MCUmgr

Overview

The management subsystem allows remote management of Zephyr-enabled devices. The following management operations are available:

Image management
File System management
OS management
Settings (config) management
Shell management
Statistic management
Zephyr management

over the following transports:

BLE (Bluetooth Low Energy)
Serial (UART)
UDP over IP

The management subsystem is based on the Simple Management Protocol (SMP) provided by MCUmgr, an open source project that provides a management subsystem that is portable across multiple real-time operating systems.

The management subsystem is located in subsys/mgmt/ inside of the Zephyr tree.

Additionally, there is a sample server that provides management functionality over BLE and serial.

Tools/libraries

There are various tools and libraries available which enable usage of MCUmgr functionality on a device which are listed below. Note that these tools are not part of or related to the Zephyr project.

Tools and Libraries for MCUmgr
Name	OS support					Transports			Groups							Type	Language	License
Name	Windows	Linux	mac	Mobile	Embedded	Serial	Bluetooth	UDP	OS	IMG	Stat	Settings	FS	Shell	Zephyr	Type	Language	License
AuTerm	✓	✓	✓	✕	✕	✓	✓	✓	✓	✓	✓	✓	✓	✓	✓	Application	C++ (Qt)	GPLv3
mcumgr-client	✓	✓	✓	✕	✕	✓	✕	✕	✕	✓	✕	✕	✕	✕	✕	Application	Rust	BSD
mcumgr-web	✓	✓	✓	✕	✕	✕	✓	✕	✕	✓	✕	✕	✕	✕	✕	Web page (chrome only)	Javascript	MIT
nRF Connect Device Manager: Android and iOS	✕	✕	✕	✓	✕	✕	✓	✕	✓	✓	✓	✓	✓	✓	✓	Library and application	Java, Kotlin, Swift	Apache
Zephyr MCUmgr client (in-tree)	✕	✓	✕	✕	✓	✓	✓	✓	✓	✓	✕	✕	✕	✕	✕	Library	C	Apache

Note that a tick for a particular group indicates basic support for that group in the code, it is possible that not all commands/features of a group are supported by the implementation.

Command-line Tool

MCUmgr provides a command-line tool, mcumgr, for managing remote devices. The tool is written in the Go programming language.

Note

This tool is provided for evaluation use only and is not recommended for use in a production environment. It has known issues and will not respect the MCUmgr protocol properly e.g. when an error is received, instead of aborting will, in some circumstances, sit in an endless loop of sending the same command over and over again. A universal replacement for this tool is currently in development and once released, support for the go tool will be dropped entirely. It is recommended that usage of tools listed above in the Tools/libraries section are used instead of the go client.

To install the tool:

go get github.com/apache/mynewt-mcumgr-cli/mcumgr

go install github.com/apache/mynewt-mcumgr-cli/mcumgr@latest

Configuring the transport

There are two command-line options that are responsible for setting and configuring the transport layer to use when communicating with managed device:

--conntype is used to choose the transport used, and
--connstring is used to pass a comma separated list of options in the key=value format, where each valid key depends on the particular conntype.

Valid transports for --conntype are serial, ble and udp. Each transport expects a different set of key/value options:

--connstring accepts the following key values:

`dev`	the device name for the OS `mcumgr` is running on (eg, `/dev/ttyUSB0`, `/dev/tty.usbserial`, `COM1`, etc).
`baud`	the communication speed; must match the baudrate of the server.
`mtu`	aka Maximum Transmission Unit, the maximum protocol packet size.

--connstring accepts the following key values:

`ctlr_name`	an OS specific string for the controller name.
`own_addr_type`	can be one of `public`, `random`, `rpa_pub`, `rpa_rnd`, where `random` is the default.
`peer_name`	the name the peer BLE device advertises, this should match the configuration specified with `CONFIG_BT_DEVICE_NAME`.
`peer_id`	the peer BLE device address or UUID. Only required when `peer_name` was not given. The format depends on the OS where `mcumgr` is run, it is a 6 bytes hexadecimal string separated by colons on Linux, or a 128-bit UUID on macOS.
`conn_timeout`	a float number representing the connection timeout in seconds.

