Skip to content

type: worklist status: current summary: Open work on the external flash loaders: the L432 state-capture script, watchdog refresh in the ST entry points, untested MassErase, and NAND loader distribution.

Loaders Worklist

Delete an item when it is done. Runtime rules are in Loader Runtime Design.

  • tag_capture_state.py does not work on STM32L432. Its SRAM step requests 256 KB (the U375's size) and fails, and its backup-register step also fails. Both come from U375 constants in the script. For the backup registers it writes the U375's RCC_APB1ENR1 at 0x40030C9C (RTCAPBEN is bit 30 there). On the L432 that address maps to nothing; the register is at 0x40021058, with RTCAPBEN at bit 10. The internal-flash regions capture correctly. Until this is fixed, read the backup registers by hand: under reset, enable RCC_APB1ENR1_RTCAPBEN, then read 0x40002850, 128 bytes.
  • The ST entry points do not refresh the watchdog. Serve() does: it writes IWDG_KR = 0xAAAA while it waits for commands. The ST entry points do not. A tag whose option bytes select the hardware watchdog (FLASH_OPTR.IWDG_SW = 0) has it running from reset, and the debug freeze covers only a halted core. A long run of loader code would then be cut off by a reset, a mass erase above all. The bench PresTag uses the software watchdog, so it is unaffected. The fix is an IWDG_KR = 0xAAAA write in the poll loops, which has no effect when the watchdog is not running; see SWD Capture and Recovery Library.
  • MassErase is untested. It has not been established whether -e all with a loader loaded also erases internal flash, and it was not tried on a tag with firmware worth keeping. Sector erase is tested, through the programmer's erase-before-write.
  • Distribute the NAND loader? IMUTagNandBmp581 is a distributed tag, but GD5F2GM7RE_IMUTagNandv2 is not marked DISTRIBUTE, unlike the AT25XE loaders, so it does not ship in the firmware release beside the tag it reads.
  • Run the linear path through tag-capture on a PresTag. The AT25 loaders use the same code through tag-capture as through tag-xflash dump. The U375 loader plan recorded on 2026-10-02 that this combination had not yet been run on a PresTag.