I have a bug I have a particularly hard time understanding.
My rp2040 runs a dual core firmware that presents a USB device to a computer. A button is connected to one of its GPIO. When the button is pressed, is called. But nothing happens and the device goes on about its day, seemingly functioning properly ! Then, when a seemingly unrelated event happens (a non-control USB transfer..), the device reboots, depending on whether the button got pushed before.
That's the part that's incredibly confusing to me. watchdog_reboot is called once, when the button is released. Nothing else is done, it just calls that function. I've tried replacing the call with lighting a led, and the led gets lit up. So how can a watchdog reboot possibly end up delayed until some other event happen ? The only thing I can think of is that the interrupt is set, but ignored. But how can a watchdog interrupt ever get ignored ? Not only, to the best of my knowledge, are the interrupts not disabled at the watchdog_reboot point, but does save_and_disable_interrupts even disable the watchdog ?.. I mean, the watchdog interrupt isn't part of the 26 'user' IRQs, the ones connected to the NVIC. And it preempts all but hard faults.
Am I misunderstanding some thing about the watchdog ? Do I need to enable something ? And I don't mean enable the watchdog in the traditional sense to reboot if the device freezes up; I mean, prerequisites for watchdog_reboot(0, 0, 0); to reboot instantly. I'm just calling watchdog_reboot out of the blue here, no set-up done whatsoever. And it gets buffered, somehow. Thanks for the suggestions should have any.
My rp2040 runs a dual core firmware that presents a USB device to a computer. A button is connected to one of its GPIO. When the button is pressed,
Code:
watchdog_reboot(0, 0, 0);That's the part that's incredibly confusing to me. watchdog_reboot is called once, when the button is released. Nothing else is done, it just calls that function. I've tried replacing the call with lighting a led, and the led gets lit up. So how can a watchdog reboot possibly end up delayed until some other event happen ? The only thing I can think of is that the interrupt is set, but ignored. But how can a watchdog interrupt ever get ignored ? