rom_reset_usb_boot does not function when Watchdog is enabled #2689

New issue

Open

opened 2025-10-06 04:05:14 -05:00 by jondurrant · 4 comments

jondurrant commented

2025-10-06 04:05:14 -05:00

(Migrated from github.com)

On SDK 2.2.0 if you enable the watchdog then calling rom_reset_usb_boot just halts the calling core but does not trigger bootsel mode. This doesn't seem to be mentioned in the documentation and seems an unusual side effect.

lurch commented

2025-10-06 04:30:16 -05:00

(Migrated from github.com)

It looks like on RP2350 rom_reset_usb_boot eventually calls rom_reboot which is documented as "Reboot using the watchdog. Resets the chip and uses the watchdog facility to restart.", so maybe that's the "smoking gun" as to why this isn't working? (But I've no idea if this is a deliberate side-effect or an unintentional side-effect.)

It _looks_ like on RP2350 `rom_reset_usb_boot` eventually calls `rom_reboot` which is [documented](https://www.raspberrypi.com/documentation/pico-sdk/runtime.html#group_pico_bootrom_1gadbb1f51a04fd0c27513468996d24f3ce) as "Reboot using the watchdog. Resets the chip and uses the watchdog facility to restart.", so maybe that's the "smoking gun" as to why this isn't working? (But I've no idea if this is a deliberate side-effect or an unintentional side-effect.)

jondurrant commented

2025-10-06 04:46:33 -05:00

(Migrated from github.com)

In terms of issue, either the side effect needs to be documented or a fix put in place.

Good spot. Would explain it as I have the other RP2350 core resetting the watchdog timer. With this information I could write a workaround so the core resetting the watchdog is the one to trigger bootsel. In terms of issue, either the side effect needs to be documented or a fix put in place.

👍 1

will-v-pi commented

2025-10-07 03:27:40 -05:00

(Migrated from github.com)

I think the full explanation here is that the core calling rom_reboot sets the watchdog up for a BOOTSEL reboot, then loads 10ms into the watchdog and goes __wfi() to wait for the reboot. If the other core prevents that watchdog reboot (by eg reloading the watchdog) then the calling core just remains in __wfi(). Fix is to not use the watchdog for 2 things at the same time, as there's only 1 watchdog.

I think the full explanation here is that the core calling `rom_reboot` sets the watchdog up for a BOOTSEL reboot, then loads 10ms into the watchdog and goes `__wfi()` to wait for the reboot. If the other core prevents that watchdog reboot (by eg reloading the watchdog) then the calling core just remains in `__wfi()`. Fix is to not use the watchdog for 2 things at the same time, as there's only 1 watchdog.

lurch commented

2025-10-07 03:35:35 -05:00

(Migrated from github.com)