Skip to contentSkip to conversation

hw/nvme: Support for Namespaces Management from guest OS

PATCH Superseded Jonathan Derrick 4 participants 6 messages Dec 28, 2022 – Jan 9, 2023 Arm/Amlogic Meson SoC support

v4 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

v4 cover hw/nvme: Support for Namespaces Management from guest OS

From: Michael Kropaczek <[email protected]>

Description:

Currently namespaces could be configured as follows:
1. Legacy Namespace - just one namespace within Nvme controller's where the back-end was specified for nvme device by -drive parameter pointing directly to the image file.
2. Additional Namespaces - specified by nvme-ns devices each having its own back-end. To have multiple namespaces each needed to be specified at Qemu's command line being associated with the most recently defined nvme-bus from nvme device.
If a such additional namespace should be attached and/or detached by the guest OS, nvme controller has to be linked with another device nvme-subsys.

All that have a static nature, all need to be specified at the Qemu's command line, all specified virtual nvme entities will be processed during Qemu's start-up then created and provided to the guest OS.

To have a support for nvme create-ns and delete-ns commands with specified parameters a different approach is needed.
Virtual devices representing namespaces need to be created and/or deleted during Qemu's running session, at anytime. The back-end image sizes for a namespace must accommodate the payload size and size of metadata resulted from specified parameters. The total capacity of the nvme controller altogether with un-allocated capacity needs to be taken into account and updated according to nvme create-ns and delete-ns commands respectively.

Here is the approach:
The nvme device will get new parameter:
- auto-ns-path, which specifies the path to the storage area where back-end image and necessary config files located stored.

The virtual devices representing namespaces will be created dynamically during the Qemu running session following issuance of nvme create-ns and delete-ns commands from the guest OS. QOM classes and instances will be created utilizing existing configuration scheme used during Qemu's start-up. Back-end image files will be neither created nor deleted during Qemu's startup or running session. Instead a set of back-end image files and relevant config will be created by qemu-img tool with createns sub-command prior to Qemu's session. Required parameters are: -S serial number which must match serial parameter of qemu-system-xx -device nvme command line specification, -C total capacity, and optional -N that will set a maximal limit on number of allowed namespaces (default 256) which will be followed by path name pointing to storage location corresponding to auto-ns-path of qemu-system-xx -device nvme parameter.

Those created back-end image files will be pre-loaded during Qemu's start-up and then during running Qemu's session will be associated or disassociated with QOM namespaces virtual instances, as a result of issuing nvme create-ns or delete-ns commands. The associated back-end image file for relevant namespace will be re-sized as follows: delete-ns command will truncate image file to the size of 0, whereas create-ns command will re-size the image file to the size provided by nvme create-ns command parameters. Truncating/re-sizing is a result of blk_truncate() API which utilizes co-routines and should not block Qemu main thread while scheduling AIO operations. It is assured that all settings will retain over Qemu's start-ups and shutdowns. The implementation makes it possible to combine the existing "Additional Namespace" implementation with the new "Managed Namespaces". Those will coexist with obvious restrictions, like both will share the same NsIds space, "static" namespaces cannot be deleted or if its NsId specified at Qemu's command line will conflicts with previously created one by nvme create-ns (and retained), this will lead to an abort of Qemu at its start up.

More than one of NVMe controllers associated with NVMe subsystem are supported. This feature requires that parameters serial= and subsys= of additional controllers must match those of the primary controller and auto-ns-path= must not be specified.

Michael Kropaczek (2):
hw/nvme: Support for Namespaces Management from guest OS - create-ns hw/nvme: Support for Namespaces Management from guest OS - delete-ns

 docs/system/devices/nvme.rst |  60 +++++-
 hw/nvme/cfg_key_checker.c    |  51 +++++
 hw/nvme/ctrl-cfg.c           | 224 +++++++++++++++++++++
 hw/nvme/ctrl.c               | 313 +++++++++++++++++++++++++++++-
 hw/nvme/meson.build          |   2 +-
 hw/nvme/ns-backend.c         | 288 +++++++++++++++++++++++++++
 hw/nvme/ns.c                 | 365 +++++++++++++++++++++++++++++++----
 hw/nvme/nvme.h               |  32 ++-
 hw/nvme/subsys.c             |  11 +-
 hw/nvme/trace-events         |   3 +
 include/block/nvme.h         |  31 +++
 include/hw/nvme/ctrl-cfg.h   |  24 +++
 include/hw/nvme/ns-cfg.h     |  28 +++
 include/hw/nvme/nvme-cfg.h   | 188 ++++++++++++++++++
 qemu-img-cmds.hx             |   6 +
 qemu-img.c                   | 132 +++++++++++++
 16 files changed, 1704 insertions(+), 54 deletions(-)
 create mode 100644 hw/nvme/cfg_key_checker.c
 create mode 100644 hw/nvme/ctrl-cfg.c
 create mode 100644 hw/nvme/ns-backend.c
 create mode 100644 include/hw/nvme/ctrl-cfg.h
 create mode 100644 include/hw/nvme/ns-cfg.h
 create mode 100644 include/hw/nvme/nvme-cfg.h
Signature · 1 line
2.37.3
#1 lore raw

PATCH v4 1/2 hw/nvme: Support for Namespaces Management from guest OS - create-ns

From: Michael Kropaczek <[email protected]>

Added support for NVMEe NameSpaces Mangement allowing the guest OS to create namespaces by issuing nvme create-ns command.
It is an extension to currently implemented Qemu nvme virtual device. Virtual devices representing namespaces will be created and/or deleted during Qemu's running session, at anytime.

---
 docs/system/devices/nvme.rst |  59 ++++++-
 hw/nvme/cfg_key_checker.c    |  51 ++++++
 hw/nvme/ctrl-cfg.c           | 224 ++++++++++++++++++++++++++
 hw/nvme/ctrl.c               | 244 ++++++++++++++++++++++++++++-
 hw/nvme/meson.build          |   2 +-
 hw/nvme/ns-backend.c         | 283 +++++++++++++++++++++++++++++++++
 hw/nvme/ns.c                 | 293 ++++++++++++++++++++++++++++++-----
 hw/nvme/nvme.h               |  31 +++-
 hw/nvme/subsys.c             |  11 +-
 hw/nvme/trace-events         |   2 +
 include/block/nvme.h         |  30 ++++
 include/hw/nvme/ctrl-cfg.h   |  24 +++
 include/hw/nvme/ns-cfg.h     |  28 ++++
 include/hw/nvme/nvme-cfg.h   | 188 ++++++++++++++++++++++
 qemu-img-cmds.hx             |   6 +
 qemu-img.c                   | 132 ++++++++++++++++
 16 files changed, 1554 insertions(+), 54 deletions(-)
 create mode 100644 hw/nvme/cfg_key_checker.c
 create mode 100644 hw/nvme/ctrl-cfg.c
 create mode 100644 hw/nvme/ns-backend.c
 create mode 100644 include/hw/nvme/ctrl-cfg.h
 create mode 100644 include/hw/nvme/ns-cfg.h
 create mode 100644 include/hw/nvme/nvme-cfg.h
docs/system/devices/nvme.rst +58 -1
index 30f841ef62..6b3bee5e5d 100644
@@ -92,6 +92,63 @@ There are a number of parameters available:
   attach the namespace to a specific ``nvme`` device (identified by an ``id``
   parameter on the controller device).
 
