Crate cortex_m_quickstart [] [src]

A template for building applications for ARM Cortex-M microcontrollers


  • Nightly Rust toolchain newer than nightly-2018-04-08: rustup default nightly
  • Cargo clone subcommand: cargo install cargo-clone
  • ARM toolchain: sudo apt-get install gcc-arm-none-eabi (on Ubuntu)
  • GDB: sudo apt-get install gdb-arm-none-eabi (on Ubuntu)
  • OpenOCD: sudo apt-get install OpenOCD (on Ubuntu)
  • [Optional] Cargo add subcommand: cargo install cargo-edit


  1. Figure out the cross compilation target to use.

  • Use thumbv6m-none-eabi for ARM Cortex-M0 and Cortex-M0+

  • Use thumbv7m-none-eabi for ARM Cortex-M3

  • Use thumbv7em-none-eabi for ARM Cortex-M4 and Cortex-M7 (no FPU support)

  • Use thumbv7em-none-eabihf for ARM Cortex-M4F and Cortex-M7F (with FPU support)

  1. Install the rust-std component for your target, if you haven't done so already
$ rustup target add thumbv7em-none-eabihf
  1. Clone this crate
$ git clone cortex-m-quickstart --vers 0.3.0
  1. Change the crate name, author and version
$ edit Cargo.toml && head $_
authors = ["Jorge Aparicio <>"]
name = "demo"
version = "0.1.0"
  1. Specify the memory layout of the target device

NOTE board support crates sometimes provide this file for you (check the crate documentation). If you are using one that does then remove both memory.x and from the root of this crate.

$ cat >memory.x <<'EOF'
  /* NOTE K = KiBi = 1024 bytes */
  FLASH : ORIGIN = 0x08000000, LENGTH = 256K
  RAM : ORIGIN = 0x20000000, LENGTH = 40K
  1. Optionally, set a default build target. This way you don't have to pass --target to each Cargo invocation.
$ cat >>.cargo/config <<'EOF'
target = "thumbv7em-none-eabihf"
  1. Optionally, depend on a device, HAL implementation or a board support crate.
$ # add a device crate, OR
$ cargo add stm32f30x

$ # add a HAL implementation crate, OR
$ cargo add stm32f30x-hal

$ # add a board support crate
$ cargo add f3
  1. Write the application or start from one of the examples
$ rm -r src/* && cp examples/ src/
  1. Build the application
$ cargo build --release

$ # sanity check
$ arm-none-eabi-readelf -A target/thumbv7em-none-eabihf/release/demo
Attribute Section: aeabi
File Attributes
  Tag_conformance: "2.09"
  Tag_CPU_arch: v7E-M
  Tag_CPU_arch_profile: Microcontroller
  Tag_THUMB_ISA_use: Thumb-2
  Tag_FP_arch: VFPv4-D16
  Tag_ABI_PCS_GOT_use: direct
  Tag_ABI_FP_denormal: Needed
  Tag_ABI_FP_exceptions: Needed
  Tag_ABI_FP_number_model: IEEE 754
  Tag_ABI_align_needed: 8-byte
  Tag_ABI_align_preserved: 8-byte, except leaf SP
  Tag_ABI_HardFP_use: SP only
  Tag_ABI_VFP_args: VFP registers
  Tag_ABI_optimization_goals: Aggressive Speed
  Tag_CPU_unaligned_access: v6
  Tag_FP_HP_extension: Allowed
  Tag_ABI_FP_16bit_format: IEEE 754
  1. Flash and debug the program
$ # Launch OpenOCD on a terminal
$ openocd -f (..)
$ # Start a debug session in another terminal
$ arm-none-eabi-gdb target/thumbv7em-none-eabihf/release/demo

Alternatively, you can use cargo run to build, flash and debug the program in a single step.

$ cargo run --example hello
> # drops you into a GDB session


Check the examples module


This section contains fixes for common errors encountered when the cortex-m-quickstart template is misused.

Used the standard main interface

Error message:

$ cargo build
   Compiling demo v0.1.0 (file:///home/japaric/tmp/demo)

error: requires `start` lang_item

Solution: Use #![no_main] and entry! as shown in the examples.

Forgot to launch an OpenOCD instance

Error message:

$ arm-none-eabi-gdb target/..
Reading symbols from hello...done.
.gdbinit:1: Error in sourced command file:
:3333: Connection timed out.

Solution: Launch OpenOCD on other terminal. See Usage section.

