Boards that are shipped with a DFU bootloader define the `dfu-util` configuration in their `Makefile.include`. However, when `riotboot_dfu` is used as the DFU bootloader, the board's `dfu-util` configuration must be overridden by the configuration as required by `riotboot_dfu` to use it to flash applications. Therefore, all `dfu-util` configurations are defined as overridable in the board's `Makefile.include` and the configuration as required by `riotboot_dfu` is included before the board's `Makefile.include`.
STM32F2/4/7 MCUs use sectors instead of pages, where the minimum sector size is defined by FLASHPAGE_MIN_SECTOR_SIZE, which is 16KB or 32KB (the first sector) depending on the CPU_MODEL. In this case SLOT0_OFFSET must be a multiple of the minimum sector size to cover a whole sector.
The `dfu-util` command uses `--device $(DFU_USB_ID)`, where `DFU_USB_ID` is derived from `USB_VID` and `USB_PID` to specify the DFU device to use. Without specifying `USB_VID` and `USB_PID` in the make command, `dfu-util` is called with `--device :`, which only works if there is only one DFU device. Also, the STM32 make system forbids the use of `dfu-util` as programmer without setting the variable `DFU_USB_ID`. Therefore, the documentation is changed to use `USB_VID`/`USB_PID` or `DFU_USB_ID` in the make command.
If the STM32 MCU does not have an internal D+ pullup and there is no dedicated GPIO on the board to simulate a USB disconnect, the D+ GPIO
has to be temporarily configured as an output and pushed down to simulate a disconnect/connect cycle to allow the host to recognize the device. This is done correctly in `cpu/stm32/periph/usbdev_fs.c` but not in tinyUSB STM32 hardware driver. That is, pressing the RESET button doesn't reset the USB connection.
fixup! pkg/tinyusb: fix reset for STM32 boards without internal D+ pullup
Previously, the callback was incorrectly passed a channel of zero as
argument regardless of the channel that triggered the IRQ. This fixes
the issue and also uses `bitarithm_test_and_clear()` to only iterate
over the channels that actually have an IRQ flag set, rather than
all channels.
The linter was unhappy that `unsinged long` and `uint32_t` were used
inconsistency (in the `timer_init()` declaration, implementation, as
well as in the `DEBUG()` format specifiers).
To synchronize communication via shared memory between ISR context and
thread context it is a common misconception that `volatile` is
sufficient. This is however is not the cause and the cause of many
subtle data race bugs. This fixes the issue.
