The functions smsc9220_receive_by_chunks and smsc9220_send_by_chunks are
supposed to implement receiving and sending packets by chunks. However,
the functions SMSC9220_EMAC::low_level_input and SMSC9220_EMAC::link_out,
which call them respectively, already require or assemble the full packet.
Also, smsc9220_receive_by_chunks doesn't implement the "chunks" part.
This commit renames the functions to smsc9220_receive_packet and
smsc9220_send_packet. The functions now do their operations by word
instead of by bytes. The functions SMSC9220_EMAC::low_level_input and
SMSC9220_EMAC::link_out already handle allocation, continuity and word
alignment of the packet buffer.
The function smsc9220_receive_by_chunks loads data from the Ethernet port.
It is expected to return the Ethernet frame without the 4 CRC bytes.
However, it is required to call the Ethernet data port register (32-bit)
an amount equal to the number of frame words (including the 4 CRC bytes)
to pop all frame words. The current code doesn't call the register for the
last word (which has CRC data). This causes subsequent calls to have this
missed word at the beginning. The impact of this is huge as the high level
API is getting fed wrong data. The fix adds one additional call to the data
port register.
The HAL gpio_irq_api stores object IDs, which serve as a form of context
for the dispatch of the interrupt handler in the drivers level
InterruptIn Class. The way this is achieved is that the InterruptIn
Class casts its address to uint32_t, which is stored as the ID.
This results in compilation failure when the size of an object pointer
is greater than uint32_t, for example when building on a PC for unit
testing.
In order to allow Unit Testing of the InterruptIn Class, we replace the
use of uint32_t with uintptr_t (type capable of holding a pointer),
which allows portability and expresses intentions more clearly.
In aid of this latter goal, we also replace the use of the name "id"
with "context", to improve clarity - these are addresses of the context
related to that callback.
Import the latest partition headers from upstream. We no longer need to
patch the headers to define the BL2 macro as we now define it in the
scatter files for the MUSCA targets.
We worked around an issue with mbed-cli1 not recognising the BL2 macro
from targets.json by adding patched versions of region_defs.h and
flash_layout.h for ARM_MUSCA targets. In the patched headers we defined
the BL2 macro to ensure it can be picked up by the ARM scatter files
that include the headers.
The current solution is not robust, because it means that the
aforementioned headers easily become out of date. A workaround of
defining the macros in the scatter files which need them was suggested
in https://github.com/ARMmbed/mbed-os/issues/14762
This commit applies the suggested changes to the ARM_MUSCA scatter
files.
tfm_ns_interface.c is intended to be overriden by clients to support
different targets. We copy this file from upstream into the mbed-os
platform library. We also have a specific "strong" overridden version
for the NU_M2354 target, which is located in its target library.
Previously the implementations in the platform library were decorated
with __attribute__(weak), and we provided a strong definition for the
NU_M2354 target. This worked fine because of weak linking, the linker
will pick the first "strong" definition and use that, avoiding any ODR
violations. However, upstream have removed __attribute__(weak) from the
function definitions, which caused multiply defined symbol errors when
trying to build the NU_M2354 target.
To work around the above issue, we remove the common definition in the
platform library; instead we copy the file to the Musca B1 and Musca S1
target libaries. This means the appropriate tfm_ns_interface.c is only
included in the build when compiling for the specific target which uses
it.
- list files included via module path
- <project-name>_SOURCE_DIR for sources that are out of the current processed CMake
- CMAKE_CURRENT_LIST_DIR for listfiles
The signing keys were previously imported from trusted-firmware-m
and located in mbed-os/tools/targets/musca_* (path for Mbed CLI 1).
This PR copie them into each target's directory as per the
convention of the new tools. Keys in the old path remain untouched
for backward compatibility, but they will be eventually removed
once we stop supporting Mbed CLI 1.
In targets.json, ARM_MUSCA_B1 and ARM_MUSCA_S1 have alias target
names suffixed with `_NS`. They are identical to targets without
`_NS` and exist purely for compatibility with the old naming
convention we had. The CI builds them as separate targets and uses
extra resources.
As we are upgrading Musca targets to TF-M v1.2, it's time to clean
up the aliases.
The vector table needs to be copied from ROM to RAM, in order for us
to set IRQ handlers at run time. The address in RAM is defined by
`NVIC_RAM_VECTOR_ADDRESS` in `cmsis_nvic.h`, but its inclusion
was missing from Musca S1's `cmsis.h` and consequently the vector
table was not copied.
On most targets this results in a memory access error when we set
vectors. But Musca S1's ROM is in its MRAM (which can be accessed
like any RAMs), and this causes the ROM image to be modified
with no error/warning. On the next boot, MCUboot fails the image
integrity check.
This commit adds the missing include, in the same spirit as
01dd997d55.
* Partition files are synced with TF-M v1.2
* To have uniformity with TF-M v1.2, rename the following:
** image_macros_preprocessed_ns.c to `signing_layout_ns.c`
** image_macros_preprocessed_s.c to `signing_layout_s.c`
* `MCUBOOT_IMAGE_NUMBER` is set to 2 by default for TF-M v1.2,
therefore it is necessary that Mbed OS compiles the right macros
for when linking and using the partition files.
* Partition files are synced with TF-M v1.2
* To have uniformity with TF-M v1.2, rename the following:
** image_macros_preprocessed_ns.c to `signing_layout_ns.c`
** image_macros_preprocessed_s.c to `signing_layout_s.c`
* `MCUBOOT_IMAGE_NUMBER` is set to 2 by default for TF-M v1.2,
therefore it is necessary that Mbed OS compiles the right macros
for when linking and using the partition files
** Workaround **
The `region_defs.h` has an explicit definition of `BL2`, even
though it is already defined in target.json for `ARM_MUSCA_B1`.
This is because of Mbed CLI 1, as it can't seem to use the right
macro when linking the files for Mbed OS application when using
the ARMCLANG toolchain.
The heap size was incorrectly calculated.
This fixes it by subtracting the Stack size, any memory chunks allocated
before the start of the application (for vectors and/or crash report), and
finally the size of the application from the total RAM size.
Add license identifier to files which Arm owns the copyright to,
and contain either BSD-3 or Apache-2.0 licenses. This is to address
license errors raised by scancode analysis.
Workaround a bug where the boot stack size configuration option is not
passed on to armlink, the Arm Compiler's linker. Prefer
MBED_CONF_TARGET_BOOT_STACK_SIZE if present, as this is what the
configuration system should provide. Fall back to MBED_BOOT_STACK_SIZE
if MBED_CONF_TARGET_BOOT_STACK_SIZE is not defined, as in the case of
buggy tools. If both MBED_CONF_TARGET_BOOT_STACK_SIZE and
MBED_BOOT_STACK_SIZE are not defined, then we fall back to a hard-coded
value provided by the linkerscript. See
https://github.com/ARMmbed/mbed-os/issues/13474 for more information.
To allow overriding of the boot stack size from the Mbed configuration
system, consistently use MBED_CONF_TARGET_BOOT_STACK_SIZE rather than
MBED_BOOT_STACK_SIZE.
Fixes#10319
Rami Elkhatib
Hari Limaye
Robert Walton
Martin Kojtal
Lingkai Dong
Martin Kojtal
George Psimenos
Vikas Katariya
Hugues Kamba
Harrison Mutai
Mark Horvath
Gabor Toth
Jaeden Amero