STM32 MPU Flash mapping

Applicable for

STM32MP13x lines, STM32MP15x lines, STM32MP21x lines, STM32MP23x lines, STM32MP25x lines

This article shows the mapping used in STM32 MPU Embedded Software distribution for ST boards, that can be used as starting point for other boards.

1. Supported Flash memory technologies[edit | edit source]

STM32MP1 series and STM32MP2 series support the different types of Flash memory. The bullets below show which interfaces are supported on STMicroelectronics boards:

SD card on the SDMMC interface

e•MMC on the SDMMC interface

NAND Flash memory on the FMC interface

used on STM32MP15 EVAL board only.

Serial NOR Flash memory on the Dual QSPI/OCTOSPI interface

used on STM32MP15 EVAL board
used on STM32MP25 EVAL board

Serial NAND Flash memory on the Dual QSPI/OCTOSPI interface

not present on ST boards

HyperFlash memory on the OCTOSPI interface

not present on ST boards

The next section lists all partitions used on STM32MPU boards (size, name, and content), and the following sections show how they are mapped on the different types of Flash memory.

2. Flash partitions[edit | edit source]

The tables below list the partitions defined for STMP32MP1 and STM32MP2 boards in Cortex-A main processor mode and gives typical sizes with OpenSTLinux, that can be tuned, depending on the targeted application needs.

Size	Component	Comment
Remaining area	userfs	The user file system contains user data and examples
768 Mbytes	rootfs	Linux root file system contains all user space binaries (executable, libraries, and so on), and kernel modules
16 Mbytes	vendorfs	This partition is preferred to the rootfs for third-party proprietary binaries, and ensures that they are not contaminated by any open source licence, such as GPL v3
64 Mbytes	bootfs or boot (NAND)	The boot file system contains: (option) the init RAM file system, which can be copied to the external RAM and used by Linux before mounting a fatter rootfs Linux kernel device tree (can be in a Flattened Image Tree - FIT) Linux kernel U-Boot image (can be in a Flattened Image Tree - FIT) For all flashes except for NOR: the boot loader splash screen image, displayed by U-Boot U-Boot distro config file extlinux.conf
2 * 256 Kbytes (*)	u-boot-env or uboot_config / uboot_config_r (NAND)	This partition is used to store the U-Boot environment. uboot_config_r is used for redundancy management with uboot_config, in NAND Flash.
(2 *) 4 Mbytes	fip	The TF-A firmware image package (FIP) is a binary file that encapsulates several binaries, and their certificates (optionally, for authentication), that will be loaded by TF-A BL2. OpenSTLinux FIP contains: the second stage boot loader (SSBL): U-Boot binary U-Boot device tree blob the OP-TEE binary the firmware configuration file used by TF-A BL2 There are two partitions, fip-a and fip-b, when A/B (seamless) FIP firmware update is activated in TF-A BL2.
256 Kbytes to 512 Kbytes (*)	metadata	Metadata are used by TF-A BL2 to localize the FIP version to use, when the FIP is subject to firmware update. Notes: The metadata are not used when the FIP firmware update is not activated in TF-A BL2; these partitions can be removed or left empty in this case. The metadata update is failsafe against power loss during the FIP update process with two copies.
2 * 256 Kbytes to 512 Kbytes (*)	fsbla	The first stage boot loader is Arm Trusted Firmware (TF-A). At least two copies are embedded (fsbla1, fsbla2). Notes: The FSBL-A payload is limited to 128 Kbytes on STM32MP13 and 247 Kbytes on STM32MP15 and STM32MP2. The TF-A BL2 update is failsafe against power loss during the process with two copies (see ROM code page for details).

(*): The partition size depends on the Flash technology, to be aligned to the block erase size of the Flash memory present on the board: NOR (256 Kbytes) / NAND (512 Kbytes).

Information

Some boards can be equipped with multiple Flash devices, like the STM32MP15 EVAL board, where all of the Flash devices can be programmed with STM32CubeProgrammer. However, caution must be taken for the serial NOR/NAND and SLC NAND because a static bootable MTD partitioning is defined in U-Boot default config (see MTD configuration for details), with the consequence that up to 6 Mbytes of space is lost at the beginning of each such device, even those which are not bootable.

3. SD card memory mapping[edit | edit source]

The SD card has to be partitioned with GPT format in order to be recognized by the STM32MP1 and STM32MP2 ROM code. The easiest way to achieve this is to use STM32CubeProgrammer.
The ROM code looks for the GPT entries whose name begins with "fsbl": fsbl1 and fsbl2 for example.
Note: The SD card can be unplugged from the board and inserted into a Linux host computer for direct partitioning with Linux utilities and access to the bootfs, rootfs and userfs partitions. The file system is Linux EXT4.

4. e•MMC memory mapping[edit | edit source]

The e•MMC embeds four physical partitions:

Boot area partition 1: one copy of the FSBL
Boot area partition 2: one copy of the FSBL
User data area: formatted with GPT partitioning and used to store all remaining partitions
Replay Protected Memory Block (RPMB): not shown in the figure below, since not involved in the current boot chain.

STM32CubeProgrammer has to be used to prepare the e•MMC with the layout shown below, and to populate each partition.

Information

The boot area partition used by the e•MMC boot sequence is selected via the EXT_CSD[179] register in the e•MMC. The STM32CubeProgrammer execution is concluded with the selection of the last written partition from the flashlayout file, typically partition 2. The other copy is never used as long as the user does not explicitly change the e•MMC EXT_CSD[179] register to select it.

5. NOR memory mapping[edit | edit source]

As NOR Flash memory is expensive, its size is usually limited to the minimum needed to store only the bootloaders. The system files (bootfs, rootfs and userfs) are usually stored in another Flash memory, such as the SD card in OpenSTLinux distribution.
STM32CubeProgrammer must be used to prepare the NOR Flash and the SD card with the layout shown below, and to populate each partition.

It is possible to use an e•MMC card or NAND as second-level Flash memory, rather than an SD card. This requires the following aspects to be changed:

The Flash memory layout, using STM32CubeProgrammer in order to write the rootfs and userfs to the targeted Flash memory
The Linux kernel parameters, using U-Boot, in order to indicate where the rootfs and userfs have to be mounted.

6. NAND memory mapping[edit | edit source]

STM32CubeProgrammer has to be used to prepare the NAND Flash memory with the layout shown below, and to populate each partition.

Warning

In the RAW/MTD area, the number of copies to embed and the number of blocks to reserve at the end of each partition, for future bad blocks replacement, are a critical part of NAND based product dimensioning !

The strategy has to take into account many parameters such as:

the binaries sizes,
the read accesses to those partitions during product life, that may generate read disturb effect,
the capability of the product to refresh the partitions content when erroneous bits are detected,
the number of software updates estimated on this partition,
the selected NAND flash characteristics

ST set foundations in the STM32MP1 and STM32MP2 devices in order to allow the integration of NAND flash memories but the product definition remains the customer responsibility. Please, contact your memory provider for further advice.