Add rescue lane for targets whose firmware steals the SWD pins

[sensoryelectronics]

Problem

On boards where the debug pins are muxed with a peripheral that the running firmware enables, the target becomes permanently undebuggable: AttachChip is refused with protocol error 0x55 within milliseconds of every boot, and the condition survives power cycles because it lives in flash.

Concrete case: the WeAct nanoCH32H417 routes SWCLK/SWDIO to PB8/PB9, which are also USB2 D+/D-. Flashing any USBHS device firmware kills SWD on the next boot — every subsequent attach fails 0x55, through replugs, full power-offs, and probe mode-switch round-trips. The same trap exists for CH32V003 users who remux SWIO as GPIO.

What this adds

• wlink reset force — fires the raw 0x0b reset ladder (chip/soft/normal) without attaching. Diagnostic: separates probe-alive from target-unresponsive (the resets are visibly acknowledged by the board while attach still refuses).
• wlink reset rescue and a global --rescue flag — hammers AttachChip for 15 s while the operator taps the target RESET, catching the brief post-reset window before user code runs, then immediately halts the core. With the global flag, any command (flash, erase, dump) can attach this way.
• ProbeSession::reset_mcu_and_halt() — ndmreset with haltreq held, so the caught core comes back with clean peripheral/clock state while still never executing a user instruction (activated from the commented reference block).
• wlink erase --method dmi — erases the first 64 K via direct FLASH-register access over DMI (also activated from the reference block; register map verified against the CH32H417 RM ch. 46). Needed because the probe firmware's own flash stub aborts (fastprogram 0x05) on a rescue-halted core; blanking the vectors kills the pin-stealing firmware permanently, after which the normal attach + flash lane works again.
• Rescue sessions skip unprotect_flash — its reattach_chip() detaches (resuming the CPU), which re-steals the pins mid-operation.

Field test

Recovered a nanoCH32H417 (CH32H417QEU, ChipID 0x4170052d) that USBHS CDC firmware had made undebuggable: --rescue attach caught the boot window on the first try, erase --method dmi blanked the vectors, and a plain wlink flash then restored the board. Attach has been reliable since.

🤖 Generated with Claude Code