+Additional Namespaces managed by guest OS Namespaces Management
+---------------------------------------------------------------------
+
+.. code-block:: console
+
+   -device nvme,id=nvme-ctrl,serial=1234,subsys=nvme-subsys,auto-ns-path=path
+
+Parameters:
+
+``auto-ns-path=<path to nvme storage location>``
+  If specified indicates a support for dynamic management of nvme namespaces
+  by means of nvme create-ns command. This path points
+  to the storage area for backend images must exist. Additionally it requires
+  that parameter `ns-subsys` must be specified whereas parameter `drive`
+  must not. The legacy namespace backend is disabled, instead, a pair of
+  files 'nvme_<ctrl SN>_ns_<NS-ID>.cfg' and 'nvme_<ctrl SN>_ns_<NS-ID>.img'
+  will refer to respective namespaces. The create-ns, attach-ns
+  and detach-ns commands, issued at the guest side, will make changes to
+  those files accordingly.
+  For each namespace exists an image file in raw format and a config file
+  containing namespace parameters and state of the attachment allowing QEMU
+  to configure namespaces accordingly during start up. If for instance an
+  image file has a size of 0 bytes this will be interpreted as non existent
+  namespace. Issuing create-ns command will change the status in the config
+  files and and will re-size the image file accordingly so the image file
+  will be associated with the respective namespace. The main config file
+  nvme_<ctrl SN>_ctrl.cfg keeps the track of allocated capacity to the
+  namespaces within the nvme controller.
+  As it is the case of a typical hard drive, backend images together with
+  config files need to be created. For this reason the qemu-img tool has
+  been extended by adding createns command.
+
+   qemu-img createns {-S <serial> -C <total NVMe capacity>}
+                     [-N <NsId max>] {<path>}
+
+  Parameters:
+  -S and -C and <path> are mandatory, `-S` must match `serial` parameter
+  and <path> must match `auto-ns-path` parameter of "-device nvme,..."
+  specification.
+  -N is optional, if specified it will set a limit to the number of potential
+  namespaces and will reduce the number of backend images and config files
+  accordingly. As a default, a set of images of 0 bytes size and default
+  config files for 256 namespaces will be created, a total of 513 files.
+
+Please note that ``nvme-ns`` device is not required to support of dynamic
+namespaces management feature. It is not prohibited to assign a such device to
+``nvme`` device specified to support dynamic namespace management if one has
+an use case to do so, however, it will only coexist and be out of the scope of
+Namespaces Management. NsIds will be consistently managed, creation (create-ns)
+of a namespace will not allocate the NsId already being taken. If ``nvme-ns``
+device conflicts with previously created one by create-ns (the same NsId),
+it will break QEMU's start up.
+More than one of NVMe controllers associated with NVMe subsystem are supported.
+This feature requires that parameters ``serial=`` and ``subsys=`` of additional
+controllers must match those of the primary controller and ``auto-ns-path=``
+must not be specified.
+
 NVM Subsystems
 --------------
 
@@ -320,4 +377,4 @@ controller are:
 
 .. code-block:: console
 
-   echo 0000:01:00.1 > /sys/bus/pci/drivers/nvme/bind
\ No newline at end of file
+   echo 0000:01:00.1 > /sys/bus/pci/drivers/nvme/bind
hw/nvme/cfg_key_checker.c new file +51 -0

51 lines not on this page · raw message

hw/nvme/ctrl-cfg.c new file +224 -0

224 lines not on this page · raw message

hw/nvme/ctrl.c +237 -7

350 lines not on this page · raw message

hw/nvme/meson.build +1 -1
index 3cf40046ee..8900831701 100644
@@ -1 +1 @@
-softmmu_ss.add(when: 'CONFIG_NVME_PCI', if_true: files('ctrl.c', 'dif.c', 'ns.c', 'subsys.c'))
+softmmu_ss.add(when: 'CONFIG_NVME_PCI', if_true: files('ctrl.c', 'dif.c', 'ns.c', 'subsys.c', 'ns-backend.c', 'cfg_key_checker.c', 'ctrl-cfg.c'))
hw/nvme/ns-backend.c new file +283 -0

283 lines not on this page · raw message

hw/nvme/ns.c +255 -38

379 lines not on this page · raw message

hw/nvme/nvme.h +28 -3
index 7adf042ec3..15b790cbdd 100644
@@ -23,9 +23,8 @@
 #include "hw/block/block.h"
 
 #include "block/nvme.h"
+#include "hw/nvme/ctrl-cfg.h"
 
-#define NVME_MAX_CONTROLLERS 256
-#define NVME_MAX_NAMESPACES  256
 #define NVME_EUI64_DEFAULT ((uint64_t)0x5254000000000000)
 
