Running applications with Trusted Firmware-M
On nRF5340 and nRF9160, Trusted Firmware-M (TF-M) is used to configure and boot an application as non-secure.
TF-M is the reference implementation of Platform Security Architecture (PSA).
It provides a highly configurable set of software components to create a Trusted Execution Environment. This is achieved by a set of secure run time services such as Secure Storage, Cryptography, Audit Logs, and Attestation. Additionally, secure boot through MCUboot in TF-M ensures integrity of runtime software and supports firmware upgrade.
For official documentation, see the TF-M documentation.
The TF-M implementation in nRF Connect SDK is currently demonstrated in the following samples:
TF-M is one of the images that are built as part of a multi-image application. For more information about multi-image builds, see Multi-image builds.
To add TF-M to your build, enable the
CONFIG_BUILD_WITH_TFM configuration option by adding it to your
If you use menuconfig to enable
CONFIG_BUILD_WITH_TFM, you must also enable its dependencies.
By default, TF-M is configured to build the minimal version.
To use the full TF-M, you must disable the
You must build TF-M using a non-secure build target. The following platforms are currently supported:
TF-M uses UART1 for logging from the secure application.
To disable logging, enable the
When building TF-M with logging enabled, UART1 must be disabled in the non-secure application, otherwise the non-secure application will fail to run.
The recommended way to do this is to copy the .overlay file from the TF-M Hello World sample.
When using the Nordic Security Module, if
CONFIG_BUILD_WITH_TFM is enabled together with
CONFIG_NORDIC_SECURITY_BACKEND, the TF-M secure image will enable the use of the hardware acceleration of Arm CryptoCell.
In such case, the Kconfig configurations in the Nordic Security Backend control the features enabled through TF-M.
You can configure what crypto modules to include in TF-M by using the
TFM_CRYPTO_ Kconfig options found in file
TF-M utilizes hardware unique keys when the PSA Crypto key derivation APIs are used, and
psa_key_derivation_setup is called with the algorithm
For more information about the PSA cryptography and the API, see PSA Cryptography API 1.0.1.
The default configuration of TF-M has all supported features enabled, which results in a significant memory footprint. A minimal version of the TF-M secure application is provided in nRF Connect SDK to show how to configure a reduced version of TF-M.
The secure services supported by this minimal version allow for generating random numbers, and the platform services.
The minimal version of TF-M is disabled by setting the
CONFIG_TFM_PROFILE_TYPE_NOT_SET option or one of the other build profiles.
CONFIG_TFM_PROFILE_TYPE_MINIMAL is set, the configurability of TF-M is severely limited.
Hence, it is not possible to modify the TF-M minimal configuration to create your own variant of the minimal configuration.
Instead, the default configuration must be used as a starting point.
The procedure for programming an application with TF-M is the same as for other multi-image applications in nRF Connect SDK.
After building the application, a
merged.hex file is created that contains MCUboot, TF-M, and the application.
merged.hex file can be then programmed using nRF Connect for Visual Studio Code.
When using the command line, the file is programmed automatically when you call
ninja flash or
TF-M employs two UART interfaces for logging: one for the secure part (MCUboot and TF-M), and one for the non-secure application. The logs arrive on different COM ports on the host PC.
On the nRF5340 DK v1.0.0, you must connect specific wires on the kit to receive secure logs on the host PC. Specifically, wire the pins P0.25 and P0.26 of the P2 connector to RxD and TxD of the P24 connector respectively. See Getting logging output on the Working with nRF5340 DK page for more information.
On the nRF5340 DK v2.0.0, there are only two virtual COM ports available. By default, one of the ports is used by the non-secure UART0 peripheral from the application and the other by the UART1 peripheral from the network core.
There are several options to get UART output from the secure TF-M:
Disable the output for the network core and change the pins used by TF-M. The network core will usually have an nRF Connect SDK child image. To configure a child image, see Configuration of the child image section described in Multi-image builds. To configure logging in an nRF Connect SDK image, see Logging in nRF Connect SDK. To change the pins used by TF-M, the RXD (
CONFIG_TFM_UART1_RXD_PIN) and TXD (
CONFIG_TFM_UART1_TXD_PIN) Kconfig options in the application image can be set to P1.00 (32) and P1.01 (33).
The secure and non-secure UART peripherals can be wired to the same pins. Specifically, physically wire together the pins P0.25 and P0.26 to P0.20 and P0.22, respectively.
If the non-secure application, network core and TF-M outputs are all needed simultaneously, additional UART <-> USB hardware is needed. A second nRF DK can be used if available. Pin P0.25 needs to be wired to the TXD pin, and P0.26 to the RXD pin of the external hardware. These pins will provide the secure TF-M output, while the two native COM ports of the DK will be used for the non-secure application and the network core output.
The following limitations apply to TF-M and its usage:
Firmware Update service is not supported.
The following crypto modules or ciphers are not supported:
AES output feedback (AES-OFB) mode.
AES cipher feedback (AES-CFB) mode.
Isolation level 3 is not supported.
In Isolation level 2 or higher, the number of peripherals configured as secure in Application Root of Trust (ARoT) is limited by the number of available MPU regions.
Nordic Semiconductor devices only support the GCC toolchain for building TF-M.
TF-M requires that secure and non-secure partition addresses must be aligned to the NRF_SPU flash region size
nRF Connect SDK ensures that they in fact are aligned and comply with the TF-M requirements.
TF-M requires this alignment because it uses the SPU to enforce the security policy between the partitions. When the Partition Manager is enabled, it will take into consideration the alignment requirements. But when the static partitions are used, the user is responsible for following the alignment requirements.
If you are experiencing any partition alignment issues when using the Partition Manager, check the Known issues page on the main branch.
The partitions which need to be aligned to the SPU flash region size are partitions
Both the partition start address and the partition size need to be aligned with the NRF_SPU flash region size
Note that the
tfm_nonsecure partition is placed after the
tfm_secure partition, thus the end address of the
tfm_secure partition is the same as the start address of the
As a result, altering the size of the
tfm_secure partition affects the start address of the
The following static partition snippet shows a non-aligned configuration for nRF5340 which has a SPU flash region size
CONFIG_NRF_SPU_FLASH_REGION_SIZE of 0x4000.
span: [mcuboot_pad, tfm]
In the above example, the
tfm_nonsecure partition starts at address 0x8200, which is not aligned with the SPU requirement of 0x4000.
tfm_secure spans the
tfm partitions we can decrease the size of any of them by 0x200 to fix the alignment issue.
We will decrease the size of the (optional)
mcuboot_pad partition and thus the size of the
tfm_secure partition as follows:
span: [mcuboot_pad, tfm]
The interface to TF-M is different from the interface to SPM. Due to that, the application code that uses the SPM Secure Services needs to be ported to use TF-M instead.
TF-M can replace the following SPM services:
The following SPM services have no replacement in TF-M:
By default, TF-M configures memory regions as secure memory, while SPM configures memory regions as non-secure.
nonsecure_storage are configured as non-secure flash memory regions.
sram_nonsecure is configured as a non-secure RAM region.
If a static partition file is used for the application, make the following changes:
Add a partition called
mcuboot_pad(if MCUboot is enabled) and
Add a partition called
tfm_nonsecurethat spans the application, and other possible application partitions that must be non-secure.
For non-secure storage partitions, place the partitions inside the