Crate cortex_m_rt[−][src]
Expand description
Startup code and minimal runtime for Cortex-M microcontrollers
This crate contains all the required parts to build a no_std
application (binary crate) that
targets a Cortex-M microcontroller.
Features
This crates takes care of:
-
The memory layout of the program. In particular, it populates the vector table so the device can boot correctly, and properly dispatch exceptions and interrupts.
-
Initializing
static
variables before the program entry point. -
Enabling the FPU before the program entry point if the target is
thumbv7em-none-eabihf
.
This crate also provides the following attributes:
#[entry]
to declare the entry point of the program#[exception]
to override an exception handler. If not overridden all exception handlers default to an infinite loop.#[pre_init]
to run code beforestatic
variables are initialized
This crate also implements a related attribute called #[interrupt]
, which allows you
to define interrupt handlers. However, since which interrupts are available depends on the
microcontroller in use, this attribute should be re-exported and used from a device crate.
The documentation for these attributes can be found in the Attribute Macros section.
Requirements
memory.x
This crate expects the user, or some other crate, to provide the memory layout of the target
device via a linker script named memory.x
. This section covers the contents of memory.x
MEMORY
The linker script must specify the memory available in the device as, at least, two MEMORY
regions: one named FLASH
and one named RAM
. The .text
and .rodata
sections of the
program will be placed in the FLASH
region, whereas the .bss
and .data
sections, as well
as the heap,will be placed in the RAM
region.
/* Linker script for the STM32F103C8T6 */ MEMORY { FLASH : ORIGIN = 0x08000000, LENGTH = 64K RAM : ORIGIN = 0x20000000, LENGTH = 20K }
_stack_start
This optional symbol can be used to indicate where the call stack of the program should be
placed. If this symbol is not used then the stack will be placed at the end of the RAM
region – the stack grows downwards towards smaller address. This symbol can be used to place
the stack in a different memory region, for example:
/* Linker script for the STM32F303VCT6 */ MEMORY { FLASH : ORIGIN = 0x08000000, LENGTH = 256K /* .bss, .data and the heap go in this region */ RAM : ORIGIN = 0x20000000, LENGTH = 40K /* Core coupled (faster) RAM dedicated to hold the stack */ CCRAM : ORIGIN = 0x10000000, LENGTH = 8K } _stack_start = ORIGIN(CCRAM) + LENGTH(CCRAM);
_stext
This optional symbol can be used to control where the .text
section is placed. If omitted the
.text
section will be placed right after the vector table, which is placed at the beginning of
FLASH
. Some devices store settings like Flash configuration right after the vector table;
for these devices one must place the .text
section after this configuration section –
_stext
can be used for this purpose.
MEMORY { /* .. */ } /* The device stores Flash configuration in 0x400-0x40C so we place .text after that */ _stext = ORIGIN(FLASH) + 0x40C
An example
This section presents a minimal application built on top of cortex-m-rt
. Apart from the
mandatory memory.x
linker script describing the memory layout of the device, the hard fault
handler and the default exception handler must also be defined somewhere in the dependency
graph (see [#[exception]
]). In this example we define them in the binary crate:
// IMPORTANT the standard `main` interface is not used because it requires nightly #![no_main] #![no_std] extern crate cortex_m_rt as rt; // makes `panic!` print messages to the host stderr using semihosting extern crate panic_semihosting; use rt::entry; // use `main` as the entry point of this application // `main` is not allowed to return #[entry] fn main() -> ! { // initialization loop { // application logic } }
To actually build this program you need to place a memory.x
linker script somewhere the linker
can find it, e.g. in the current directory; and then link the program using cortex-m-rt
’s
linker script: link.x
. The required steps are shown below:
$ cat > memory.x <<EOF /* Linker script for the STM32F103C8T6 */ MEMORY { FLASH : ORIGIN = 0x08000000, LENGTH = 64K RAM : ORIGIN = 0x20000000, LENGTH = 20K } EOF $ cargo rustc --target thumbv7m-none-eabi -- \ -C link-arg=-nostartfiles -C link-arg=-Tlink.x $ file target/thumbv7m-none-eabi/debug/app app: ELF 32-bit LSB executable, ARM, EABI5 version 1 (SYSV), statically linked, (..)