 QEMU_BUILD_BUG_ON(NVME_MAX_NAMESPACES > NVME_NSID_BROADCAST - 1);
@@ -279,6 +278,8 @@ int nvme_ns_setup(NvmeNamespace *ns, Error **errp);
 void nvme_ns_drain(NvmeNamespace *ns);
 void nvme_ns_shutdown(NvmeNamespace *ns);
 void nvme_ns_cleanup(NvmeNamespace *ns);
+void nvme_validate_flbas(uint8_t flbas,  Error **errp);
+NvmeNamespace * nvme_ns_create(NvmeCtrl *n, uint32_t nsid, NvmeIdNsMgmt *id_ns, Error **errp);
 
 typedef struct NvmeAsyncEvent {
     QTAILQ_ENTRY(NvmeAsyncEvent) entry;
@@ -339,6 +340,7 @@ static inline const char *nvme_adm_opc_str(uint8_t opc)
     case NVME_ADM_CMD_SET_FEATURES:     return "NVME_ADM_CMD_SET_FEATURES";
     case NVME_ADM_CMD_GET_FEATURES:     return "NVME_ADM_CMD_GET_FEATURES";
     case NVME_ADM_CMD_ASYNC_EV_REQ:     return "NVME_ADM_CMD_ASYNC_EV_REQ";
+    case NVME_ADM_CMD_NS_MGMT:          return "NVME_ADM_CMD_NS_MGMT";
     case NVME_ADM_CMD_NS_ATTACHMENT:    return "NVME_ADM_CMD_NS_ATTACHMENT";
     case NVME_ADM_CMD_VIRT_MNGMT:       return "NVME_ADM_CMD_VIRT_MNGMT";
     case NVME_ADM_CMD_DBBUF_CONFIG:     return "NVME_ADM_CMD_DBBUF_CONFIG";
@@ -427,6 +429,7 @@ typedef struct NvmeParams {
     uint16_t sriov_vi_flexible;
     uint8_t  sriov_max_vq_per_vf;
     uint8_t  sriov_max_vi_per_vf;
+    char     *ns_directory;     /* if empty (default) one legacy ns will be created */
 } NvmeParams;
 
 typedef struct NvmeCtrl {
@@ -485,8 +488,9 @@ typedef struct NvmeCtrl {
 
     NvmeSubsystem   *subsys;
 
-    NvmeNamespace   namespace;
+    NvmeNamespace   namespace;                  /* if ns_directory is empty this will be used */
     NvmeNamespace   *namespaces[NVME_MAX_NAMESPACES + 1];
+    BlockBackend    *preloaded_blk[NVME_MAX_NAMESPACES + 1];
     NvmeSQueue      **sq;
     NvmeCQueue      **cq;
     NvmeSQueue      admin_sq;
@@ -509,6 +513,7 @@ typedef struct NvmeCtrl {
         uint16_t    vqrfap;
         uint16_t    virfap;
     } next_pri_ctrl_cap;    /* These override pri_ctrl_cap after reset */
+    uint16_t nsidMax;
 } NvmeCtrl;
 
 typedef enum NvmeResetType {
@@ -575,6 +580,9 @@ static inline NvmeSecCtrlEntry *nvme_sctrl_for_cntlid(NvmeCtrl *n,
     return NULL;
 }
 
+BlockBackend *ns_blockdev_init(const char *file, Error **errp);
+void ns_blockdev_activate(BlockBackend *blk,  uint64_t image_size, Error **errp);
+int nvme_ns_backend_setup(NvmeCtrl *n, Error **errp);
 void nvme_attach_ns(NvmeCtrl *n, NvmeNamespace *ns);
 uint16_t nvme_bounce_data(NvmeCtrl *n, void *ptr, uint32_t len,
                           NvmeTxDirection dir, NvmeRequest *req);
@@ -583,5 +591,22 @@ uint16_t nvme_bounce_mdata(NvmeCtrl *n, void *ptr, uint32_t len,
 void nvme_rw_complete_cb(void *opaque, int ret);
 uint16_t nvme_map_dptr(NvmeCtrl *n, NvmeSg *sg, size_t len,
                        NvmeCmd *cmd);
+char *ns_create_image_name(NvmeCtrl *n, uint32_t nsid, Error **errp);
+int ns_storage_path_check(NvmeCtrl *n, Error **errp);
+int ns_auto_check(NvmeCtrl *n, NvmeNamespace *ns, uint32_t nsid);
+int ns_cfg_save(NvmeCtrl *n, NvmeNamespace *ns, uint32_t nsid);
+int ns_cfg_load(NvmeCtrl *n, NvmeNamespace *ns, uint32_t nsid);
+int64_t qdict_get_int_chkd(const QDict *qdict, const char *key, Error **errp);
+QList *qdict_get_qlist_chkd(const QDict *qdict, const char *key, Error **errp);
+void ns_cfg_clear(NvmeNamespace *ns);
+int nvme_cfg_save(NvmeCtrl *n);
+int nvme_cfg_load(NvmeCtrl *n);
+
+typedef enum NvmeNsAllocAction {
+    NVME_NS_ALLOC_CHK,
+    NVME_NS_ALLOC,
+    NVME_NS_DEALLOC,
+} NvmeNsAllocAction;
+int nvme_cfg_update(NvmeCtrl *n, uint64_t ammount, NvmeNsAllocAction action);
 
 #endif /* HW_NVME_NVME_H */
hw/nvme/subsys.c +7 -4
index 9d2643678b..5f4bb0e6a2 100644
@@ -87,10 +87,13 @@ int nvme_subsys_register_ctrl(NvmeCtrl *n, Error **errp)
 
     subsys->ctrls[cntlid] = n;
 
-    for (nsid = 1; nsid < ARRAY_SIZE(subsys->namespaces); nsid++) {
-        NvmeNamespace *ns = subsys->namespaces[nsid];
-        if (ns && ns->params.shared && !ns->params.detached) {
-            nvme_attach_ns(n, ns);
+    if (!n->params.ns_directory) {
+        /* legacy implementation */
+        for (nsid = 1; nsid < ARRAY_SIZE(subsys->namespaces); nsid++) {
+            NvmeNamespace *ns = subsys->namespaces[nsid];
+            if (ns && ns->params.shared && !ns->params.detached) {
+                nvme_attach_ns(n, ns);
+            }
         }
     }
 
