ARM: Review READMEs.

This commit is contained in:
Kelvin Lawson
2013-03-18 23:34:38 +00:00
parent 8ecb72c755
commit 65edc4ab09
2 changed files with 57 additions and 50 deletions

View File

@@ -14,15 +14,18 @@ architectures.
The same common/core port can be used on a wide variety of ARM devices
but platform-specific code has been separated out into multiple "platform"
folders. This allows the common ARM port to be shared among several
different ARM platforms and boards. For example different ARM devices and
(BSP) folders. This allows the common ARM code to be shared among several
different ARM platforms and boards. For example, different ARM devices and
platforms might use different interrupt controllers, timer subsystems and
UARTs.
UARTs but share the same core OS context-switching routines etc.
An example platform is in the "platforms/qemu_integratorcp" folder. The
Makefile for all ARM ports is in the platform-specific folders, so you
may wish to head straight there for full instructions on how to build
and use Atomthreads on a particular ARM platform.
platform-specific folders such as this contain the Makefile to build the
project for that platform, so you may wish to head straight there if you
wish to quickly get started building and running Atomthreads on ARM. The
qemu_integratorcp platform is designed for the Integrator/CP platform with
ARM926EJ-S processor, and can be run within QEMU for quick evaluation of
Atomthreads without real hardware.
---------------------------------------------------------------------------
@@ -31,7 +34,7 @@ FILES IN THE COMMON ARM PORT FOLDER
* tests-main.c: Contains a sample Atomthreads application starting at
main() that initialises the operating system and runs the automated test
suite applications. You will make your own main() function or similar
suite applications. You will normally make your own main() function
suitable for your application, possibly using this as a basis.
* atomport-asm.s: Contains the main assembler code that forms the portion
of the core ARM architecture port that must be implemented in assembler
@@ -43,13 +46,14 @@ FILES IN THE COMMON ARM PORT FOLDER
functions typically required if you want to do anything with stdio
(e.g. printf() calls etc). This is a very simple implementation that
always writes to the UART regardless of which file is "opened". Use of
printf() and friends with GCC toolchains typically requires heap, which
is supported via the _sbrk() function in here. Your linker script should
specify where the heap starts and stops using "end" and "heap_top"
definitions respectively. Note that in QEMU environments this may not be
required as some "semi-hosted" toolchains implement these functions and
the UART driver for you. In that case these functions will be ignored
because they are defined with weak linkage.
printf() and friends with GCC toolchains typically requires a heap, and
thus a heap is supported in this file via the _sbrk() function. Your
linker script should specify where the heap starts and stops using "end"
and "heap_top" definitions respectively. Note that in QEMU environments
this may not be required as some "semi-hosted" toolchains implement
these functions and the UART driver for you. In that case these
functions will be left out of the build because they are defined with
weak linkage.
---------------------------------------------------------------------------
@@ -60,24 +64,24 @@ To port Atomthreads to your ARM platform you must provide the following
functionality in your platform folder (in all cases example filenames are
from the qemu_integratorcp sample platform):
* Startup code: see Reset_Handler in startup.s. Typically this will be
at least initially some assembly code, and will set up the initial
stack pointer, perform any copying of segments from flash to RAM,
zero the BSS section etc, before branching to the main application
* Startup code: see Reset_Handler in startup.s. Typically this will be (at
least in the first few instructions) some assembly code, and will set up
the initial stack pointer, perform any copying of segments from flash to
RAM, zero the BSS section etc, before branching to the main application
function (e.g. main()). At some point during initialisation the timer
tick interrupt required by the OS should be started (100 ticks per
second) and this might be done here in the very early startup code.
Note that some compiler toolchains will provide a portion of the C
startup code e.g. the function _mainCRTStartup() provided by some
GCC toolchains.
second) and this might be done here in the very early startup code. Note
that some compiler toolchains will provide a portion of the C startup
code e.g. the function _mainCRTStartup() provided by some GCC
toolchains.
* Interrupt vector table: see __interrupt_vector_table in startup.s.
Typically this will contain at least an entry for the startup/reset
code, and an entry for the hardware IRQ handler. In order to share
common code amongst all platforms, the hardware IRQ handler
(archIRQHandler()) is actually implemented in the common ARM port
folder but calls back out to a dispatcher function in your platform
folder to determine the source of the interrupt and handle
it appropriately. Your platform folder should contain this dispatcher
folder to determine the source of the interrupt and handle it
appropriately. Your platform folder should contain this dispatcher
function (__interrupt_dispatcher). It must support at least the timer
interrupt service routine (atomTimerTick()) required by the OS. The
dispatcher also handles other hardware interrupt sources; it determines
@@ -86,17 +90,18 @@ from the qemu_integratorcp sample platform):
* Linker script: Here you should specify the location of the interrupt
vector table within RAM, location of code/text segment etc. The
Atomthreads ARM port does not dictate particular section names or
layout unless your toolchain does not provide a suitable syscalls.c
and you instead get the heap _sbrk() implementation from Atomthreads'
sample syscalls.c which expects "heap_top" and "end" (heap_base) to
be defined by the linker script (which can be easily changed in
ports/arm/syscalls.c if necessary).
layout unless your toolchain does not provide a suitable syscalls.c and
you wish to use heap. In that case you will (at least initially) be
using the heap implementation _sbrk() in syscalls.c in the common ARM
port which expects "heap_top" and "end" (heap_base) to be defined by the
linker script. These names can be easily changed in ports/arm/syscalls.c
if necessary.
* UART driver: You should provide at least a UART write routine If you
would like to see debug statements, for example to see the results of
running the automated test suite. See uart.c for a simple example.
Note that for QEMU targets some semihosted toolchains will implement
this for you, in which case you won't need either
ports/arm/syscalls.c or a UART driver.
running the automated test suite. See uart.c for a simple example. Note
that for QEMU targets some semihosted toolchains will implement this for
you, in which case you won't need either ports/arm/syscalls.c or a UART
driver.
---------------------------------------------------------------------------

