))
+    {
+        cur_op->reply.hdr.retval = -EINTR;
+        goto continue_others;
+    }
     if (op_data->scheme == POOL_SCHEME_REPLICATED)
     {
         // Set bitmap bits
@@ -265,7 +271,7 @@ continue_others:
             next_op = next_it->second;
     }
     // finish_op would invalidate next_it if it cleared pg.write_queue, but it doesn't do that :)
-    finish_op(cur_op, cur_op->req.rw.len);
+    finish_op(cur_op, cur_op->reply.hdr.retval);
     if (next_op)
     {
         // Continue next write to the same object

src/qemu_driver.c

@@ -26,7 +26,7 @@
 #define qobject_unref QDECREF
 #endif
 
-#include "qemu_proxy.h"
+#include "vitastor_c.h"
 
 void qemu_module_dummy(void)
 {
@@ -48,6 +48,7 @@ typedef struct VitastorClient
     uint64_t pool;
     uint64_t size;
     long readonly;
+    int use_rdma;
     char *rdma_device;
     int rdma_port_num;
     int rdma_gid_index;
@@ -65,10 +66,11 @@ typedef struct VitastorRPC
 } VitastorRPC;
 
 static void vitastor_co_init_task(BlockDriverState *bs, VitastorRPC *task);
-static void vitastor_co_generic_bh_cb(long retval, void *opaque);
+static void vitastor_co_generic_bh_cb(void *opaque, long retval);
+static void vitastor_co_read_cb(void *opaque, long retval, uint64_t version);
 static void vitastor_close(BlockDriverState *bs);
 
-static char *qemu_rbd_next_tok(char *src, char delim, char **p)
+static char *qemu_vitastor_next_tok(char *src, char delim, char **p)
 {
     char *end;
     *p = NULL;
@@ -87,7 +89,7 @@ static char *qemu_rbd_next_tok(char *src, char delim, char **p)
     return src;
 }
 
-static void qemu_rbd_unescape(char *src)
+static void qemu_vitastor_unescape(char *src)
 {
     char *p;
     for (p = src; *src; ++src, ++p)
@@ -100,7 +102,7 @@ static void qemu_rbd_unescape(char *src)
 }
 
 // vitastor[:key=value]*
-// vitastor[:etcd_host=127.0.0.1]:inode=1:pool=1[:rdma_gid_index=3]
+// vitastor[:(etcd|etcd_host|etcd_address)=127.0.0.1]:inode=1:pool=1[:rdma_gid_index=3]
 // vitastor:config_path=/etc/vitastor/vitastor.conf:image=testimg
 static void vitastor_parse_filename(const char *filename, QDict *options, Error **errp)
 {
@@ -120,18 +122,19 @@ static void vitastor_parse_filename(const char *filename, QDict *options, Error
     while (p)
     {
         char *name, *value;
-        name = qemu_rbd_next_tok(p, '=', &p);
+        name = qemu_vitastor_next_tok(p, '=', &p);
         if (!p)
         {
             error_setg(errp, "conf option %s has no value", name);
             break;
         }
-        qemu_rbd_unescape(name);
-        value = qemu_rbd_next_tok(p, ':', &p);
-        qemu_rbd_unescape(value);
+        qemu_vitastor_unescape(name);
+        value = qemu_vitastor_next_tok(p, ':', &p);
+        qemu_vitastor_unescape(value);
         if (!strcmp(name, "inode") ||
             !strcmp(name, "pool") ||
             !strcmp(name, "size") ||
+            !strcmp(name, "use_rdma") ||
             !strcmp(name, "rdma_port_num") ||
             !strcmp(name, "rdma_gid_index") ||
             !strcmp(name, "rdma_mtu"))
@@ -181,7 +184,7 @@ static void coroutine_fn vitastor_co_get_metadata(VitastorRPC *task)
     task->co = qemu_coroutine_self();
 
     qemu_mutex_lock(&client->mutex);
-    vitastor_proxy_watch_metadata(client->proxy, client->image, vitastor_co_generic_bh_cb, task);
+    vitastor_c_watch_inode(client->proxy, client->image, vitastor_co_generic_bh_cb, task);
     qemu_mutex_unlock(&client->mutex);
 
     while (!task->complete)
@@ -196,15 +199,21 @@ static int vitastor_file_open(BlockDriverState *bs, QDict *options, int flags, E
     int64_t ret = 0;
     qemu_mutex_init(&client->mutex);
     client->config_path = g_strdup(qdict_get_try_str(options, "config_path"));
-    client->etcd_host = g_strdup(qdict_get_try_str(options, "etcd_host"));
+    if (qdict_get_try_str(options, "etcd_address"))
+        client->etcd_host = g_strdup(qdict_get_try_str(options, "etcd_address"));
+    else if (qdict_get_try_str(options, "etcd_host"))
+        client->etcd_host = g_strdup(qdict_get_try_str(options, "etcd_host"));
+    else if (qdict_get_try_str(options, "etcd"))
+        client->etcd_host = g_strdup(qdict_get_try_str(options, "etcd"));
     client->etcd_prefix = g_strdup(qdict_get_try_str(options, "etcd_prefix"));
+    client->use_rdma = qdict_get_try_int(options, "use_rdma", -1);
     client->rdma_device = g_strdup(qdict_get_try_str(options, "rdma_device"));
     client->rdma_port_num = qdict_get_try_int(options, "rdma_port_num", 0);
     client->rdma_gid_index = qdict_get_try_int(options, "rdma_gid_index", 0);
     client->rdma_mtu = qdict_get_try_int(options, "rdma_mtu", 0);
-    client->proxy = vitastor_proxy_create(
-        bdrv_get_aio_context(bs), client->config_path, client->etcd_host, client->etcd_prefix,
-        client->rdma_device, client->rdma_port_num, client->rdma_gid_index, client->rdma_mtu
+    client->proxy = vitastor_c_create_qemu(
+        (QEMUSetFDHandler*)aio_set_fd_handler, bdrv_get_aio_context(bs), client->config_path, client->etcd_host, client->etcd_prefix,
+        client->use_rdma, client->rdma_device, client->rdma_port_num, client->rdma_gid_index, client->rdma_mtu, 0
     );
     client->image = g_strdup(qdict_get_try_str(options, "image"));
     client->readonly = (flags & BDRV_O_RDWR) ? 1 : 0;
@@ -224,9 +233,9 @@ static int vitastor_file_open(BlockDriverState *bs, QDict *options, int flags, E
         }
         BDRV_POLL_WHILE(bs, !task.complete);
         client->watch = (void*)task.ret;
-        client->readonly = client->readonly || vitastor_proxy_get_readonly(client->watch);
-        client->size = vitastor_proxy_get_size(client->watch);
-        if (!vitastor_proxy_get_inode_num(client->watch))
+        client->readonly = client->readonly || vitastor_c_inode_get_readonly(client->watch);
+        client->size = vitastor_c_inode_get_size(client->watch);
+        if (!vitastor_c_inode_get_num(client->watch))
         {
             error_setg(errp, "image does not exist");
             vitastor_close(bs);
@@ -255,12 +264,15 @@ static int vitastor_file_open(BlockDriverState *bs, QDict *options, int flags, E
     }
     bs->total_sectors = client->size / BDRV_SECTOR_SIZE;
     //client->aio_context = bdrv_get_aio_context(bs);
+    qdict_del(options, "use_rdma");
     qdict_del(options, "rdma_mtu");
     qdict_del(options, "rdma_gid_index");
     qdict_del(options, "rdma_port_num");
     qdict_del(options, "rdma_device");
     qdict_del(options, "config_path");
     qdict_del(options, "etcd_host");
+    qdict_del(options, "etcd_address");
+    qdict_del(options, "etcd");
     qdict_del(options, "etcd_prefix");
     qdict_del(options, "image");
     qdict_del(options, "inode");
@@ -272,7 +284,7 @@ static int vitastor_file_open(BlockDriverState *bs, QDict *options, int flags, E
 static void vitastor_close(BlockDriverState *bs)
 {
     VitastorClient *client = bs->opaque;
-    vitastor_proxy_destroy(client->proxy);
+    vitastor_c_destroy(client->proxy);
     qemu_mutex_destroy(&client->mutex);
     if (client->config_path)
         g_free(client->config_path);
@@ -387,7 +399,7 @@ static void vitastor_co_init_task(BlockDriverState *bs, VitastorRPC *task)
     };
 }
 