--connstring takes the form [addr]:port where:

`addr`	can be a DNS name (if it can be resolved to the device IP), IPv4 addr (app must be built with `CONFIG_MCUMGR_TRANSPORT_UDP_IPV4`), or IPv6 addr (app must be built with `CONFIG_MCUMGR_TRANSPORT_UDP_IPV6`)
`port`	any valid UDP port.

Saving the connection config

The transport configuration can be managed with the conn sub-command and later used with --conn (or -c) parameter to skip typing both --conntype and --connstring. For example a new config for a serial device that would require typing mcumgr --conntype serial --connstring dev=/dev/ttyACM0,baud=115200,mtu=512 can be saved with:

mcumgr conn add acm0 type="serial" connstring="dev=/dev/ttyACM0,baud=115200,mtu=512"

Accessing this port can now be done with:

mcumgr -c acm0

General options

Some options work for every mcumgr command and might be helpful to debug and fix issues with the communication, among them the following deserve special mention:

`-l <log-level>`	Configures the log level, which can be one of `critical`, `error`, `warn`, `info` or `debug`, from less to most verbose. When there are communication issues, `-lDEBUG` might be useful to dump the packets for later inspection.
`-t <timeout>`	Changes the timeout waiting for a response from the default of 10s to a given value. Some commands might take a long time of processing, eg, the erase before an image upload, and might need incrementing the timeout to a larger value.
`-r <tries>`	Changes the number of retries on timeout from the default of 1 to a given value.

List of Commands

Not all commands defined by mcumgr (and SMP protocol) are currently supported on Zephyr. The ones that are supported are described in the following table:

Tip

Running mcumgr with no parameters, or -h will display the list of commands.

Command	Description
`echo`	Send data to a device and display the echoed back data. This command is part of the `OS` group, which must be enabled by setting `CONFIG_MCUMGR_GRP_OS`. The `echo` command itself can be enabled by setting `CONFIG_MCUMGR_GRP_OS_ECHO`.
`fs`	Access files on a device. More info in Filesystem Management.
`image`	Manage images on a device. More info in Image Management.
`reset`	Perform a soft reset of a device. This command is part of the `OS` group, which must be enabled by setting `CONFIG_MCUMGR_GRP_OS`. The `reset` command itself is always enabled and the time taken for a reset to happen can be set with `CONFIG_MCUMGR_GRP_OS_RESET_MS` (in ms).
`shell`	Execute a command in the remote shell. This option is disabled by default and can be enabled with `CONFIG_MCUMGR_GRP_SHELL` = `y`. To know more about the shell in Zephyr check Shell.
`stat`	Read statistics from a device. More info in Statistics Management.
`taskstat`	Read task statistics from a device. This command is part of the `OS` group, which must be enabled by setting `CONFIG_MCUMGR_GRP_OS`. The `taskstat` command itself can be enabled by setting `CONFIG_MCUMGR_GRP_OS_TASKSTAT`. `CONFIG_THREAD_MONITOR` also needs to be enabled otherwise a `-8` (`MGMT_ERR_ENOTSUP`) will be returned.

Tip

taskstat has a few options that might require tweaking. The CONFIG_THREAD_NAME must be set to display the task names, otherwise the priority is displayed. Since the taskstat packets are large, they might need increasing the CONFIG_MCUMGR_TRANSPORT_NETBUF_SIZE option.

Warning

To display the correct stack size in the taskstat command, the CONFIG_THREAD_STACK_INFO option must be set. To display the correct stack usage in the taskstat command, both CONFIG_THREAD_STACK_INFO and CONFIG_INIT_STACKS options must be set.