Optional features
device
If this feature is disabled then this crate populates the whole vector table. All the interrupts
in the vector table, even the ones unused by the target device, will be bound to the default
exception handler. This makes the final application device agnostic: you will be able to run it
on any Cortex-M device – provided that you correctly specified its memory layout in memory.x
– without hitting undefined behavior.
If this feature is enabled then the interrupts section of the vector table is left unpopulated
and some other crate, or the user, will have to populate it. This mode is meant to be used in
conjunction with crates generated using svd2rust
. Those device crates will populate the
missing part of the vector table when their "rt"
feature is enabled.
Inspection
This section covers how to inspect a binary that builds on top of cortex-m-rt
.
Sections (size
)
cortex-m-rt
uses standard sections like .text
, .rodata
, .bss
and .data
as one would
expect. cortex-m-rt
separates the vector table in its own section, named .vector_table
. This
lets you distinguish how much space is taking the vector table in Flash vs how much is being
used by actual instructions (.text
) and constants (.rodata
).
$ size -Ax target/thumbv7m-none-eabi/examples/app target/thumbv7m-none-eabi/release/examples/app : section size addr .vector_table 0x400 0x8000000 .text 0x88 0x8000400 .rodata 0x0 0x8000488 .data 0x0 0x20000000 .bss 0x0 0x20000000
Without the -A
argument size
reports the sum of the sizes of .text
, .rodata
and
.vector_table
under “text”.
$ size target/thumbv7m-none-eabi/examples/app text data bss dec hex filename 1160 0 0 1660 67c target/thumbv7m-none-eabi/release/app
Symbols (objdump
, nm
)
One will always find the following (unmangled) symbols in cortex-m-rt
applications:
-
Reset
. This is the reset handler. The microcontroller will executed this function upon booting. This function will call the user program entry point (cf.#[entry]
) using themain
symbol so you may also find that symbol in your program; if you do,main
will contain your application code. Some other timesmain
gets inlined intoReset
so you won’t find it. -
DefaultHandler
. This is the default handler. If not overridden using#[exception] fn DefaultHandler(..
this will be an infinite loop. -
HardFaultTrampoline
. This is the real hard fault handler. This function is simply a trampoline that jumps into the user defined hard fault handler namedHardFault
. The trampoline is required to set up the pointer to the stacked exception frame. -
HardFault
. This is the user defined hard fault handler. If not overridden using#[exception] fn HardFault(..
it will default to an infinite loop. -
__STACK_START
. This is the first entry in the.vector_table
section. This symbol contains the initial value of the stack pointer; this is where the stack will be located – the stack grows downwards towards smaller addresses. -
__RESET_VECTOR
. This is the reset vector, a pointer into theReset
handler. This vector is located in the.vector_table
section after__STACK_START
. -
__EXCEPTIONS
. This is the core exceptions portion of the vector table; it’s an array of 14 exception vectors, which includes exceptions likeHardFault
andSysTick
. This array is located after__RESET_VECTOR
in the.vector_table
section. -
__INTERRUPTS
. This is the device specific interrupt portion of the vector table; its exact size depends on the target device but if the"device"
feature has not been enabled it will have a size of 32 vectors (on ARMv6-M) or 240 vectors (on ARMv7-M). This array is located after__EXCEPTIONS
in the.vector_table
section. -
__pre_init
. This is a function to be run before RAM is initialized. It defaults to an empty function. The function called can be changed by applying the#[pre_init]
attribute to a function. The empty function is not optimized out by default, but if an empty function is passed to#[pre_init]
the function call will be optimized out.
If you override any exception handler you’ll find it as an unmangled symbol, e.g. SysTick
or
SVCall
, in the output of objdump
,
If you are targeting the thumbv7em-none-eabihf
target you’ll also see a ResetTrampoline
symbol in the output. To avoid the compiler placing FPU instructions before the FPU has been
enabled (cf. vpush
) Reset
calls the function ResetTrampoline
which is marked as
#[inline(never)]
and ResetTrampoline
calls main
. The compiler is free to inline main
into ResetTrampoline
but it can’t inline ResetTrampoline
into Reset
– the FPU is enabled
in Reset
.
Advanced usage
Setting the program entry point
This section describes how #[entry]
is implemented. This information is useful
to developers who want to provide an alternative to #[entry]
that provides extra
guarantees.
The Reset
handler will call a symbol named main
(unmangled) after initializing .bss
and
.data
, and enabling the FPU (if the target is thumbv7em-none-eabihf
). A function with the
entry
attribute will be set to have the export name “main
”; in addition, its mutable
statics are turned into safe mutable references (see #[entry]
for details).