-static void vitastor_co_generic_bh_cb(long retval, void *opaque)
+static void vitastor_co_generic_bh_cb(void *opaque, long retval)
 {
     VitastorRPC *task = opaque;
     task->ret = retval;
@@ -403,6 +415,11 @@ static void vitastor_co_generic_bh_cb(long retval, void *opaque)
     }
 }
 
+static void vitastor_co_read_cb(void *opaque, long retval, uint64_t version)
+{
+    vitastor_co_generic_bh_cb(opaque, retval);
+}
+
 static int coroutine_fn vitastor_co_preadv(BlockDriverState *bs, uint64_t offset, uint64_t bytes, QEMUIOVector *iov, int flags)
 {
     VitastorClient *client = bs->opaque;
@@ -410,9 +427,9 @@ static int coroutine_fn vitastor_co_preadv(BlockDriverState *bs, uint64_t offset
     vitastor_co_init_task(bs, &task);
     task.iov = iov;
 
-    uint64_t inode = client->watch ? vitastor_proxy_get_inode_num(client->watch) : client->inode;
+    uint64_t inode = client->watch ? vitastor_c_inode_get_num(client->watch) : client->inode;
     qemu_mutex_lock(&client->mutex);
-    vitastor_proxy_rw(0, client->proxy, inode, offset, bytes, iov->iov, iov->niov, vitastor_co_generic_bh_cb, &task);
+    vitastor_c_read(client->proxy, inode, offset, bytes, iov->iov, iov->niov, vitastor_co_read_cb, &task);
     qemu_mutex_unlock(&client->mutex);
 
     while (!task.complete)
@@ -430,9 +447,9 @@ static int coroutine_fn vitastor_co_pwritev(BlockDriverState *bs, uint64_t offse
     vitastor_co_init_task(bs, &task);
     task.iov = iov;
 
-    uint64_t inode = client->watch ? vitastor_proxy_get_inode_num(client->watch) : client->inode;
+    uint64_t inode = client->watch ? vitastor_c_inode_get_num(client->watch) : client->inode;
     qemu_mutex_lock(&client->mutex);
-    vitastor_proxy_rw(1, client->proxy, inode, offset, bytes, iov->iov, iov->niov, vitastor_co_generic_bh_cb, &task);
+    vitastor_c_write(client->proxy, inode, offset, bytes, 0, iov->iov, iov->niov, vitastor_co_generic_bh_cb, &task);
     qemu_mutex_unlock(&client->mutex);
 
     while (!task.complete)
@@ -462,7 +479,7 @@ static int coroutine_fn vitastor_co_flush(BlockDriverState *bs)
     vitastor_co_init_task(bs, &task);
 
     qemu_mutex_lock(&client->mutex);
-    vitastor_proxy_sync(client->proxy, vitastor_co_generic_bh_cb, &task);
+    vitastor_c_sync(client->proxy, vitastor_co_generic_bh_cb, &task);
     qemu_mutex_unlock(&client->mutex);
 
     while (!task.complete)
@@ -501,6 +518,8 @@ static const char *vitastor_strong_runtime_opts[] = {
     "inode",
     "pool",
     "config_path",
+    "etcd",
+    "etcd_address",
     "etcd_host",
     "etcd_prefix",
 