J-Link Virtual MSD Interaction Note

On boards where a J-Link OB is present which has both CDC and MSC (virtual Mass Storage Device, also known as drag-and-drop) support, the MSD functionality can prevent MCUmgr commands over the CDC UART port from working due to how USB endpoints are configured in the J-Link firmware (for example on the Nordic nrf52840dk/nrf52840 board) because of limiting the maximum packet size (most likely to occur when using image management commands for updating firmware). This issue can be resolved by disabling MSD functionality on the J-Link device, follow the instructions on Disabling the Mass Storage Device functionality to disable MSD support.

Image Management

The image management provided by mcumgr is based on the image format defined by MCUboot. For more details on the internals see MCUboot design and Signing Binaries.

To list available images in a device:

mcumgr <connection-options> image list

This should result in an output similar to this:

$ mcumgr -c acm0 image list
Images:
  image=0 slot=0
    version: 1.0.0
    bootable: true
    flags: active confirmed
    hash: 86dca73a3439112b310b5e033d811ec2df728d2264265f2046fced5a9ed00cc7
Split status: N/A (0)

Where image is the number of the image pair in a multi-image system, and slot is the number of the slot where the image is stored, 0 for primary and 1 for secondary. This image being active and confirmed means it will run again on next reset. Also relevant is the hash, which is used by other commands to refer to this specific image when performing operations.

An image can be manually erased using:

mcumgr <connection-options> image erase

The behavior of erase is defined by the server (MCUmgr in the device). The current implementation is limited to erasing the image in the secondary partition.

To upload a new image:

mcumgr <connection-options> image upload [-n] [-e] [-u] [-w] <signed-bin>

-n: This option allows uploading a new image to a specific set of images in a multi-image system, and is currently only supported by MCUboot when the CONFIG_MCUBOOT_SERIAL option is enabled.
-e: This option avoids performing a full erase of the partition before starting a new upload.

Tip

The -e option should always be passed in because the upload command already checks if an erase is required, respecting the CONFIG_IMG_ERASE_PROGRESSIVELY setting.

Tip

If the upload command times out while waiting for a response from the device, -t might be used to increase the wait time to something larger than the default of 10s. See general_options.

Warning

mcumgr does not understand .hex files, when uploading a new image always use the .bin file.

-u: This option allows upgrading only to newer image version.
-w: This option allows setting the maximum size for the window of outstanding chunks in transit. It is set to 5 by default.

Tip

If the option is set to a value lower than the default one, for example -w 1, less chunks are transmitted on the window, resulting in lower risk of errors. Conversely, setting a value higher than 5 increases risk of errors and may impact performance.

After an image upload is finished, a new image list would now have an output like this:

$ mcumgr -c acm0 image upload -e build/zephyr/zephyr.signed.bin
  35.69 KiB / 92.92 KiB [==========>---------------]  38.41% 2.97 KiB/s 00m19

Now listing the images again:

$ mcumgr -c acm0 image list
Images:
 image=0 slot=0
  version: 1.0.0
  bootable: true
  flags: active confirmed
  hash: 86dca73a3439112b310b5e033d811ec2df728d2264265f2046fced5a9ed00cc7
 image=0 slot=1
  version: 1.1.0
  bootable: true
  flags:
  hash: e8cf0dcef3ec8addee07e8c4d5dc89e64ba3fae46a2c5267fc4efbea4ca0e9f4
Split status: N/A (0)

To test a new upgrade image the test command is used:

mcumgr <connection-options> image test <hash>

This command should mark a test upgrade, which means that after the next reboot the bootloader will execute the upgrade and jump into the new image. If no other image operations are executed on the newly running image, it will revert back to the image that was previously running on the device on the subsequent reset. When a test is requested, flags will be updated with pending to inform that a new image will be run after a reset:

$ mcumgr -c acm0 image test e8cf0dcef3ec8addee07e8c4d5dc89e64ba3fae46a2c5267fc4efbea4ca0e9f4
Images:
 image=0 slot=0
  version: 1.0.0
  bootable: true
  flags: active confirmed
  hash: 86dca73a3439112b310b5e033d811ec2df728d2264265f2046fced5a9ed00cc7
 image=0 slot=1
  version: 1.1.0
  bootable: true
  flags: pending
  hash: e8cf0dcef3ec8addee07e8c4d5dc89e64ba3fae46a2c5267fc4efbea4ca0e9f4
Split status: N/A (0)

After a reset the output with change to:

$ mcumgr -c acm0 image list
Images:
 image=0 slot=0
  version: 1.1.0
  bootable: true
  flags: active
  hash: e8cf0dcef3ec8addee07e8c4d5dc89e64ba3fae46a2c5267fc4efbea4ca0e9f4
 image=0 slot=1
  version: 1.0.0
  bootable: true
  flags: confirmed
  hash: 86dca73a3439112b310b5e033d811ec2df728d2264265f2046fced5a9ed00cc7
Split status: N/A (0)

Tip

It’s important to mention that an upgrade only ever happens if the image is valid. The first thing MCUboot does when an upgrade is requested is to validate the image, using the SHA-256 and/or the signature (depending on the configuration). So before uploading an image, one way to be sure it is valid is to run imgtool verify -k <your-signature-key> <your-image>, where -k <your-signature-key can be skipped if no signature validation was enabled.

The confirmed flag in the secondary slot tells that after the next reset a revert upgrade will be performed to switch back to the original layout.

The confirm command used to confirm that an image is OK and no revert should happen (empty hash required) is:

mcumgr <connection-options> image confirm ""

The confirm command can also be run passing in a hash so that instead of doing a test/revert procedure, the image in the secondary partition is directly upgraded to, eg:

mcumgr <connection-options> image confirm <hash>

Tip

The whole test/revert cycle does not need to be done using only the mcumgr command-line tool. A better alternative is to perform a test and allow the new running image to self-confirm after any checks by calling boot_write_img_confirmed().

Tip

Building with CONFIG_MCUMGR_GRP_IMG_VERBOSE_ERR enables better error messages when failures happen (but increases the application size).

Statistics Management

Statistics are used for troubleshooting, maintenance, and usage monitoring; it consists basically of user-defined counters which are tightly connected to mcumgr and can be used to track any information for easy retrieval. The available sub-commands are:

mcumgr <connection-options> stat list
mcumgr <connection-options> stat <section-name>

Statistics are organized in sections (also called groups), and each section can be individually queried. Defining new statistics sections is done by using macros available under zephyr/stats/stats.h. Each section consists of multiple variables (or counters), all with the same size (16, 32 or 64 bits).

To create a new section my_stats:

STATS_SECT_START(my_stats)
  STATS_SECT_ENTRY(my_stat_counter1)
  STATS_SECT_ENTRY(my_stat_counter2)
  STATS_SECT_ENTRY(my_stat_counter3)
STATS_SECT_END;

STATS_SECT_DECL(my_stats) my_stats;

Each entry can be declared with STATS_SECT_ENTRY (or the equivalent STATS_SECT_ENTRY32), STATS_SECT_ENTRY16 or STATS_SECT_ENTRY64. All statistics in a section must be declared with the same size.

The statistics counters can either have names or not, depending on the setting of the CONFIG_STATS_NAMES option. Using names requires an extra declaration step:

STATS_NAME_START(my_stats)
  STATS_NAME(my_stats, my_stat_counter1)
  STATS_NAME(my_stats, my_stat_counter2)
  STATS_NAME(my_stats, my_stat_counter3)
STATS_NAME_END(my_stats);

Tip

Disabling CONFIG_STATS_NAMES will free resources. When this option is disabled the STATS_NAME* macros output nothing, so adding them in the code does not increase the binary size.