The unmangled main
symbol must have signature extern "C" fn() -> !
or its invocation from
Reset
will result in undefined behavior.
Incorporating device specific interrupts
This section covers how an external crate can insert device specific interrupt handlers into the
vector table. Most users don’t need to concern themselves with these details, but if you are
interested in how device crates generated using svd2rust
integrate with cortex-m-rt
read on.
The information in this section applies when the "device"
feature has been enabled.
__INTERRUPTS
The external crate must provide the interrupts portion of the vector table via a static
variable named__INTERRUPTS
(unmangled) that must be placed in the .vector_table.interrupts
section of its object file.
This static
variable will be placed at ORIGIN(FLASH) + 0x40
. This address corresponds to the
spot where IRQ0 (IRQ number 0) is located.
To conform to the Cortex-M ABI __INTERRUPTS
must be an array of function pointers; some spots
in this array may need to be set to 0 if they are marked as reserved in the data sheet /
reference manual. We recommend using a union
to set the reserved spots to 0
; None
(Option<fn()>
) may also work but it’s not guaranteed that the None
variant will always be
represented by the value 0
.
Let’s illustrate with an artificial example where a device only has two interrupt: Foo
, with
IRQ number = 2, and Bar
, with IRQ number = 4.
union Vector { handler: extern "C" fn(), reserved: usize, } extern "C" { fn Foo(); fn Bar(); } #[link_section = ".vector_table.interrupts"] #[no_mangle] pub static __INTERRUPTS: [Vector; 5] = [ // 0-1: Reserved Vector { reserved: 0 }, Vector { reserved: 0 }, // 2: Foo Vector { handler: Foo }, // 3: Reserved Vector { reserved: 0 }, // 4: Bar Vector { handler: Bar }, ];
device.x
Linking in __INTERRUPTS
creates a bunch of undefined references. If the user doesn’t set a
handler for all the device specific interrupts then linking will fail with "undefined reference"
errors.
We want to provide a default handler for all the interrupts while still letting the user
individually override each interrupt handler. In C projects, this is usually accomplished using
weak aliases declared in external assembly files. In Rust, we could achieve something similar
using global_asm!
, but that’s an unstable feature.
A solution that doesn’t require global_asm!
or external assembly files is to use the PROVIDE
command in a linker script to create the weak aliases. This is the approach that cortex-m-rt
uses; when the "device"
feature is enabled cortex-m-rt
’s linker script (link.x
) depends on
a linker script named device.x
. The crate that provides __INTERRUPTS
must also provide this
file.
For our running example the device.x
linker script looks like this:
/* device.x */ PROVIDE(Foo = DefaultHandler); PROVIDE(Bar = DefaultHandler);
This weakly aliases both Foo
and Bar
. DefaultHandler
is the default exception handler and
that the core exceptions use unless overridden.
Because this linker script is provided by a dependency of the final application the dependency
must contain build script that puts device.x
somewhere the linker can find. An example of such
build script is shown below:
use std::env; use std::fs::File; use std::io::Write; use std::path::PathBuf; fn main() { // Put the linker script somewhere the linker can find it let out = &PathBuf::from(env::var_os("OUT_DIR").unwrap()); File::create(out.join("device.x")) .unwrap() .write_all(include_bytes!("device.x")) .unwrap(); println!("cargo:rustc-link-search={}", out.display()); }
Uninitialized static variables
The .uninit
linker section can be used to leave static mut
variables uninitialized. One use
case of unitialized static variables is to avoid zeroing large statically allocated buffers (say
to be used as thread stacks) – this can considerably reduce initialization time on devices that
operate at low frequencies.
The only correct way to use this section is by placing static mut
variables with type
MaybeUninit
in it.
use core::mem::MaybeUninit; const STACK_SIZE: usize = 8 * 1024; const NTHREADS: usize = 4; #[link_section = ".uninit.STACKS"] static mut STACKS: MaybeUninit<[[u8; STACK_SIZE]; NTHREADS]> = MaybeUninit::uninit();
Be very careful with the link_section
attribute because it’s easy to misuse in ways that cause
undefined behavior. At some point in the future we may add an attribute to safely place static
variables in this section.
Structs
Registers stacked (pushed onto the stack) during an exception.
Functions
Returns a pointer to the start of the heap