View File

@@ -16,10 +16,11 @@ platform running under QEMU.
All of the cross-platform kernel code is contained in the top-level
'kernel' folder, while ports to specific CPU architectures are contained in
the 'ports' folder tree. To support multiple ARM boards/platforms using a
single common ARM architecture port, the ARM port contains platform
single common ARM architecture port, the ARM port contains 'platform'
sub-folders in which the board/platform-specific code is situated. This
allows many different ARM boards with different interrupt controllers, UARTs
etc but which all reuse the same common core ARM context-switching code.
allows the sharing of common ARM port code between many different ARM
boards with different interrupt controllers, UARTs etc but which all reuse
the same common core ARM context-switching code.
This platform contains a few key platform-specific files:
@@ -63,7 +64,7 @@ The port works out-of-the-box with the GCC tools (for building) and QEMU
was tested using the CodeSourcery toolchain (2009q3 non-Linux but others
should be supported) and self-built toolchains such as hosted toolchains
built by build_arm_toolchain.sh (see http://navaro.nl for details). Note
that the Makefile for this platform assumes your GCC binary is named
that the Makefile for this platform assumes that your GCC binary is named
"arm-none-eabi-gcc".
Currently we assume that the toolchain will provide some header files like
@@ -91,8 +92,8 @@ OTHER PREREQUISITES
QEMU is used for simulation of the target and versions 0.14.1, 1.2.0 &
1.4.0 were tested.
Running the entire automated test suite via "make qemutests" also requires
the "expect" program.
Running the entire automated test suite in one command via "make qemutests"
also requires the "expect" program.
---------------------------------------------------------------------------
@@ -108,9 +109,9 @@ carry out the full build using the following:
All objects are built into the 'build' folder under
ports/arm/platforms/qemu_integrator_cp. The build process builds separate
target applications for each automated test, and appropriate ELF files can be
found in the build folder ready for running on the target or within QEMU.
Each test is built and run as a separate application.
target applications for each automated test, and appropriate ELF files can
be found in the build folder ready for running on the target or within
QEMU. Each test is built and run as a separate application.
All built objects etc can be cleaned using:
@@ -126,7 +127,7 @@ both the kernel and port documentation from this folder using:
---------------------------------------------------------------------------
IntegratorCP SPECIFICS
Integrator/CP SPECIFICS
The test applications make use of the Integrator's UART to print out
pass/fail indications and other information. For this you should connect a
@@ -147,9 +148,10 @@ these tests is built as an independent application in the 'build' folder.
These can be run on the target or within QEMU using the instructions below.
To view the test results, connect a serial debug cable to your target
platform or view the console if using QEMU. On starting, the test applications
print out "Go" on the UART. Once the test is complete they will print
out "Pass" or "Fail", along with other information if the test failed.
platform or view the console if using QEMU. On starting, the test
applications print out "Go" on the UART. Once the test is complete they
will print out "Pass" or "Fail", along with other information if the test
failed.
Most of the tests complete within a few seconds, but some (particularly
the stress tests) can take longer, so be patient.
@@ -169,11 +171,11 @@ RUNNING TESTS WITHIN THE QEMU SIMULATOR
It is possible to run the full automated test suite in a simulator without
programming the test applications into real hardware. This is very useful
for quick verification of the entire test suite after making any software
changes, and is much faster than download each test application to a real
target.
changes, and is much faster than downloading each test application to a
real target.
A single command runs every single test application, and checks the
(simulated) UART output to verify that each test case passes.
A single command runs every single test application, and automatically
parses the (simulated) UART output to verify that each test case passes.
This requires two applications on your development PC: expect and QEMU.