src/qemu_proxy.cpp

@@ -1,177 +0,0 @@
-// Copyright (c) Vitaliy Filippov, 2019+
-// License: VNPL-1.1 or GNU GPL-2.0+ (see README.md for details)
-
-// C-C++ proxy for the QEMU driver
-// (QEMU headers don't compile with g++)
-
-#include <sys/epoll.h>
-
-#include "cluster_client.h"
-
-typedef void* AioContext;
-#include "qemu_proxy.h"
-
-extern "C"
-{
-    // QEMU
-    typedef void IOHandler(void *opaque);
-    void aio_set_fd_handler(AioContext *ctx, int fd, int is_external, IOHandler *fd_read, IOHandler *fd_write, void *poll_fn, void *opaque);
-}
-
-struct QemuProxyData
-{
-    int fd;
-    std::function<void(int, int)> callback;
-};
-
-class QemuProxy
-{
-    std::map<int, QemuProxyData> handlers;
-
-public:
-
-    timerfd_manager_t *tfd;
-    cluster_client_t *cli;
-    AioContext *ctx;
-
-    QemuProxy(AioContext *ctx, const char *config_path, const char *etcd_host, const char *etcd_prefix,
-        const char *rdma_device, int rdma_port_num, int rdma_gid_index, int rdma_mtu)
-    {
-        this->ctx = ctx;
-        json11::Json::object cfg;
-        if (config_path)
-            cfg["config_path"] = std::string(config_path);
-        if (etcd_host)
-            cfg["etcd_address"] = std::string(etcd_host);
-        if (etcd_prefix)
-            cfg["etcd_prefix"] = std::string(etcd_prefix);
-        if (rdma_device)
-            cfg["rdma_device"] = std::string(rdma_device);
-        if (rdma_port_num)
-            cfg["rdma_port_num"] = rdma_port_num;
-        if (rdma_gid_index)
-            cfg["rdma_gid_index"] = rdma_gid_index;
-        if (rdma_mtu)
-            cfg["rdma_mtu"] = rdma_mtu;
-        json11::Json cfg_json(cfg);
-        tfd = new timerfd_manager_t([this](int fd, bool wr, std::function<void(int, int)> callback) { set_fd_handler(fd, wr, callback); });
-        cli = new cluster_client_t(NULL, tfd, cfg_json);
-    }
-
-    ~QemuProxy()
-    {
-        delete cli;
-        delete tfd;
-    }
-
-    void set_fd_handler(int fd, bool wr, std::function<void(int, int)> callback)
-    {
-        if (callback != NULL)
-        {
-            handlers[fd] = { .fd = fd, .callback = callback };
-            aio_set_fd_handler(ctx, fd, false, &QemuProxy::read_handler, wr ? &QemuProxy::write_handler : NULL, NULL, &handlers[fd]);
-        }
-        else
-        {
-            handlers.erase(fd);
-            aio_set_fd_handler(ctx, fd, false, NULL, NULL, NULL, NULL);
-        }
-    }
-
-    static void read_handler(void *opaque)
-    {
-        QemuProxyData *data = (QemuProxyData *)opaque;
-        data->callback(data->fd, EPOLLIN);
-    }
-
-    static void write_handler(void *opaque)
-    {
-        QemuProxyData *data = (QemuProxyData *)opaque;
-        data->callback(data->fd, EPOLLOUT);
-    }
-};
-
-extern "C" {
-
-void* vitastor_proxy_create(AioContext *ctx, const char *config_path, const char *etcd_host, const char *etcd_prefix,
-    const char *rdma_device, int rdma_port_num, int rdma_gid_index, int rdma_mtu)
-{
-    QemuProxy *p = new QemuProxy(ctx, config_path, etcd_host, etcd_prefix, rdma_device, rdma_port_num, rdma_gid_index, rdma_mtu);
-    return p;
-}
-
-void vitastor_proxy_destroy(void *client)
-{
-    QemuProxy *p = (QemuProxy*)client;
-    delete p;
-}
-
-void vitastor_proxy_rw(int write, void *client, uint64_t inode, uint64_t offset, uint64_t len,
-    iovec *iov, int iovcnt, VitastorIOHandler cb, void *opaque)
-{
-    QemuProxy *p = (QemuProxy*)client;
-    cluster_op_t *op = new cluster_op_t;
-    op->opcode = write ? OSD_OP_WRITE : OSD_OP_READ;
-    op->inode = inode;
-    op->offset = offset;
-    op->len = len;
-    for (int i = 0; i < iovcnt; i++)
-    {
-        op->iov.push_back(iov[i].iov_base, iov[i].iov_len);
-    }
-    op->callback = [cb, opaque](cluster_op_t *op)
-    {
-        cb(op->retval, opaque);
-        delete op;
-    };
-    p->cli->execute(op);
-}
-
-void vitastor_proxy_sync(void *client, VitastorIOHandler cb, void *opaque)
-{
-    QemuProxy *p = (QemuProxy*)client;
-    cluster_op_t *op = new cluster_op_t;
-    op->opcode = OSD_OP_SYNC;
-    op->callback = [cb, opaque](cluster_op_t *op)
-    {
-        cb(op->retval, opaque);
-        delete op;
-    };
-    p->cli->execute(op);
-}
-
-void vitastor_proxy_watch_metadata(void *client, char *image, VitastorIOHandler cb, void *opaque)
-{
-    QemuProxy *p = (QemuProxy*)client;
-    p->cli->on_ready([=]()
-    {
-        auto watch = p->cli->st_cli.watch_inode(std::string(image));
-        cb((long)watch, opaque);
-    });
-}
-
-void vitastor_proxy_close_watch(void *client, void *watch)
-{
-    QemuProxy *p = (QemuProxy*)client;
-    p->cli->st_cli.close_watch((inode_watch_t*)watch);
-}
-
-uint64_t vitastor_proxy_get_size(void *watch_ptr)
-{
-    inode_watch_t *watch = (inode_watch_t*)watch_ptr;
-    return watch->cfg.size;
-}
-
-uint64_t vitastor_proxy_get_inode_num(void *watch_ptr)
-{
-    inode_watch_t *watch = (inode_watch_t*)watch_ptr;
-    return watch->cfg.num;
-}
-
-int vitastor_proxy_get_readonly(void *watch_ptr)
-{
-    inode_watch_t *watch = (inode_watch_t*)watch_ptr;
-    return watch->cfg.readonly;
-}
-
-}

src/qemu_proxy.h

@@ -1,35 +0,0 @@
-// Copyright (c) Vitaliy Filippov, 2019+
-// License: VNPL-1.1 or GNU GPL-2.0+ (see README.md for details)
-
-#ifndef VITASTOR_QEMU_PROXY_H
-#define VITASTOR_QEMU_PROXY_H
-
-#ifndef POOL_ID_BITS
-#define POOL_ID_BITS 16
-#endif
-#include <stdint.h>
-#include <sys/uio.h>
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-// Our exports
-typedef void VitastorIOHandler(long retval, void *opaque);
-void* vitastor_proxy_create(AioContext *ctx, const char *config_path, const char *etcd_host, const char *etcd_prefix,
-    const char *rdma_device, int rdma_port_num, int rdma_gid_index, int rdma_mtu);
-void vitastor_proxy_destroy(void *client);
-void vitastor_proxy_rw(int write, void *client, uint64_t inode, uint64_t offset, uint64_t len,
-    struct iovec *iov, int iovcnt, VitastorIOHandler cb, void *opaque);
-void vitastor_proxy_sync(void *client, VitastorIOHandler cb, void *opaque);
-void vitastor_proxy_watch_metadata(void *client, char *image, VitastorIOHandler cb, void *opaque);
-void vitastor_proxy_close_watch(void *client, void *watch);
-uint64_t vitastor_proxy_get_size(void *watch);
-uint64_t vitastor_proxy_get_inode_num(void *watch);
-int vitastor_proxy_get_readonly(void *watch);
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif

src/test_cas.cpp

@@ -0,0 +1,135 @@
+// Copyright (c) Vitaliy Filippov, 2019+
+// License: VNPL-1.1 (see README.md for details)
+
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "epoll_manager.h"
+#include "cluster_client.h"
+
+void send_read(cluster_client_t *cli, uint64_t inode, std::function<void(int, uint64_t)> cb)
+{
+    cluster_op_t *op = new cluster_op_t();
+    op->opcode = OSD_OP_READ;
+    op->inode = inode;
+    op->offset = 0;
+    op->len = 4096;
+    op->iov.push_back(malloc_or_die(op->len), op->len);
+    op->callback = [cb](cluster_op_t *op)
+    {
+        uint64_t version = op->version;
+        int retval = op->retval;
+        if (retval == op->len)
+            retval = 0;
+        free(op->iov.buf[0].iov_base);
+        delete op;
+        if (cb != NULL)
+            cb(retval, version);
+    };
+    cli->execute(op);
+}
+
+void send_write(cluster_client_t *cli, uint64_t inode, int byte, uint64_t version, std::function<void(int)> cb)
+{
+    cluster_op_t *op = new cluster_op_t();
+    op->opcode = OSD_OP_WRITE;
+    op->inode = inode;
+    op->offset = 0;
+    op->len = 4096;
+    op->version = version;
+    op->iov.push_back(malloc_or_die(op->len), op->len);
+    memset(op->iov.buf[0].iov_base, byte, op->len);
+    op->callback = [cb](cluster_op_t *op)
+    {
+        int retval = op->retval;
+        if (retval == op->len)
+            retval = 0;
+        free(op->iov.buf[0].iov_base);
+        delete op;
+        if (cb != NULL)
+            cb(retval);
+    };
+    cli->execute(op);
+}
+
+int main(int narg, char *args[])
+{
+    json11::Json::object cfgo;
+    for (int i = 1; i < narg; i++)
+    {
+        if (args[i][0] == '-' && args[i][1] == '-')
+        {
+            const char *opt = args[i]+2;
+            cfgo[opt] = i == narg-1 ? "1" : args[++i];
+        }
+    }
+    json11::Json cfg(cfgo);
+    uint64_t inode = (cfg["pool_id"].uint64_value() << (64-POOL_ID_BITS))
+        | cfg["inode_id"].uint64_value();
+    uint64_t base_ver = 0;
+    // Create client
+    auto ringloop = new ring_loop_t(512);
+    auto epmgr = new epoll_manager_t(ringloop);
+    auto cli = new cluster_client_t(ringloop, epmgr->tfd, cfg);
+    cli->on_ready([&]()
+    {
+        send_read(cli, inode, [&](int r, uint64_t v)
+        {
+            if (r < 0)
+            {
+                fprintf(stderr, "Initial read operation failed\n");
+                exit(1);
+            }
+            base_ver = v;
+            // CAS v=1 = compare with zero, non-existing object
+            send_write(cli, inode, 0x01, base_ver+1, [&](int r)
+            {
+                if (r < 0)
+                {
+                    fprintf(stderr, "CAS for non-existing object failed\n");
+                    exit(1);
+                }
+                // Check that read returns the new version
+                send_read(cli, inode, [&](int r, uint64_t v)
+                {
+                    if (r < 0)
+                    {
+                        fprintf(stderr, "Read operation failed after write\n");
+                        exit(1);
+                    }
+                    if (v != base_ver+1)
+                    {
+                        fprintf(stderr, "Read operation failed to return the new version number\n");
+                        exit(1);
+                    }
+                    // CAS v=2 = compare with v=1, existing object
+                    send_write(cli, inode, 0x02, base_ver+2, [&](int r)
+                    {
+                        if (r < 0)
+                        {
+                            fprintf(stderr, "CAS for existing object failed\n");
+                            exit(1);
+                        }
+                        // CAS v=2 again = compare with v=1, but version is 2. Must fail with -EINTR
+                        send_write(cli, inode, 0x03, base_ver+2, [&](int r)
+                        {
+                            if (r != -EINTR)
+                            {
+                                fprintf(stderr, "CAS conflict detection failed\n");
+                                exit(1);
+                            }
+                            printf("Basic CAS test succeeded\n");
+                            exit(0);
+                        });
+                    });
+                });
+            });
+        });
+    });
+    while (1)
+    {
+        ringloop->loop();
+        ringloop->wait();
+    }
+    return 0;
+}

src/vitastor_c.cpp

@@ -0,0 +1,254 @@
+// Copyright (c) Vitaliy Filippov, 2019+
+// License: VNPL-1.1 or GNU GPL-2.0+ (see README.md for details)
+
+// Simplified C client library for QEMU, fio and other external drivers
+// Also acts as a C-C++ proxy for the QEMU driver (QEMU headers don't compile with g++)
+
+#include <sys/epoll.h>
+
+#include "ringloop.h"
+#include "epoll_manager.h"
+#include "cluster_client.h"
+
+#include "vitastor_c.h"
+
+struct vitastor_qemu_fd_t
+{
+    int fd;
+    std::function<void(int, int)> callback;
+};
+
+struct vitastor_c
+{
+    std::map<int, vitastor_qemu_fd_t> handlers;
+    ring_loop_t *ringloop = NULL;
+    epoll_manager_t *epmgr = NULL;
+    timerfd_manager_t *tfd = NULL;
+    cluster_client_t *cli = NULL;
+
+    QEMUSetFDHandler *aio_set_fd_handler = NULL;
+    void *aio_ctx = NULL;
+};
+
+extern "C" {
+
+static json11::Json vitastor_c_common_config(const char *config_path, const char *etcd_host, const char *etcd_prefix,
+    int use_rdma, const char *rdma_device, int rdma_port_num, int rdma_gid_index, int rdma_mtu, int log_level)
+{
+    json11::Json::object cfg;
+    if (config_path)
+        cfg["config_path"] = std::string(config_path);
+    if (etcd_host)
+        cfg["etcd_address"] = std::string(etcd_host);
+    if (etcd_prefix)
+        cfg["etcd_prefix"] = std::string(etcd_prefix);
+    // -1 means unspecified
+    if (use_rdma >= 0)
+        cfg["use_rdma"] = use_rdma > 0;
+    if (rdma_device)
+        cfg["rdma_device"] = std::string(rdma_device);
+    if (rdma_port_num)
+        cfg["rdma_port_num"] = rdma_port_num;
+    if (rdma_gid_index)
+        cfg["rdma_gid_index"] = rdma_gid_index;
+    if (rdma_mtu)
+        cfg["rdma_mtu"] = rdma_mtu;
+    if (log_level)
+        cfg["log_level"] = log_level;
+    return json11::Json(cfg);
+}
+
+static void vitastor_c_read_handler(void *opaque)
+{
+    vitastor_qemu_fd_t *data = (vitastor_qemu_fd_t *)opaque;
+    data->callback(data->fd, EPOLLIN);
+}
+
+static void vitastor_c_write_handler(void *opaque)
+{
+    vitastor_qemu_fd_t *data = (vitastor_qemu_fd_t *)opaque;
+    data->callback(data->fd, EPOLLOUT);
+}
+
+vitastor_c *vitastor_c_create_qemu(QEMUSetFDHandler *aio_set_fd_handler, void *aio_context,
+    const char *config_path, const char *etcd_host, const char *etcd_prefix,
+    bool use_rdma, const char *rdma_device, int rdma_port_num, int rdma_gid_index, int rdma_mtu, int log_level)
+{
+    json11::Json cfg_json = vitastor_c_common_config(
+        config_path, etcd_host, etcd_prefix, use_rdma,
+        rdma_device, rdma_port_num, rdma_gid_index, rdma_mtu, log_level
+    );
+    vitastor_c *self = new vitastor_c;
+    self->aio_set_fd_handler = aio_set_fd_handler;
+    self->aio_ctx = aio_context;
+    self->tfd = new timerfd_manager_t([self](int fd, bool wr, std::function<void(int, int)> callback)
+    {
+        if (callback != NULL)
+        {
+            self->handlers[fd] = { .fd = fd, .callback = callback };
+            self->aio_set_fd_handler(self->aio_ctx, fd, false,
+                vitastor_c_read_handler, wr ? vitastor_c_write_handler : NULL, NULL, &self->handlers[fd]);
+        }
+        else
+        {
+            self->handlers.erase(fd);
+            self->aio_set_fd_handler(self->aio_ctx, fd, false, NULL, NULL, NULL, NULL);
+        }
+    });
+    self->cli = new cluster_client_t(NULL, self->tfd, cfg_json);
+    return self;
+}
+
+vitastor_c *vitastor_c_create_uring(const char *config_path, const char *etcd_host, const char *etcd_prefix,
+    int use_rdma, const char *rdma_device, int rdma_port_num, int rdma_gid_index, int rdma_mtu, int log_level)
+{
+    json11::Json cfg_json = vitastor_c_common_config(
+        config_path, etcd_host, etcd_prefix, use_rdma,
+        rdma_device, rdma_port_num, rdma_gid_index, rdma_mtu, log_level
+    );
+    vitastor_c *self = new vitastor_c;
+    self->ringloop = new ring_loop_t(512);
+    self->epmgr = new epoll_manager_t(self->ringloop);
+    self->cli = new cluster_client_t(self->ringloop, self->epmgr->tfd, cfg_json);
+    return self;
+}
+
+vitastor_c *vitastor_c_create_uring_json(const char **options, int options_len)
+{
+    json11::Json::object cfg;
+    for (int i = 0; i < options_len-1; i += 2)
+    {
+        cfg[options[i]] = std::string(options[i+1]);
+    }
+    json11::Json cfg_json(cfg);
+    vitastor_c *self = new vitastor_c;
+    self->ringloop = new ring_loop_t(512);
+    self->epmgr = new epoll_manager_t(self->ringloop);
+    self->cli = new cluster_client_t(self->ringloop, self->epmgr->tfd, cfg_json);
+    return self;
+}
+
+void vitastor_c_destroy(vitastor_c *client)
+{
+    delete client->cli;
+    if (client->epmgr)
+        delete client->epmgr;
+    else
+        delete client->tfd;
+    if (client->ringloop)
+        delete client->ringloop;
+    delete client;
+}
+
+int vitastor_c_is_ready(vitastor_c *client)
+{
+    return client->cli->is_ready();
+}
+
+void vitastor_c_uring_wait_ready(vitastor_c *client)
+{
+    while (!client->cli->is_ready())
+    {
+        client->ringloop->loop();
+        if (client->cli->is_ready())
+            break;
+        client->ringloop->wait();
+    }
+}
+
+void vitastor_c_uring_handle_events(vitastor_c *client)
+{
+    client->ringloop->loop();
+}
+
+void vitastor_c_uring_wait_events(vitastor_c *client)
+{
+    client->ringloop->wait();
+}
+
+void vitastor_c_read(vitastor_c *client, uint64_t inode, uint64_t offset, uint64_t len,
+    struct iovec *iov, int iovcnt, VitastorReadHandler cb, void *opaque)
+{
+    cluster_op_t *op = new cluster_op_t;
+    op->opcode = OSD_OP_READ;
+    op->inode = inode;
+    op->offset = offset;
+    op->len = len;
+    for (int i = 0; i < iovcnt; i++)
+    {
+        op->iov.push_back(iov[i].iov_base, iov[i].iov_len);
+    }
+    op->callback = [cb, opaque](cluster_op_t *op)
+    {
+        cb(opaque, op->retval, op->version);
+        delete op;
+    };
+    client->cli->execute(op);
+}
+
+void vitastor_c_write(vitastor_c *client, uint64_t inode, uint64_t offset, uint64_t len, uint64_t check_version,
+    struct iovec *iov, int iovcnt, VitastorIOHandler cb, void *opaque)
+{
+    cluster_op_t *op = new cluster_op_t;
+    op->opcode = OSD_OP_WRITE;
+    op->inode = inode;
+    op->offset = offset;
+    op->len = len;
+    op->version = check_version;
+    for (int i = 0; i < iovcnt; i++)
+    {
+        op->iov.push_back(iov[i].iov_base, iov[i].iov_len);
+    }
+    op->callback = [cb, opaque](cluster_op_t *op)
+    {
+        cb(opaque, op->retval);
+        delete op;
+    };
+    client->cli->execute(op);
+}
+
+void vitastor_c_sync(vitastor_c *client, VitastorIOHandler cb, void *opaque)
+{
+    cluster_op_t *op = new cluster_op_t;
+    op->opcode = OSD_OP_SYNC;
+    op->callback = [cb, opaque](cluster_op_t *op)
+    {
+        cb(opaque, op->retval);
+        delete op;
+    };
+    client->cli->execute(op);
+}
+
+void vitastor_c_watch_inode(vitastor_c *client, char *image, VitastorIOHandler cb, void *opaque)
+{
+    client->cli->on_ready([=]()
+    {
+        auto watch = client->cli->st_cli.watch_inode(std::string(image));
+        cb(opaque, (long)watch);
+    });
+}
+
+void vitastor_c_close_watch(vitastor_c *client, void *handle)
+{
+    client->cli->st_cli.close_watch((inode_watch_t*)handle);
+}
+
+uint64_t vitastor_c_inode_get_size(void *handle)
+{
+    inode_watch_t *watch = (inode_watch_t*)handle;
+    return watch->cfg.size;
+}
+
+uint64_t vitastor_c_inode_get_num(void *handle)
+{
+    inode_watch_t *watch = (inode_watch_t*)handle;
+    return watch->cfg.num;
+}
+
+int vitastor_c_inode_get_readonly(void *handle)
+{
+    inode_watch_t *watch = (inode_watch_t*)handle;
+    return watch->cfg.readonly;
+}
+
+}

src/vitastor_c.h

@@ -0,0 +1,55 @@
+// Copyright (c) Vitaliy Filippov, 2019+
+// License: VNPL-1.1 or GNU GPL-2.0+ (see README.md for details)
+
+// Simplified C client library for QEMU, fio and other external drivers
+
+#ifndef VITASTOR_QEMU_PROXY_H
+#define VITASTOR_QEMU_PROXY_H
+
+#ifndef POOL_ID_BITS
+#define POOL_ID_BITS 16
+#endif
+#include <stdint.h>
+#include <sys/uio.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct vitastor_c;
+typedef struct vitastor_c vitastor_c;
+
+typedef void VitastorReadHandler(void *opaque, long retval, uint64_t version);
+typedef void VitastorIOHandler(void *opaque, long retval);
+
+// QEMU
+typedef void IOHandler(void *opaque);
+typedef void QEMUSetFDHandler(void *ctx, int fd, int is_external, IOHandler *fd_read, IOHandler *fd_write, void *poll_fn, void *opaque);
+
+vitastor_c *vitastor_c_create_qemu(QEMUSetFDHandler *aio_set_fd_handler, void *aio_context,
+    const char *config_path, const char *etcd_host, const char *etcd_prefix,
+    bool use_rdma, const char *rdma_device, int rdma_port_num, int rdma_gid_index, int rdma_mtu, int log_level);
+vitastor_c *vitastor_c_create_uring(const char *config_path, const char *etcd_host, const char *etcd_prefix,
+    int use_rdma, const char *rdma_device, int rdma_port_num, int rdma_gid_index, int rdma_mtu, int log_level);
+vitastor_c *vitastor_c_create_uring_json(const char **options, int options_len);
+void vitastor_c_destroy(vitastor_c *client);
+int vitastor_c_is_ready(vitastor_c *client);
+void vitastor_c_uring_wait_ready(vitastor_c *client);
+void vitastor_c_uring_handle_events(vitastor_c *client);
+void vitastor_c_uring_wait_events(vitastor_c *client);
+void vitastor_c_read(vitastor_c *client, uint64_t inode, uint64_t offset, uint64_t len,
+    struct iovec *iov, int iovcnt, VitastorReadHandler cb, void *opaque);
+void vitastor_c_write(vitastor_c *client, uint64_t inode, uint64_t offset, uint64_t len, uint64_t check_version,
+    struct iovec *iov, int iovcnt, VitastorIOHandler cb, void *opaque);
+void vitastor_c_sync(vitastor_c *client, VitastorIOHandler cb, void *opaque);
+void vitastor_c_watch_inode(vitastor_c *client, char *image, VitastorIOHandler cb, void *opaque);
+void vitastor_c_close_watch(vitastor_c *client, void *handle);
+uint64_t vitastor_c_inode_get_size(void *handle);
+uint64_t vitastor_c_inode_get_num(void *handle);
+int vitastor_c_inode_get_readonly(void *handle);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

tests/run_3osds.sh

@@ -0,0 +1,43 @@
+#!/bin/bash -ex
+
+. `dirname $0`/common.sh
+
+OSD_SIZE=${OSD_SIZE:-1024}
+
+dd if=/dev/zero of=./testdata/test_osd1.bin bs=1024 count=1 seek=$((OSD_SIZE*1024-1))
+dd if=/dev/zero of=./testdata/test_osd2.bin bs=1024 count=1 seek=$((OSD_SIZE*1024-1))
+dd if=/dev/zero of=./testdata/test_osd3.bin bs=1024 count=1 seek=$((OSD_SIZE*1024-1))
+
+build/src/vitastor-osd --osd_num 1 --bind_address 127.0.0.1 $OSD_ARGS --etcd_address $ETCD_URL $(node mon/simple-offsets.js --format options --device ./testdata/test_osd1.bin 2>/dev/null) &>./testdata/osd1.log &
+OSD1_PID=$!
+build/src/vitastor-osd --osd_num 2 --bind_address 127.0.0.1 $OSD_ARGS --etcd_address $ETCD_URL $(node mon/simple-offsets.js --format options --device ./testdata/test_osd2.bin 2>/dev/null) &>./testdata/osd2.log &
+OSD2_PID=$!
+build/src/vitastor-osd --osd_num 3 --bind_address 127.0.0.1 $OSD_ARGS --etcd_address $ETCD_URL $(node mon/simple-offsets.js --format options --device ./testdata/test_osd3.bin 2>/dev/null) &>./testdata/osd3.log &
+OSD3_PID=$!
+
+cd mon
+npm install
+cd ..
+node mon/mon-main.js --etcd_url http://$ETCD_URL --etcd_prefix "/vitastor" &>./testdata/mon.log &
+MON_PID=$!
+
+if [ -n "$GLOBAL_CONF" ]; then
+    $ETCDCTL put /vitastor/config/global "$GLOBAL_CONF"
+fi
+
+$ETCDCTL put /vitastor/config/pools '{"1":{"name":"testpool","scheme":"xor","pg_size":3,"pg_minsize":2,"parity_chunks":1,"pg_count":1,"failure_domain":"osd"}}'
+
+sleep 2
+
+if ! ($ETCDCTL get /vitastor/config/pgs --print-value-only | jq -s -e '(. | length) != 0 and (.[0].items["1"]["1"].osd_set | sort) == ["1","2","3"]'); then
+    format_error "FAILED: 1 PG NOT CONFIGURED"
+fi
+
+if ! ($ETCDCTL get /vitastor/pg/state/1/1 --print-value-only | jq -s -e '(. | length) != 0 and .[0].state == ["active"]'); then
+    format_error "FAILED: 1 PG NOT UP"
+fi
+
+if ! cmp build/src/block-vitastor.so /usr/lib/x86_64-linux-gnu/qemu/block-vitastor.so; then
+    sudo rm -f /usr/lib/x86_64-linux-gnu/qemu/block-vitastor.so
+    sudo ln -s "$(realpath .)/build/src/block-vitastor.so" /usr/lib/x86_64-linux-gnu/qemu/block-vitastor.so
+fi