Tip

CONFIG_MCUMGR_GRP_STAT_MAX_NAME_LEN sets the maximum length of a section name that can can be accepted as parameter for showing the section data, and might require tweaking for long section names.

The final steps to use a statistics section is to initialize and register it:

rc = STATS_INIT_AND_REG(my_stats, STATS_SIZE_32, "my_stats");
assert (rc == 0);

In the running code a statistics counter can be incremented by 1 using STATS_INC, by N using STATS_INCN or reset with STATS_CLEAR.

Let’s suppose we want to increment those counters by 1, 2 and 3 every second. To get a list of stats:

$ mcumgr --conn acm0 stat list
stat groups:
  my_stats

To get the current value of the counters in my_stats:

$ mcumgr --conn acm0 stat my_stats
stat group: my_stats
      13 my_stat_counter1
      26 my_stat_counter2
      39 my_stat_counter3

$ mcumgr --conn acm0 stat my_stats
stat group: my_stats
      16 my_stat_counter1
      32 my_stat_counter2
      48 my_stat_counter3

When CONFIG_STATS_NAMES is disabled the output will look like this:

$ mcumgr --conn acm0 stat my_stats
stat group: my_stats
       8 s0
      16 s1
      24 s2

Filesystem Management

The filesystem module is disabled by default due to security concerns: because of a lack of access control by default, every file in the FS will be accessible, including secrets, etc. To enable it CONFIG_MCUMGR_GRP_FS must be set (y). Once enabled the following sub-commands can be used:

mcumgr <connection-options> fs download <remote-file> <local-file>
mcumgr <connection-options> fs upload <local-file> <remote-file>

Using the fs command, requires CONFIG_FILE_SYSTEM to be enabled, and that some particular filesystem is enabled and properly mounted by the running application, eg for littlefs this would mean enabling CONFIG_FILE_SYSTEM_LITTLEFS, defining a storage partition Flash map and mounting the filesystem in the startup (fs_mount()).

Uploading a new file to a littlefs storage, mounted under /lfs, can be done with:

$ mcumgr -c acm0 fs upload foo.txt /lfs/foo.txt
25
Done

Where 25 is the size of the file.

For downloading a file, let’s first use the fs command (CONFIG_FILE_SYSTEM_SHELL must be enabled) in a remote shell to create a new file:

uart:~$ fs write /lfs/bar.txt 41 42 43 44 31 32 33 34 0a
uart:~$ fs read /lfs/bar.txt
File size: 9
00000000  41 42 43 44 31 32 33 34 0A                       ABCD1234.

Now it can be downloaded using:

$ mcumgr -c acm0 fs download /lfs/bar.txt bar.txt
0
9
Done
$ cat bar.txt
ABCD1234

Where 0 is the return code, and 9 is the size of the file.

Warning

The commands might exhaust the system workqueue, if its size is not large enough, so increasing CONFIG_SYSTEM_WORKQUEUE_STACK_SIZE might be required for correct behavior.

The size of the stack allocated buffer used to store the blocks, while transferring a file can be adjusted with CONFIG_MCUMGR_GRP_FS_DL_CHUNK_SIZE; this allows saving RAM resources.

Tip

CONFIG_MCUMGR_GRP_FS_PATH_LEN sets the maximum PATH accepted for a file name. It might require tweaking for longer file names.

Note

To add security to the filesystem management group, callbacks for MCUmgr hooks can be registered by a user application when the upload/download functions are ran which allows the application to control if access to a file is allowed or denied. See the MCUmgr Callbacks section for details.

Bootloader Integration

The Device Firmware Upgrade subsystem integrates the management subsystem with the bootloader, providing the ability to send and upgrade a Zephyr image to a device.

Currently only the MCUboot bootloader is supported. See MCUboot for more information.

Discord channel

Developers welcome!

Discord mcumgr channel: https://discord.com/invite/Ck7jw53nU2

API Reference

MCUmgr mgmt API