This refactoring of the chip specific bootloader creates the internal boot instance and aligned
buffer in the prepare stage, so that they are automatically dropped after. This unlocks a use
case where peripherals owning the flash need to be Drop'ed before load() happens.
* Allow manipulating state without accessing DFU partition.
* Provide aligned buffer when creating updater to reduce potential wrong parameters passed.
It seems as if the arm compiler can does not care about whether the bootloader symbols are undefined if the default() function is never used. The x64 compiler does care however, so this change ensures that we can instantiate the types from tests.
1324: Add MCO support for L4 and F4 families r=Dirbaio a=m-dupont
Add MCO support for L4 and F4 as already done in F7.
When the 'HSI' source is selected as MCO source, 'HSI' is activated (`set_hsion(true)`) . This is done to operate the MCO in case 'MSI' is chosen as the clock source for the CPU. The same applies to PLL, etc.
1327: Avoid write before erase r=Dirbaio a=rmja
This introduces an additional marker to the state partition right after the magic which indicates whether the current progress is valid or not. Validation in tests that we never write without an erase is added.
There is currently a FIXME in the FirmwareUpdater. Let me know if we should take the erase value as a parameter. I opened a feature request in embedded-storage to get this value in the trait. Before this, the assumption about ERASE_VALUE=0xFF was the same.
I have made some thoughts about whether this is a breaking change between the app and firmware, i.e. whether adding the "Progress valid" field is breaking. My conclusion is that it is not a breaking change. For the case where an app uses this new FirmwareUpdater together with an old bootloader, what it now does, is that it:
1. Writes the progress valid field to all zeros. This field is not known in the old bootloader, so it actually writes a "current progress" index.
2. The entire state partition is erased - effectively removing any trace of 1.
3. Set magic
This should be compatible.
Co-authored-by: Mathieu Dupont <mdupont@cppm.in2p3.fr>
Co-authored-by: Rasmus Melchior Jacobsen <rmja@laesoe.org>
This introduces an additional marker to the state partition right after the magic which indicates whether the current progress is valid or not. Validation in tests that we never write without an erase is added.
There is currently a FIXME in the FirmwareUpdater. Let me know if we should take the erase value as a parameter. I opened a feature request in embedded-storage to get this value in the trait. Before this, the assumption about ERASE_VALUE=0xFF was the same.
FirmwareWriter currently has a "max-write-size" parameter, but this is a limitation that should be handled by chunking inside the NorFlash driver, and not "up here" in user code. In case that the driver (e.g. qspi driver) is unaware of any max-write limitations, one could simply add an intermediate NorFlash adapter providing the chunk'ing capability.
This commit provides a method to verify that firmware has been signed with a private key given its public key. The implementation uses ed25519-dalek as the signature verifier. An "ed25519" feature is required to enable the functionality. When disabled (the default), calling the firmware updater's verify method will return a failure.
1139: Wdt config changes r=lulf a=huntc
Per commits:
* By passing WDT config around we can control it more easily and promote sharing it between files.
* The memory layout of the s140 crept into a number of memory files, which can cause confusion (well, it did for me!).
* Obtaining the current WDT config is useful so that we do not have to duplicate configurations around the place. A constructor method has been introduced that attempts to return the current running WDT config from the WDT peripheral. The bootloader example has also been updated to show how the watchdog can be obtained and used.
Co-authored-by: huntc <huntchr@gmail.com>
Fixes a bug in the partition assertions that ensures that the state
page(s) have enough space for 2x active partition range.
Add unit test to verify that panic is observed.
Compiler will infer a different lifetime for BootFlash than for the
borrowed flash, which makes it require more type annotations than if it
was just owning the type. Since it doesn't really matter if it owns or
borrows in practical use, change it to own so that it simplifies usage.
* Remove the need for generic const expressions and use buffers provided in the flash config.
* Extend embedded-storage traits to simplify generics.
* Document all public APIs
* Add toplevel README
* Expose AlignedBuffer type for convenience.
* Update examples