hw/nvme/trace-events +2 -0
index fccb79f489..28b025ac42 100644
@@ -77,6 +77,8 @@ pci_nvme_aer(uint16_t cid) "cid %"PRIu16""
 pci_nvme_aer_aerl_exceeded(void) "aerl exceeded"
 pci_nvme_aer_masked(uint8_t type, uint8_t mask) "type 0x%"PRIx8" mask 0x%"PRIx8""
 pci_nvme_aer_post_cqe(uint8_t typ, uint8_t info, uint8_t log_page) "type 0x%"PRIx8" info 0x%"PRIx8" lid 0x%"PRIx8""
+pci_nvme_ns_mgmt(uint16_t cid, uint32_t nsid, uint8_t sel, uint8_t csi, uint8_t psdt) "cid %"PRIu16", nsid=%"PRIu32", sel=0x%"PRIx8", csi=0x%"PRIx8", psdt=0x%"PRIx8""
+pci_nvme_ns_create(uint16_t nsid, uint64_t nsze, uint64_t ncap, uint8_t flbas) "nsid %"PRIu16", nsze=%"PRIu64", ncap=%"PRIu64", flbas=%"PRIu8""
 pci_nvme_ns_attachment(uint16_t cid, uint8_t sel) "cid %"PRIu16", sel=0x%"PRIx8""
 pci_nvme_ns_attachment_attach(uint16_t cntlid, uint32_t nsid) "cntlid=0x%"PRIx16", nsid=0x%"PRIx32""
 pci_nvme_enqueue_event(uint8_t typ, uint8_t info, uint8_t log_page) "type 0x%"PRIx8" info 0x%"PRIx8" lid 0x%"PRIx8""
include/block/nvme.h +30 -0
index 8027b7126b..9d2e121f1a 100644
@@ -592,6 +592,7 @@ enum NvmeAdminCommands {
     NVME_ADM_CMD_SET_FEATURES   = 0x09,
     NVME_ADM_CMD_GET_FEATURES   = 0x0a,
     NVME_ADM_CMD_ASYNC_EV_REQ   = 0x0c,
+    NVME_ADM_CMD_NS_MGMT        = 0x0d,
     NVME_ADM_CMD_ACTIVATE_FW    = 0x10,
     NVME_ADM_CMD_DOWNLOAD_FW    = 0x11,
     NVME_ADM_CMD_NS_ATTACHMENT  = 0x15,
@@ -897,14 +898,18 @@ enum NvmeStatusCodes {
     NVME_FEAT_NOT_CHANGEABLE    = 0x010e,
     NVME_FEAT_NOT_NS_SPEC       = 0x010f,
     NVME_FW_REQ_SUSYSTEM_RESET  = 0x0110,
+    NVME_NS_INSUFFICIENT_CAPAC  = 0x0115,
+    NVME_NS_IDNTIFIER_UNAVAIL   = 0x0116,
     NVME_NS_ALREADY_ATTACHED    = 0x0118,
     NVME_NS_PRIVATE             = 0x0119,
     NVME_NS_NOT_ATTACHED        = 0x011a,
+    NVME_THIN_PROVISION_NOTSPRD = 0x011b,
     NVME_NS_CTRL_LIST_INVALID   = 0x011c,
     NVME_INVALID_CTRL_ID        = 0x011f,
     NVME_INVALID_SEC_CTRL_STATE = 0x0120,
     NVME_INVALID_NUM_RESOURCES  = 0x0121,
     NVME_INVALID_RESOURCE_ID    = 0x0122,
+    NVME_NS_ATTACH_MGMT_NOTSPRD = 0x0129,
     NVME_CONFLICTING_ATTRS      = 0x0180,
     NVME_INVALID_PROT_INFO      = 0x0181,
     NVME_WRITE_TO_RO            = 0x0182,
@@ -1184,6 +1189,10 @@ enum NvmeIdCtrlCmic {
     NVME_CMIC_MULTI_CTRL    = 1 << 1,
 };
 
+enum NvmeNsManagementOperation {
+    NVME_NS_MANAGEMENT_CREATE = 0x0,
+};
+
 enum NvmeNsAttachmentOperation {
     NVME_NS_ATTACHMENT_ATTACH = 0x0,
     NVME_NS_ATTACHMENT_DETACH = 0x1,
@@ -1345,6 +1354,26 @@ typedef struct QEMU_PACKED NvmeIdNs {
     uint8_t     vs[3712];
 } NvmeIdNs;
 
+typedef struct QEMU_PACKED NvmeIdNsMgmt {
+    uint64_t    nsze;
+    uint64_t    ncap;
+    uint8_t     rsvd16[10];
+    uint8_t     flbas;
+    uint8_t     rsvd27[2];
+    uint8_t     dps;
+    uint8_t     nmic;
+    uint8_t     rsvd31[61];
+    uint32_t    anagrpid;
+    uint8_t     rsvd96[4];
+    uint16_t    nvmsetid;
+    uint16_t    endgid;
+    uint8_t     rsvd104[280];
+    uint64_t    lbstm;
+    uint8_t     rsvd392[120];
+    uint8_t     rsvd512[512];
+    uint8_t     vs[3072];
+} NvmeIdNsMgmt;
+
 #define NVME_ID_NS_NVM_ELBAF_PIF(elbaf) (((elbaf) >> 7) & 0x3)
 
 typedef struct QEMU_PACKED NvmeIdNsNvm {
@@ -1646,6 +1675,7 @@ static inline void _nvme_check_size(void)
     QEMU_BUILD_BUG_ON(sizeof(NvmeLBAF) != 4);
     QEMU_BUILD_BUG_ON(sizeof(NvmeLBAFE) != 16);
     QEMU_BUILD_BUG_ON(sizeof(NvmeIdNs) != 4096);
+    QEMU_BUILD_BUG_ON(sizeof(NvmeIdNsMgmt) != 4096);
     QEMU_BUILD_BUG_ON(sizeof(NvmeIdNsNvm) != 4096);
     QEMU_BUILD_BUG_ON(sizeof(NvmeIdNsZoned) != 4096);
     QEMU_BUILD_BUG_ON(sizeof(NvmeSglDescriptor) != 16);
include/hw/nvme/ctrl-cfg.h new file +24 -0
new file mode 100644
index 0000000000..1be44cb8df
@@ -0,0 +1,24 @@
+/*
+ * QEMU NVM Express Virtual Dynamic Namespace Management
+ * Common configuration handling for qemu-img tool and and qemu-system-xx
+ *
+ *
+ * Copyright (c) 2022 Solidigm
+ *
+ * Authors:
+ *  Michael Kropaczek      <[email protected]>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef CTRL_CFG_DEF
+#define NVME_STR_(s) #s
+#define NVME_STRINGIFY(s) NVME_STR_(s)
+#define NVME_MAX_NAMESPACES  256
+#define NVME_MAX_CONTROLLERS 256
+#else
+CTRL_CFG_DEF(int, "tnvmcap", int128_get64(tnvmcap128), tnvmcap64)
+CTRL_CFG_DEF(int, "unvmcap", int128_get64(unvmcap128), unvmcap64)
+#endif
include/hw/nvme/ns-cfg.h new file +28 -0
new file mode 100644
index 0000000000..5f27a6cece
@@ -0,0 +1,28 @@
+/*
+ * QEMU NVM Express Virtual Dynamic Namespace Management
+ * Common configuration handling for qemu-img tool and qemu-system-xx
+ *
+ *
+ * Copyright (c) 2022 Solidigm
+ *
+ * Authors:
+ *  Michael Kropaczek      <[email protected]>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ *
+ */
+
+#ifdef NS_CFG_DEF
+NS_CFG_DEF(int, "params.nsid", (int64_t)ns->params.nsid, nsid)
+NS_CFG_DEF(obj, "attached_ctrls", QOBJECT(ctrl_qlist), QOBJECT(ctrl_qlist))
+NS_CFG_DEF(int, "params.pi", (int64_t)ns->params.pi, 0)
+NS_CFG_DEF(int, "lbasz", (int64_t)ns->lbasz, 0)
+NS_CFG_DEF(int, "id_ns.nsze", le64_to_cpu(ns->id_ns.nsze), 0)
+NS_CFG_DEF(int, "id_ns.ncap", le64_to_cpu(ns->id_ns.ncap), 0)
+NS_CFG_DEF(int, "id_ns.nuse", le64_to_cpu(ns->id_ns.nuse), 0)
+NS_CFG_DEF(int, "id_ns.nsfeat", (int64_t)ns->id_ns.nsfeat, 0)
+NS_CFG_DEF(int, "id_ns.flbas", (int64_t)ns->id_ns.flbas, 0)
+NS_CFG_DEF(int, "id_ns.nmic", (int64_t)ns->id_ns.nmic, 0)
+NS_CFG_DEF(int, "ns_size", ns->size, 0)
+#endif
include/hw/nvme/nvme-cfg.h new file +188 -0

