The Freescale ColdFire family is a range of processors including the MCF5208 and the MCF5282. From a programmer's perspective these processors all share basically the same processor core, albeit with minor differences in the instruction set. They differ in areas like performance, on-chip peripherals and caches. Even when it comes to peripherals there is a lot of commonality. For example many but not all Coldfire processors use the same basic interrupt controller(s) as the MCF5282. Similarly the on-chip UARTs tend to use the same basic design although there are variations in the number of UARTs, the fifo sizes, and in certain details.
The MCFxxxx variant HAL package
CYGPKG_HAL_M68K_MCFxxxx provides support for
various features that are common to many but not all Coldfire
processors. This includes HAL diagnostics via an on-chip UART and
interrupt controller management for those processors which have
MCF5282-compatible controllers. The variant HAL complements the M68K
architectural HAL package. An eCos configuration should also include a
processor-specific HAL package such as
CYGPKG_HAL_M68K_MCF5272 to support the
chip-specific peripherals and cache details, and a platform HAL
package such as
CYGPKG_HAL_M68K_M5272C3 to support
board-level details like external memory chips. The processor or
platform HAL can override the functionality provided by the variant
The MCFxxxx variant HAL package should be loaded automatically when eCos is configured for appropriate target hardware. It should never be necessary to load this package explicitly. Unloading the package should only happen as a side effect of switching target hardware.
On most ColdFire platforms the variant HAL will provide the HAL
diagnostics support via one of the UARTs. Some platforms may provide
their own HAL diagnostics facility, for example output via an LCD. The
variant HAL diagnostics support is active if the processor or platform
interface. It is also active only in configurations which do not rely
on an underlying rom monitor such as RedBoot:
CYGSEM_HAL_USE_ROM_MONITOR is enabled then the
default diagnostics channel will automatically be inherited from
RedBoot. The variant HAL then provides a number of configuration
options related to diagnostics:
This selects the destination for HAL diagnostics. The number of UARTs available depends on the processor, and on any given board some of the UARTs may not be connected. Hence the variant HAL looks for configuration options
CYGHWR_HAL_M68K_MCFxxxx_UART2to see which on-chip UARTs are actually available on the processor and target hardware, and uses this information to let the user select a UART.
Instead of using a uart the diagnostics support can be set to discard all output. This is useful when other packages or application code makes unconditional use of HAL diagnostics facilities, for example to report error conditions, but the target hardware does not have a spare UART. Alternatively when debugging via a hardware debug solution such as BDM it is possible to direct the diagnostics output to a gdb hwdebug file I/O channel. By default this will also discard diagnostics output. However if the application is running inside a gdb session and the gdb set hwdebug command has been used then the diagnostics will be output via gdb.
When a UART is in use as the HAL diagnostics channel, that UART should not be used for any other purpose. In particular application code should avoid using it for I/O via the serial driver.
When a UART is selected for HAL diagnostics this option specifies the default baud rate. The most common setting is 38400. That provides a compromise between performance and reliability, especially in electrically noisy environments such as an industrial environment or a test farm. Some platforms may define
CYGNUM_HAL_M68K_MCFxxxx_DIAGNOSTICS_DEFAULT_BAUDto handle scenarios where another default baud rate is preferable, typically for compatibility with existing software.
Usually the HAL diagnostics channel is driven in polled mode but in some scenarios interrupts are required. For example, when debugging an application over a serial line on top of the gdb stubs provided by RedBoot, the user should be able to interrupt the application with a control-C. The application will not be polling the HAL diagnostics UART at this point so instead the eCos interrupt management code interacts with the gdb stubs to do the right thing. This configuration option selects the interrupt priority. It should be noted that on some processors with MCF5282-compatible interrupt controllers all priorities for enabled interrupts should be unique, and it is the responsibility of application developers to ensure this condition is satisfied.
The HAL Port
This section describes how the MCFxxxx variant HAL package implements parts of the eCos HAL specification. It should be read in conjunction with similar sections from the architectural and processor HAL documentation.
The cyg/hal/var_io.h header provides various definitions for on-chip peripherals, where the current processor has peripherals compatible with the MCF5282's or which are available on several different coldfires. This header is automatically included by the architectural cyg/hal/hal_io.h so other packages and application code will usually only include the latter.