Didn't modify the memory.x linker script

Error message:

$ cargo build
Compiling demo v0.1.0 (file:///home/japaric/tmp/demo)
error: linking with `arm-none-eabi-ld` failed: exit code: 1
= note: "arm-none-eabi-gcc" "-L" (..)
          (..)/ld: region `FLASH' overflowed by XXX bytes

Solution: Specify your device memory layout in the memory.x linker script. See Usage section.

Didn't set a default build target and forgot to pass --target to Cargo

Error message:

$ cargo build
error: language item required, but not found: `eh_personality`

error: aborting due to previous error

Solution: Set a default build target in the .cargo/config file (see Usage section), or call Cargo with --target flag: cargo build --target thumbv7em-none-eabi.

Overwrote the original .cargo/config file

Error message:

error: linking with `arm-none-eabi-gcc` failed: exit code: 1
  = note: (..)
(..)/crt0.o: In function `_start':
(.text+0x90): undefined reference to `memset'
(..)/crt0.o: In function `_start':
(.text+0xd0): undefined reference to `atexit'
(..)/crt0.o: In function `_start':
(.text+0xd4): undefined reference to `__libc_init_array'
(..)/crt0.o: In function `_start':
(.text+0xe4): undefined reference to `exit'
(..)/crt0.o: In function `_start':
(.text+0x100): undefined reference to `__libc_fini_array'
collect2: error: ld returned 1 exit status

Solution: You probably overwrote the original .cargo/config instead of appending the default build target (e.g. cat > instead of cat >>). The less error prone way to fix this is to remove the .cargo directory, clone a new copy of the template and then copy the .cargo directory from that fresh template into your current project. Don't forget to append the default build target to .cargo/config.

Called OpenOCD with wrong arguments

Error message:

$ openocd -f ..
Error: open failed
in procedure 'init'
in procedure 'ocd_bouncer'

Solution: Correct the OpenOCD arguments. Check the /usr/share/openocd/scripts directory (exact location varies per distribution / OS) for a list of scripts that can be used.

Forgot to install the rust-std component

Error message:

$ cargo build
error[E0463]: can't find crate for `core`
  = note: the `thumbv7m-none-eabi` target may not be installed

Solution: call rustup target add thumbv7m-none-eabi but with the name of your target

Used an old nightly

Error message:

$ cargo build
Compiling cortex-m-rt v0.2.0
error[E0463]: can't find crate for `core`
= note: the `thumbv7em-none-eabihf` target may not be installed

error: aborting due to previous error

Solution: Use a more recent nightly

Used the stable toolchain

Error message:

$ cargo build
error[E0463]: can't find crate for `core`
  = note: the `thumbv7em-none-eabihf` target may not be installed

Solution: We are not there yet! Switch to the nightly toolchain with rustup default nightly.

Used gdb instead of arm-none-eabi-gdb

Error message:

$ gdb target/..
Reading symbols from hello...done.
warning: Architecture rejected target-supplied description
warning: Cannot convert floating-point register value to ..
value has been optimized out
Cannot write the dashboard
Traceback (most recent call last):
File "<string>", line 353, in render
File "<string>", line 846, in lines
gdb.error: Frame is invalid.
0x00000000 in ?? ()
semihosting is enabled
Loading section .text, size 0xd88 lma 0x8000000
Start address 0x8000000, load size 3464
.gdbinit:6: Error in sourced command file:
Remote connection closed

Solution: Use arm-none-eabi-gdb target/..

Used a named piped for itm.fifo

Error message:

$ cargo run [--example ..]

Reading symbols from target/thumbv7em-none-eabihf/debug/cortex-m-quickstart...done.
cortex_m_rt::reset_handler ()
    at $REGISTRY/cortex-m-rt-0.3.12/src/
330     unsafe extern "C" fn reset_handler() -> ! {
semihosting is enabled
Ignoring packet error, continuing...
Ignoring packet error, continuing...

Note that when you reach this point OpenOCD will become unresponsive and you'll have to kill it and start a new OpenOCD process before you can invoke cargo run / start GDB.

Cause: You uncommented the monitor tpiu .. line in .gdbinit and are using a named pipe to receive the ITM data (i.e. you ran mkfifo itm.fifo). This error occurs when itmdump -f itm.fifo (or equivalent, e.g. cat itm.fifo) is not running.

Solution: Run itmdump -f itm.fifo (or equivalently cat itm.fifo) before invoking cargo run / starting GDB. Note that sometimes itmdump will exit when the GDB session ends. In that case you'll have to run itmdump before you start the next GDB session.

Alternative solution: Use a plain text file instead of a named pipe. In this scenario you omit the mkfifo itm.dump command. You can use itmdump's follow mode (-F) to get named pipe like output.



Examples sorted in increasing degree of complexity