188 lines not on this page · raw message

qemu-img-cmds.hx +6 -0
index 1b1dab5b17..9aacb88fc9 100644
@@ -57,6 +57,12 @@ SRST
 .. option:: create [--object OBJECTDEF] [-q] [-f FMT] [-b BACKING_FILE [-F BACKING_FMT]] [-u] [-o OPTIONS] FILENAME [SIZE]
 ERST
 
+DEF("createns", nsimgs_create,
+    "createns -S nvme_ctrl_serial_number -C nvme_ctrl_total_capacity [-N <NsId_max>] pathname")
+SRST
+.. option:: createns -S SERIAL_NUMBER -C TOTAL_CAPACITY [-N NSID_MAX] PATHNAME
+ERST
+
 DEF("dd", img_dd,
     "dd [--image-opts] [-U] [-f fmt] [-O output_fmt] [bs=block_size] [count=blocks] [skip=blocks] if=input of=output")
 SRST
qemu-img.c +132 -0

160 lines not on this page · raw message

Signature · 1 line
2.37.3
#2 lore raw

Re: [PATCH v4 1/2] hw/nvme: Support for Namespaces Management from guest OS - create-ns

Quoted 15 lines · Jonathan Derrick
hw/nvme/ctrl.c +15 -0
+static uint16_t nvme_ns_mgmt(NvmeCtrl *n, NvmeRequest *req)
+{
+    NvmeCtrl *n_p = NULL;     /* primary controller */
+    NvmeIdCtrl *id = &n->id_ctrl;
+    NvmeNamespace *ns;
+    NvmeIdNsMgmt id_ns = {};
+    uint8_t flags = req->cmd.flags;
+    uint32_t nsid = le32_to_cpu(req->cmd.nsid);
+    uint32_t dw10 = le32_to_cpu(req->cmd.cdw10);
+    uint32_t dw11 = le32_to_cpu(req->cmd.cdw11);
+    uint8_t sel = dw10 & 0xf;
+    uint8_t csi = (dw11 >> 24) & 0xf;
+    uint16_t i;
+    uint16_t ret;
+    Error *local_err = NULL;

hw/nvme/ctrl.c static void nvme_select_iocs_ns(NvmeCtrl *n, NvmeNamespace *ns) · line 5687

Quoted 10 lines
hw/nvme/ctrl.c +10 -0
+
+    trace_pci_nvme_ns_mgmt(nvme_cid(req), nsid, sel, csi, NVME_CMD_FLAGS_PSDT(flags));
+
+    if (!(le16_to_cpu(id->oacs) & NVME_OACS_NS_MGMT)) {
+        return NVME_NS_ATTACH_MGMT_NOTSPRD | NVME_DNR;
+    }
+
+    if (n->cntlid && !n->subsys) {
+        error_setg(&local_err, "Secondary controller without subsystem");
+        return NVME_NS_ATTACH_MGMT_NOTSPRD | NVME_DNR;

Leaking local_err. Any time you call error_setg(), the error needs to be reported or freed at some point.

hw/nvme/ctrl.c static void nvme_select_iocs_ns(NvmeCtrl *n, NvmeNamespace *ns) · line 5695

Quoted 8 lines
hw/nvme/ctrl.c +8 -0
+    }
+
+    n_p = n->subsys->ctrls[0];
+
+    switch (sel) {
+    case NVME_NS_MANAGEMENT_CREATE:
+        switch (csi) {
+        case NVME_CSI_NVM:

The following case is sufficiently large enough that the implementation should be its own function.

hw/nvme/ctrl.c static void nvme_select_iocs_ns(NvmeCtrl *n, NvmeNamespace *ns) · line 5706

Quoted 11 lines
hw/nvme/ctrl.c +11 -0
+            if (nsid) {
+                return NVME_INVALID_FIELD | NVME_DNR;
+            }
+
+            ret = nvme_h2c(n, (uint8_t *)&id_ns, sizeof(id_ns), req);
+            if (ret) {
+                return ret;
+            }
+
+            uint64_t nsze = le64_to_cpu(id_ns.nsze);
+            uint64_t ncap = le64_to_cpu(id_ns.ncap);

Please don't mix declarations with code; declare these local variables at the top of the scope.