It is up to the processor HAL to specify exactly what var_io.h should export. For example the MCF5213's proc_io.h header contains the following:
# define HAL_MCFxxxx_HAS_MCF5282_INTC 1 # define HAL_MCFxxxx_INTC0_BASE (HAL_MCF521x_IPSBAR + 0x00000C00)
This enables support within the variant HAL for a single MCF5282-compatible interrupt controller, and cases var_io.h to export symbols such as:
#ifdef HAL_MCFxxxx_HAS_MCF5282_INTC // Two 32-bit interrupt mask registers # define HAL_MCFxxxx_INTCx_IMRH 0x0008 # define HAL_MCFxxxx_INTCx_IMRL 0x000C … # define HAL_MCFxxxx_INTCx_ICRxx_IL_MASK (0x07 << 3) # define HAL_MCFxxxx_INTCx_ICRxx_IL_SHIFT 3
Symbols such as
HAL_MCFxxxx_INTCx_IMRH can be used
to access the relevant hardware registers via
HAL_WRITE_UINT32. Symbols like
HAL_MCFxxxx_INTCx_ICRxx_IL_MASK can be used to
generate or decode the contents of the hardware registers.
The header file does mostly use a naming convention, but is not guaranteed to be totally consistent. There may also be discrepancies with the documentation because the manuals for the various Coldfire processors are not always consistent about their naming schemes. All I/O definitions provided by the variant HAL will start with HAL_MCFxxxx_, followed by the name of the peripheral. If a peripheral is likely to be a singleton, for example an on-chip flash unit, then the name is unadorned. If there may be several instances of the peripheral then the name will be followed by a lower case x. For example:
# define HAL_MCFxxxx_CFM_CR 0x0000 … # define HAL_MCFxxxx_UARTx_UMR 0x00
Register names will be relative to some base address such as
HAL_MCFxxxx_UART0_BASE, so code accessing a
register would look like:
HAL_READ_UINT32(HAL_MCFxxxx_CFM_BASE + HAL_MCFxxxx_CFM_PROT, reg); … HAL_WRITE_UINT8(base + HAL_MCFxxxx_UARTx_UTB, '*');
Usually the register names are singletons, but in some cases such as
the interrupt controller priority registers there may be multiple
instances of the register and the names will be suffixed
appropriately. For example
HAL_MCFxxxx_INTCx_ICRxx_IL_MASK indicates the field
IL within one of the ICR
registers within one of the interrupt controllers.
As mentioned earlier the processor HAL's proc_io.h will control which definitions are exported by var_io.h. Sometimes the processor HAL will then go on to undefine or redefine some of the symbols, to reflect incompatibilities between the processor's devices and the equivalent devices on the MCF5282. There may also be additional symbols for the devices, and there will be additional definitions for any processor-specific hardware. In particular GPIO pin handling is handled by the processor HAL, not by the variant HAL. Application developers should examine proc_io.h as well as var_io.h and the processor-specific documentation to see exactly what I/O definitions are provided. When porting to a new Coldfire processor it is best to start with an existing processor HAL and copy code as appropriate. A search for _HAS_ in var_io.h will also be informative.
Thread Contexts and Setjmp/Longjmp
All MCFxxxx processors support interrupts and exceptions in a uniform way. When an interrupt or exception occurs the hardware pushes the current program counter, the status register, and an additional 16-bit word containing information about the interrupt source, for a total of 64 bits. Hence the PCSR part of a thread context consists of two 32-bit integers, and the variant HAL provides appropriate C and assembler macros to examine and manipulate these.
Not all MCFxxxx processors have hardware floating point, so support for this is left to the processor HAL package. Some MCFxxxx processors have additional hardware units such as a multiply-accumulator, but these are not currently supported by eCos.
The various MCFxxxx processors usually have one or more UARTs based on very similar hardware. The variant HAL package can provide HAL diagnostic support using such a UART. There are some minor differences such as fifo sizes, and the UARTs will be accessed at different memory locations. These differences are handled by a small number of macros provided by the processor and platform HAL.
The MCFxxxx variant HAL only provides HAL diagnostic support via a UART if the processor or platform HAL does not provide an alternative implementation. That copes with situations where the on-chip UARTs are not actually accessible on the target board and an alternative communication channel must be used.
If the variant HAL should implement HAL diagnostics then the processor
or platform HAL should implement the CDL interface
should also define one or more of
CYGHWR_HAL_M68K_MCFxxxx_UART2, and ensure that any
multi-purpose GPIO pins are set correctly. The variant HAL will take
care of the rest.
MCFxxxx processors support a number of different caching schemes. Partial support for some of is provided by the variant HAL's cyg/hal/var_cache.h, but it is up to the processor HAL to define which caching scheme should be used, as well as parameters such as the cache size.
All MCFxxxx processors support synchronous exceptions in a uniform
way, with the hardware pushing sufficient information on to the stack
to identify the nature of the exception. This means that the architectural
hal_m68k_exception_vsr can be used as
the default VSR for all exceptions, with no need for separate
The variant HAL does not provide any special support for recovering from exceptions.