tests/test_cas.sh

@@ -0,0 +1,7 @@
+#!/bin/bash -ex
+
+. `dirname $0`/run_3osds.sh
+
+build/src/test_cas --pool_id 1 --inode_id 1 --etcd_address $ETCD_URL
+
+format_green OK

tests/test_snapshot.sh

@@ -1,40 +1,6 @@
 #!/bin/bash -ex
 
-. `dirname $0`/common.sh
-
-dd if=/dev/zero of=./testdata/test_osd1.bin bs=1024 count=1 seek=$((1024*1024-1))
-dd if=/dev/zero of=./testdata/test_osd2.bin bs=1024 count=1 seek=$((1024*1024-1))
-dd if=/dev/zero of=./testdata/test_osd3.bin bs=1024 count=1 seek=$((1024*1024-1))
-
-build/src/vitastor-osd --osd_num 1 --bind_address 127.0.0.1 --etcd_address $ETCD_URL $(node mon/simple-offsets.js --format options --device ./testdata/test_osd1.bin 2>/dev/null) &>./testdata/osd1.log &
-OSD1_PID=$!
-build/src/vitastor-osd --osd_num 2 --bind_address 127.0.0.1 --etcd_address $ETCD_URL $(node mon/simple-offsets.js --format options --device ./testdata/test_osd2.bin 2>/dev/null) &>./testdata/osd2.log &
-OSD2_PID=$!
-build/src/vitastor-osd --osd_num 3 --bind_address 127.0.0.1 --etcd_address $ETCD_URL $(node mon/simple-offsets.js --format options --device ./testdata/test_osd3.bin 2>/dev/null) &>./testdata/osd3.log &
-OSD3_PID=$!
-
-cd mon
-npm install
-cd ..
-node mon/mon-main.js --etcd_url http://$ETCD_URL --etcd_prefix "/vitastor" &>./testdata/mon.log &
-MON_PID=$!
-
-$ETCDCTL put /vitastor/config/pools '{"1":{"name":"testpool","scheme":"xor","pg_size":3,"pg_minsize":2,"parity_chunks":1,"pg_count":1,"failure_domain":"osd"}}'
-
-sleep 2
-
-if ! ($ETCDCTL get /vitastor/config/pgs --print-value-only | jq -s -e '(. | length) != 0 and (.[0].items["1"]["1"].osd_set | sort) == ["1","2","3"]'); then
-    format_error "FAILED: 1 PG NOT CONFIGURED"
-fi
-
-if ! ($ETCDCTL get /vitastor/pg/state/1/1 --print-value-only | jq -s -e '(. | length) != 0 and .[0].state == ["active"]'); then
-    format_error "FAILED: 1 PG NOT UP"
-fi
-
-if ! cmp build/src/block-vitastor.so /usr/lib/x86_64-linux-gnu/qemu/block-vitastor.so; then
-    sudo rm -f /usr/lib/x86_64-linux-gnu/qemu/block-vitastor.so
-    sudo ln -s "$(realpath .)/build/src/block-vitastor.so" /usr/lib/x86_64-linux-gnu/qemu/block-vitastor.so
-fi
+. `dirname $0`/run_3osds.sh
 
 # Test basic write and snapshot
 

tests/test_vm_start.sh

@@ -1,40 +1,8 @@
 #!/bin/bash -ex
 
-. `dirname $0`/common.sh
+OSD_SIZE=2048
 
-dd if=/dev/zero of=./testdata/test_osd1.bin bs=2048 count=1 seek=$((1024*1024-1))
-dd if=/dev/zero of=./testdata/test_osd2.bin bs=2048 count=1 seek=$((1024*1024-1))
-dd if=/dev/zero of=./testdata/test_osd3.bin bs=2048 count=1 seek=$((1024*1024-1))
-
-build/src/vitastor-osd --osd_num 1 --bind_address 127.0.0.1 --etcd_address $ETCD_URL $(node mon/simple-offsets.js --format options --device ./testdata/test_osd1.bin 2>/dev/null) &>./testdata/osd1.log &
-OSD1_PID=$!
-build/src/vitastor-osd --osd_num 2 --bind_address 127.0.0.1 --etcd_address $ETCD_URL $(node mon/simple-offsets.js --format options --device ./testdata/test_osd2.bin 2>/dev/null) &>./testdata/osd2.log &
-OSD2_PID=$!
-build/src/vitastor-osd --osd_num 3 --bind_address 127.0.0.1 --etcd_address $ETCD_URL $(node mon/simple-offsets.js --format options --device ./testdata/test_osd3.bin 2>/dev/null) &>./testdata/osd3.log &
-OSD3_PID=$!
-
-cd mon
-npm install
-cd ..
-node mon/mon-main.js --etcd_url http://$ETCD_URL --etcd_prefix "/vitastor" &>./testdata/mon.log &
-MON_PID=$!
-
-$ETCDCTL put /vitastor/config/pools '{"1":{"name":"testpool","scheme":"xor","pg_size":3,"pg_minsize":2,"parity_chunks":1,"pg_count":1,"failure_domain":"osd"}}'
-
-sleep 2
-
-if ! ($ETCDCTL get /vitastor/config/pgs --print-value-only | jq -s -e '(. | length) != 0 and (.[0].items["1"]["1"].osd_set | sort) == ["1","2","3"]'); then
-    format_error "FAILED: 1 PG NOT CONFIGURED"
-fi
-
-if ! ($ETCDCTL get /vitastor/pg/state/1/1 --print-value-only | jq -s -e '(. | length) != 0 and .[0].state == ["active"]'); then
-    format_error "FAILED: 1 PG NOT UP"
-fi
-
-if ! cmp build/src/block-vitastor.so /usr/lib/x86_64-linux-gnu/qemu/block-vitastor.so; then
-    sudo rm -f /usr/lib/x86_64-linux-gnu/qemu/block-vitastor.so
-    sudo ln -s "$(realpath .)/build/src/block-vitastor.so" /usr/lib/x86_64-linux-gnu/qemu/block-vitastor.so
-fi
+. `dirname $0`/run_3osds.sh
 