Quoted 29 lines
hw/nvme/ctrl.c +29 -0
+
+            if (ncap > nsze) {
+                return NVME_INVALID_FIELD | NVME_DNR;
+            } else if (ncap != nsze) {
+                return NVME_THIN_PROVISION_NOTSPRD | NVME_DNR;
+            }
+
+            nvme_validate_flbas(id_ns.flbas, &local_err);
+            if (local_err) {
+                error_report_err(local_err);
+                return NVME_INVALID_FORMAT | NVME_DNR;
+            }
+
+            for (i = 1; i <= NVME_MAX_NAMESPACES; i++) {
+                if (nvme_ns(n_p, (uint32_t)i) || nvme_subsys_ns(n_p->subsys, (uint32_t)i)) {
+                    continue;
+                }
+                break;
+            }
+
+
+            if (i >  le32_to_cpu(n_p->id_ctrl.nn) || i >  NVME_MAX_NAMESPACES) {
+               return NVME_NS_IDNTIFIER_UNAVAIL | NVME_DNR;
+            }
+            nsid = i;
+
+            /* create ns here */
+            ns = nvme_ns_mgmt_create(n, nsid, &id_ns, &local_err);
+            if (!ns || local_err) {

hw/nvme/ctrl.c static void nvme_select_iocs_ns(NvmeCtrl *n, NvmeNamespace *ns) · line 5745

Quoted 10 lines
hw/nvme/ctrl.c +10 -0
+                if (local_err) {
+                    error_report_err(local_err);
+                }
+                return NVME_INVALID_FIELD | NVME_DNR;
+            }
+
+            if (nvme_cfg_update(n, ns->size, NVME_NS_ALLOC_CHK)) {
+                /* place for delete-ns */
+                error_setg(&local_err, "Insufficient capacity, an orphaned ns[%"PRIu32"] will be left behind", nsid);
+                return NVME_NS_INSUFFICIENT_CAPAC | NVME_DNR;

Leaked local_err.

hw/nvme/ctrl.c static void nvme_select_iocs_ns(NvmeCtrl *n, NvmeNamespace *ns) · line 5752

Quoted 7 lines
hw/nvme/ctrl.c +7 -0
+            }
+            (void)nvme_cfg_update(n, ns->size, NVME_NS_ALLOC);
+            if (nvme_cfg_save(n)) {
+                (void)nvme_cfg_update(n, ns->size, NVME_NS_DEALLOC);
+                /* place for delete-ns */
+                error_setg(&local_err, "Cannot save conf file, an orphaned ns[%"PRIu32"] will be left behind", nsid);
+                return NVME_INVALID_FIELD | NVME_DNR;

Another leaked local_err.

Quoted 15 lines
hw/nvme/ctrl.c +15 -0
+            }
+            req->cqe.result = cpu_to_le32(nsid);
+            break;
+        case NVME_CSI_ZONED:
+            /* fall through for now */
+        default:
+            return NVME_INVALID_FIELD | NVME_DNR;
+	    }
+        break;
+    default:
+        return NVME_INVALID_FIELD | NVME_DNR;
+    }
+
+    return NVME_SUCCESS;
+}
#3 lore raw

RE: [PATCH v4 1/2] hw/nvme: Support for Namespaces Management from guest OS - create-ns

Hi Keith,

Thank you for the review and I agree with you, the issues will be addressed.

Regards,
Michael

-----Original Message-----
From: Keith Busch <[email protected]>
Sent: Wednesday, December 28, 2022 12:10 PM
To: Jonathan Derrick <[email protected]>
Cc: [email protected]; Michael Kropaczek (CW) <[email protected]>; [email protected]; Klaus Jensen <[email protected]>; Kevin Wolf <[email protected]>; Hanna Reitz <[email protected]>
Subject: Re: [PATCH v4 1/2] hw/nvme: Support for Namespaces Management from guest OS - create-ns

[You don't often get email from [email protected]. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]

Caution: External Email

Quoted 15 lines · Jonathan Derrick
hw/nvme/ctrl.c +15 -0
+static uint16_t nvme_ns_mgmt(NvmeCtrl *n, NvmeRequest *req) {
+    NvmeCtrl *n_p = NULL;     /* primary controller */
+    NvmeIdCtrl *id = &n->id_ctrl;
+    NvmeNamespace *ns;
+    NvmeIdNsMgmt id_ns = {};
+    uint8_t flags = req->cmd.flags;
+    uint32_t nsid = le32_to_cpu(req->cmd.nsid);
+    uint32_t dw10 = le32_to_cpu(req->cmd.cdw10);
+    uint32_t dw11 = le32_to_cpu(req->cmd.cdw11);
+    uint8_t sel = dw10 & 0xf;
+    uint8_t csi = (dw11 >> 24) & 0xf;
+    uint16_t i;
+    uint16_t ret;
+    Error *local_err = NULL;
+

hw/nvme/ctrl.c static void nvme_select_iocs_ns(NvmeCtrl *n, NvmeNamespace *ns) · line 5687

Quoted 10 lines
hw/nvme/ctrl.c +10 -0
+    trace_pci_nvme_ns_mgmt(nvme_cid(req), nsid, sel, csi, 
+ NVME_CMD_FLAGS_PSDT(flags));
+
+    if (!(le16_to_cpu(id->oacs) & NVME_OACS_NS_MGMT)) {
+        return NVME_NS_ATTACH_MGMT_NOTSPRD | NVME_DNR;
+    }
+
+    if (n->cntlid && !n->subsys) {
+        error_setg(&local_err, "Secondary controller without subsystem");
+        return NVME_NS_ATTACH_MGMT_NOTSPRD | NVME_DNR;

Leaking local_err. Any time you call error_setg(), the error needs to be reported or freed at some point.

hw/nvme/ctrl.c static void nvme_select_iocs_ns(NvmeCtrl *n, NvmeNamespace *ns) · line 5695

Quoted 8 lines
hw/nvme/ctrl.c +8 -0
+    }
+
+    n_p = n->subsys->ctrls[0];
+
+    switch (sel) {
+    case NVME_NS_MANAGEMENT_CREATE:
+        switch (csi) {
+        case NVME_CSI_NVM:

The following case is sufficiently large enough that the implementation should be its own function.

hw/nvme/ctrl.c static void nvme_select_iocs_ns(NvmeCtrl *n, NvmeNamespace *ns) · line 5706

Quoted 11 lines
hw/nvme/ctrl.c +11 -0
+            if (nsid) {
+                return NVME_INVALID_FIELD | NVME_DNR;
+            }
+
+            ret = nvme_h2c(n, (uint8_t *)&id_ns, sizeof(id_ns), req);
+            if (ret) {
+                return ret;
+            }
+
+            uint64_t nsze = le64_to_cpu(id_ns.nsze);
+            uint64_t ncap = le64_to_cpu(id_ns.ncap);

Please don't mix declarations with code; declare these local variables at the top of the scope.