All MCFxxxx processors supports interrupts in a uniform way. When an
interrupt occurs the hardware pushes sufficient information on to the
stack to identify the interrupt. Therefore the architectural entry
hal_m68k_interrupt_vsr can be used as the
default VSR for all interrupts, with the variant just supplying a
small number of macros that allow the generic code to extract details of
the interrupt source. There is no need for separate trampoline
functions for every interrupt source.
On processors which have MCF5282-compatible interrupt and edge port
modules the variant HAL can provide the
HAL_INTERRUPT_CONFIGURE macros. There is support
for processors with a single interrupt controller or with two separate
interrupt controllers. Otherwise these macros are left to the
processor HAL. The allocation of interrupt vectors to the various
on-chip devices is also a characteristic of the processor HAL.
proc_intr.h should be
consulted for appropriate definitions, for example
The mask and umask operations are straightforward: if the interrupt controller has the SIMR and CIMR registers those will be used; otherwise the IRM registers will be updated by a read-modify-write cycle. The acknowledge macro is only relevant for external interrupts coming in via the edge port module and will clear the interrupt by writing to the EPIER register. There is no simple way to clear interrupts generated by the on-chip peripherals, so that is the responsibility of the various device drivers or of application code. The configure macro is only relevant for external interrupts and involves manipulating the edge port module.
HAL_INTERRUPT_SET_LEVEL macro is used
implicitly by higher level code such as
cyg_interrupt_create. With MCF5282-compatible
interrupt controllers the priority level corresponds to
the ICRxx register. The exact format depends on the
processor. Interrupt priorities corresponding to IPL level 7 are
non-maskable. Such interrupts cannot be managed safely by the usual
eCos ISR and DSR mechanisms. Instead application code will have to
install a custom VSR and manage the entire interrupt.
Some MCF5282-compatible interrupt controllers have a major
restriction: all interrupt priorities within each controller must be
unique. If two interrupts go off at the same time and have exactly the
same priority then the controllers' behaviour is undefined. In a
typical application some of the interrupts will be handled by eCos
device drivers while others will be handled directly by application
code. Since eCos cannot know which interrupts may get used, it cannot
allocate unique priorities. Instead this has to be left to the
application developer. eCos does provide configuration options such as
provide control over the eCos-managed interrupts, and provides default
values for these which are unique.
Non-unique interrupt priorities can lead to very confusing system
behaviour. For example on an MCF5282, if the PIT3 system clock
(interrupt 0x3a) and ethernet RX frame (interrupt 0x1b) are
accidentally given the same priority and go off at the same time, the
interrupt controller may actually issue an interrupt 0x3b, the bitwise
or of the two interrupt numbers. That interrupt belongs to the on-chip
flash module. There may not be an installed handler for that interrupt
at all, and even if there is a handler it will only manipulate the
flash hardware and not clear the system clock and ethernet interrupts.
Hence the system is likely to go into a spin, continually trying to
service the wrong interrupt. To track down such problems during
debugging it may prove useful to install a breakpoint on the
On processors with an MCF5282-compatible programmable interrupt timer
module or PIT, the variant HAL can provide the
HAL_CLOCK_LATENCY macros. These macros are used
by the eCos kernel to implement the system clock and may be used for
other purposes in non-kernel configurations. When multiple timers are
available it is up to the processor or platform HAL to select which
one gets used for the system clock. It is also up to the processor or
platform HAL to provide various clock-related configuration options
CYGNUM_HAL_RTC_PERIOD. Those options need
to take into account the processor clock speed, which is usually a
characteristic of the platform and hence not known to the variant HAL.
When porting to a new Coldfire processor, the processor or platform
HAL should define the symbols
_HAL_MCFxxxx_CLOCK_PIT_PRE_. Existing ports can be
examined for more details.
On processors with an MCF5282-compatible reset module or RST, the
variant HAL can provide the
macro. That macro is typically used by the gdb stubs support inside
RedBoot to reset the hardware between debug sessions, ensuring that
each session runs in as close to pristine hardware as possible. The
macro uses the SOFTRST bit of the
By default the variant HAL will provide versions of
HAL_MSBIT_INDEX which are more efficient than the
default ones in the architectural HAL. The implementation uses the
ff1.l and bitrev.l instructions.
If the Coldfire processor does not support these instructions then
the processor HAL should define
The MCFxxxx variant HAL does not affect the implementation of data types, stack size definitions, idle thread processing, linker scripts, SMP support, system startup, or debug support.