 $ETCDCTL put /vitastor/config/inode/1/1 '{"name":"debian9","size":'$((2048*1024*1024))'}'
 

tests/test_write.sh

@@ -1,44 +1,10 @@
 #!/bin/bash -ex
 
-. `dirname $0`/common.sh
-
-dd if=/dev/zero of=./testdata/test_osd1.bin bs=1024 count=1 seek=$((1024*1024-1))
-dd if=/dev/zero of=./testdata/test_osd2.bin bs=1024 count=1 seek=$((1024*1024-1))
-dd if=/dev/zero of=./testdata/test_osd3.bin bs=1024 count=1 seek=$((1024*1024-1))
-
-build/src/vitastor-osd --osd_num 1 --bind_address 127.0.0.1 --etcd_address $ETCD_URL $(node mon/simple-offsets.js --format options --device ./testdata/test_osd1.bin 2>/dev/null) &>./testdata/osd1.log &
-OSD1_PID=$!
-build/src/vitastor-osd --osd_num 2 --bind_address 127.0.0.1 --etcd_address $ETCD_URL $(node mon/simple-offsets.js --format options --device ./testdata/test_osd2.bin 2>/dev/null) &>./testdata/osd2.log &
-OSD2_PID=$!
-build/src/vitastor-osd --osd_num 3 --bind_address 127.0.0.1 --etcd_address $ETCD_URL $(node mon/simple-offsets.js --format options --device ./testdata/test_osd3.bin 2>/dev/null) &>./testdata/osd3.log &
-OSD3_PID=$!
-
-cd mon
-npm install
-cd ..
-node mon/mon-main.js --etcd_url http://$ETCD_URL --etcd_prefix "/vitastor" &>./testdata/mon.log &
-MON_PID=$!
-
-$ETCDCTL put /vitastor/config/pools '{"1":{"name":"testpool","scheme":"xor","pg_size":3,"pg_minsize":2,"parity_chunks":1,"pg_count":1,"failure_domain":"osd"}}'
-
-sleep 2
-
-if ! ($ETCDCTL get /vitastor/config/pgs --print-value-only | jq -s -e '(. | length) != 0 and (.[0].items["1"]["1"].osd_set | sort) == ["1","2","3"]'); then
-    format_error "FAILED: 1 PG NOT CONFIGURED"
-fi
-
-if ! ($ETCDCTL get /vitastor/pg/state/1/1 --print-value-only | jq -s -e '(. | length) != 0 and .[0].state == ["active"]'); then
-    format_error "FAILED: 1 PG NOT UP"
-fi
+. `dirname $0`/run_3osds.sh
 
 #LD_PRELOAD=libasan.so.5 \
 #    fio -thread -name=test -ioengine=build/src/libfio_vitastor_sec.so -bs=4k -fsync=128 `$ETCDCTL get /vitastor/osd/state/1 --print-value-only | jq -r '"-host="+.addresses[0]+" -port="+(.port|tostring)'` -rw=write -size=32M
 
-if ! cmp build/src/block-vitastor.so /usr/lib/x86_64-linux-gnu/qemu/block-vitastor.so; then
-    sudo rm -f /usr/lib/x86_64-linux-gnu/qemu/block-vitastor.so
-    sudo ln -s "$(realpath .)/build/src/block-vitastor.so" /usr/lib/x86_64-linux-gnu/qemu/block-vitastor.so
-fi
-
 # A lot of parallel syncs was crashing the primary OSD at some point
 
 LD_PRELOAD=libasan.so.5 \

tests/test_write_no_same.sh

@@ -1,40 +1,10 @@
 #!/bin/bash -ex
 # Test the `no_same_sector_overwrites` mode
 
-. `dirname $0`/common.sh
+OSD_ARGS="--journal_no_same_sector_overwrites true --journal_sector_buffer_count 1024 --disable_data_fsync 1 --immediate_commit all"
+GLOBAL_CONF='{"immediate_commit":"all"}'
 
-dd if=/dev/zero of=./testdata/test_osd1.bin bs=1024 count=1 seek=$((1024*1024-1))
-dd if=/dev/zero of=./testdata/test_osd2.bin bs=1024 count=1 seek=$((1024*1024-1))
-dd if=/dev/zero of=./testdata/test_osd3.bin bs=1024 count=1 seek=$((1024*1024-1))
-
-NO_SAME="--journal_no_same_sector_overwrites true --journal_sector_buffer_count 1024 --disable_data_fsync 1 --immediate_commit all"
-
-build/src/vitastor-osd --osd_num 1 --bind_address 127.0.0.1 $NO_SAME --etcd_address $ETCD_URL $(node mon/simple-offsets.js --format options --device ./testdata/test_osd1.bin 2>/dev/null) &>./testdata/osd1.log &
-OSD1_PID=$!
-build/src/vitastor-osd --osd_num 2 --bind_address 127.0.0.1 $NO_SAME --etcd_address $ETCD_URL $(node mon/simple-offsets.js --format options --device ./testdata/test_osd2.bin 2>/dev/null) &>./testdata/osd2.log &
-OSD2_PID=$!
-build/src/vitastor-osd --osd_num 3 --bind_address 127.0.0.1 $NO_SAME --etcd_address $ETCD_URL $(node mon/simple-offsets.js --format options --device ./testdata/test_osd3.bin 2>/dev/null) &>./testdata/osd3.log &
-OSD3_PID=$!
-
-cd mon
-npm install
-cd ..
-node mon/mon-main.js --etcd_url http://$ETCD_URL --etcd_prefix "/vitastor" &>./testdata/mon.log &
-MON_PID=$!
-
-$ETCDCTL put /vitastor/config/global '{"immediate_commit":"all"}'
-
-$ETCDCTL put /vitastor/config/pools '{"1":{"name":"testpool","scheme":"xor","pg_size":3,"pg_minsize":2,"parity_chunks":1,"pg_count":1,"failure_domain":"osd"}}'
-
-sleep 2
-
-if ! ($ETCDCTL get /vitastor/config/pgs --print-value-only | jq -s -e '(. | length) != 0 and (.[0].items["1"]["1"].osd_set | sort) == ["1","2","3"]'); then
-    format_error "FAILED: 1 PG NOT CONFIGURED"
-fi
-
-if ! ($ETCDCTL get /vitastor/pg/state/1/1 --print-value-only | jq -s -e '(. | length) != 0 and .[0].state == ["active"]'); then
-    format_error "FAILED: 1 PG NOT UP"
-fi
+. `dirname $0`/run_3osds.sh
 
 #LSAN_OPTIONS=report_objects=true:suppressions=`pwd`/testdata/lsan-suppress.txt LD_PRELOAD=libasan.so.5 \
 #    fio -thread -name=test -ioengine=build/src/libfio_vitastor_sec.so -bs=4k -fsync=128 `$ETCDCTL get /vitastor/osd/state/1 --print-value-only | jq -r '"-host="+.addresses[0]+" -port="+(.port|tostring)'` -rw=write -size=32M