Quoted 29 lines
hw/nvme/ctrl.c +29 -0
+
+            if (ncap > nsze) {
+                return NVME_INVALID_FIELD | NVME_DNR;
+            } else if (ncap != nsze) {
+                return NVME_THIN_PROVISION_NOTSPRD | NVME_DNR;
+            }
+
+            nvme_validate_flbas(id_ns.flbas, &local_err);
+            if (local_err) {
+                error_report_err(local_err);
+                return NVME_INVALID_FORMAT | NVME_DNR;
+            }
+
+            for (i = 1; i <= NVME_MAX_NAMESPACES; i++) {
+                if (nvme_ns(n_p, (uint32_t)i) || nvme_subsys_ns(n_p->subsys, (uint32_t)i)) {
+                    continue;
+                }
+                break;
+            }
+
+
+            if (i >  le32_to_cpu(n_p->id_ctrl.nn) || i >  NVME_MAX_NAMESPACES) {
+               return NVME_NS_IDNTIFIER_UNAVAIL | NVME_DNR;
+            }
+            nsid = i;
+
+            /* create ns here */
+            ns = nvme_ns_mgmt_create(n, nsid, &id_ns, &local_err);
+            if (!ns || local_err) {

hw/nvme/ctrl.c static void nvme_select_iocs_ns(NvmeCtrl *n, NvmeNamespace *ns) · line 5745

Quoted 10 lines
hw/nvme/ctrl.c +10 -0
+                if (local_err) {
+                    error_report_err(local_err);
+                }
+                return NVME_INVALID_FIELD | NVME_DNR;
+            }
+
+            if (nvme_cfg_update(n, ns->size, NVME_NS_ALLOC_CHK)) {
+                /* place for delete-ns */
+                error_setg(&local_err, "Insufficient capacity, an orphaned ns[%"PRIu32"] will be left behind", nsid);
+                return NVME_NS_INSUFFICIENT_CAPAC | NVME_DNR;

Leaked local_err.

hw/nvme/ctrl.c static void nvme_select_iocs_ns(NvmeCtrl *n, NvmeNamespace *ns) · line 5752

Quoted 7 lines
hw/nvme/ctrl.c +7 -0
+            }
+            (void)nvme_cfg_update(n, ns->size, NVME_NS_ALLOC);
+            if (nvme_cfg_save(n)) {
+                (void)nvme_cfg_update(n, ns->size, NVME_NS_DEALLOC);
+                /* place for delete-ns */
+                error_setg(&local_err, "Cannot save conf file, an orphaned ns[%"PRIu32"] will be left behind", nsid);
+                return NVME_INVALID_FIELD | NVME_DNR;

Another leaked local_err.

Quoted 15 lines
hw/nvme/ctrl.c +15 -0
+            }
+            req->cqe.result = cpu_to_le32(nsid);
+            break;
+        case NVME_CSI_ZONED:
+            /* fall through for now */
+        default:
+            return NVME_INVALID_FIELD | NVME_DNR;
+         }
+        break;
+    default:
+        return NVME_INVALID_FIELD | NVME_DNR;
+    }
+
+    return NVME_SUCCESS;
+}
#4 lore raw

PATCH v4 2/2 hw/nvme: Support for Namespaces Management from guest OS - delete-ns

In reply to Jonathan Derrick

From: Michael Kropaczek <[email protected]>

Added support for NVMEe NameSpaces Mangement allowing the guest OS to delete namespaces by issuing nvme delete-ns command.
It is an extension to currently implemented Qemu nvme virtual device. Virtual devices representing namespaces will be created and/or deleted during Qemu's running session, at anytime.

---
 docs/system/devices/nvme.rst |  9 ++--
 hw/nvme/ctrl.c               | 81 +++++++++++++++++++++++++++++++++---
 hw/nvme/ns-backend.c         |  5 +++
 hw/nvme/ns.c                 | 72 ++++++++++++++++++++++++++++++++
 hw/nvme/nvme.h               |  3 +-
 hw/nvme/trace-events         |  1 +
 include/block/nvme.h         |  1 +
 7 files changed, 161 insertions(+), 11 deletions(-)
docs/system/devices/nvme.rst +5 -4
index 6b3bee5e5d..f19072f1bc 100644
@@ -103,12 +103,12 @@ Parameters:
 
 ``auto-ns-path=<path to nvme storage location>``
   If specified indicates a support for dynamic management of nvme namespaces
-  by means of nvme create-ns command. This path points
+  by means of nvme create-ns and nvme delete-ns commands. This path points
   to the storage area for backend images must exist. Additionally it requires
   that parameter `ns-subsys` must be specified whereas parameter `drive`
   must not. The legacy namespace backend is disabled, instead, a pair of
   files 'nvme_<ctrl SN>_ns_<NS-ID>.cfg' and 'nvme_<ctrl SN>_ns_<NS-ID>.img'
-  will refer to respective namespaces. The create-ns, attach-ns
+  will refer to respective namespaces. The create-ns, delete-ns, attach-ns
   and detach-ns commands, issued at the guest side, will make changes to
   those files accordingly.
   For each namespace exists an image file in raw format and a config file
@@ -140,8 +140,9 @@ Please note that ``nvme-ns`` device is not required to support of dynamic
 namespaces management feature. It is not prohibited to assign a such device to
 ``nvme`` device specified to support dynamic namespace management if one has
 an use case to do so, however, it will only coexist and be out of the scope of
-Namespaces Management. NsIds will be consistently managed, creation (create-ns)
-of a namespace will not allocate the NsId already being taken. If ``nvme-ns``
+Namespaces Management. Deletion (delete-ns) will render an error for this
+namespace. NsIds will be consistently managed, creation (create-ns) of
+a namespace will not allocate the NsId already being taken. If ``nvme-ns``
 device conflicts with previously created one by create-ns (the same NsId),
 it will break QEMU's start up.
 More than one of NVMe controllers associated with NVMe subsystem are supported.
hw/nvme/ctrl.c +75 -6

121 lines not on this page · raw message

hw/nvme/ns-backend.c +5 -0
index 06de8f262c..57d0b695fa 100644
@@ -71,6 +71,11 @@ void ns_blockdev_activate(BlockBackend *blk,  uint64_t image_size, Error **errp)
                        errp);
 }
 
+void ns_blockdev_deactivate(BlockBackend *blk, Error **errp)
+{
+    ns_blockdev_activate(blk, 0, errp);
+}
+
 int ns_storage_path_check(NvmeCtrl *n, Error **errp)
 {
     return storage_path_check(n->params.ns_directory,  n->params.serial, errp);
hw/nvme/ns.c +72 -0

78 lines not on this page · raw message

hw/nvme/nvme.h +2 -1
index 15b790cbdd..2c3ed4a57e 100644
@@ -280,6 +280,7 @@ void nvme_ns_shutdown(NvmeNamespace *ns);
 void nvme_ns_cleanup(NvmeNamespace *ns);
 void nvme_validate_flbas(uint8_t flbas,  Error **errp);
 NvmeNamespace * nvme_ns_create(NvmeCtrl *n, uint32_t nsid, NvmeIdNsMgmt *id_ns, Error **errp);
+void nvme_ns_delete(NvmeCtrl *n, uint32_t nsid, Error **errp);
 
 typedef struct NvmeAsyncEvent {
     QTAILQ_ENTRY(NvmeAsyncEvent) entry;
@@ -513,7 +514,6 @@ typedef struct NvmeCtrl {
         uint16_t    vqrfap;
         uint16_t    virfap;
     } next_pri_ctrl_cap;    /* These override pri_ctrl_cap after reset */
-    uint16_t nsidMax;
 } NvmeCtrl;
 
 typedef enum NvmeResetType {
@@ -582,6 +582,7 @@ static inline NvmeSecCtrlEntry *nvme_sctrl_for_cntlid(NvmeCtrl *n,
 
 BlockBackend *ns_blockdev_init(const char *file, Error **errp);
 void ns_blockdev_activate(BlockBackend *blk,  uint64_t image_size, Error **errp);
+void ns_blockdev_deactivate(BlockBackend *blk, Error **errp);
 int nvme_ns_backend_setup(NvmeCtrl *n, Error **errp);
 void nvme_attach_ns(NvmeCtrl *n, NvmeNamespace *ns);
 uint16_t nvme_bounce_data(NvmeCtrl *n, void *ptr, uint32_t len,
hw/nvme/trace-events +1 -0
index 28b025ac42..0dd0c23208 100644
@@ -79,6 +79,7 @@ pci_nvme_aer_masked(uint8_t type, uint8_t mask) "type 0x%"PRIx8" mask 0x%"PRIx8"
 pci_nvme_aer_post_cqe(uint8_t typ, uint8_t info, uint8_t log_page) "type 0x%"PRIx8" info 0x%"PRIx8" lid 0x%"PRIx8""
 pci_nvme_ns_mgmt(uint16_t cid, uint32_t nsid, uint8_t sel, uint8_t csi, uint8_t psdt) "cid %"PRIu16", nsid=%"PRIu32", sel=0x%"PRIx8", csi=0x%"PRIx8", psdt=0x%"PRIx8""
 pci_nvme_ns_create(uint16_t nsid, uint64_t nsze, uint64_t ncap, uint8_t flbas) "nsid %"PRIu16", nsze=%"PRIu64", ncap=%"PRIu64", flbas=%"PRIu8""
+pci_nvme_ns_delete(uint16_t nsid) "nsid %"PRIu16""
 pci_nvme_ns_attachment(uint16_t cid, uint8_t sel) "cid %"PRIu16", sel=0x%"PRIx8""
 pci_nvme_ns_attachment_attach(uint16_t cntlid, uint32_t nsid) "cntlid=0x%"PRIx16", nsid=0x%"PRIx32""
 pci_nvme_enqueue_event(uint8_t typ, uint8_t info, uint8_t log_page) "type 0x%"PRIx8" info 0x%"PRIx8" lid 0x%"PRIx8""
include/block/nvme.h +1 -0
index 9d2e121f1a..0fe7fe9bb1 100644
@@ -1191,6 +1191,7 @@ enum NvmeIdCtrlCmic {
 
 enum NvmeNsManagementOperation {
     NVME_NS_MANAGEMENT_CREATE = 0x0,
+    NVME_NS_MANAGEMENT_DELETE = 0x1,
 };
 
 enum NvmeNsAttachmentOperation {
Signature · 1 line
2.37.3
#5 lore raw
In reply to Jonathan Derrick

Am 28.12.2022 um 20:41 hat Jonathan Derrick geschrieben:

Quoted 19 lines

Here is the approach:
The nvme device will get new parameter:
- auto-ns-path, which specifies the path to the storage area where back-end image and necessary config files located stored.

The virtual devices representing namespaces will be created dynamically during the Qemu running session following issuance of nvme create-ns and delete-ns commands from the guest OS. QOM classes and instances will be created utilizing existing configuration scheme used during Qemu's start-up. Back-end image files will be neither created nor deleted during Qemu's startup or running session. Instead a set of back-end image files and relevant config will be created by qemu-img tool with createns sub-command prior to Qemu's session. Required parameters are: -S serial number which must match serial parameter of qemu-system-xx -device nvme command line specification, -C total capacity, and optional -N that will set a maximal limit on number of allowed namespaces (default 256) which will be followed by path name pointing to storage location corresponding to auto-ns-path of qemu-system-xx -device nvme parameter.

Quoted 16 lines

Those created back-end image files will be pre-loaded during Qemu's start-up and then during running Qemu's session will be associated or disassociated with QOM namespaces virtual instances, as a result of issuing nvme create-ns or delete-ns commands. The associated back-end image file for relevant namespace will be re-sized as follows: delete-ns command will truncate image file to the size of 0, whereas create-ns command will re-size the image file to the size provided by nvme create-ns command parameters. Truncating/re-sizing is a result of blk_truncate() API which utilizes co-routines and should not block Qemu main thread while scheduling AIO operations. It is assured that all settings will retain over Qemu's start-ups and shutdowns. The implementation makes it possible to combine the existing "Additional Namespace" implementation with the new "Managed Namespaces". Those will coexist with obvious restrictions, like both will share the same NsIds space, "static" namespaces cannot be deleted or if its NsId specified at Qemu's command line will conflicts with previously created one by nvme create-ns (and retained), this will lead to an abort of Qemu at its start up.

This looks like a valid approach for a proof of concept, but from a backend perspective, I'm concerned that this approach might be too limiting and we won't have a good path forward.

For example, how can we integrate this with snapshots? You expect a specific filename for the image, but taking an external snapshot means creating an overlay image with a different name.

How do we migrate storage like this? If the management tool (probably libvirt) knows about all the namespace images and the config file (!), it can possibly migrate them individually, but note that while a mirror job is active, images can't be resized any more.

What if we don't want to use a directory on the local filesystem to store the images, but some network protocol?

It seems to me that we should define proper block layer APIs for handling namespaces, and then we can have your implementation as one possible image driver that supports these APIs, for which we can accept these limitations for now. At least this would already avoid having backend logic in the device implementation, and allow us to replace it with something better later without having to change the design of the device emulation code.

Eventually, I think, if we want to have dynamic namespaces properly supported, they need to be a feature on the image format level, so that you could keep all namespaces in a single qcow2 file.

Kevin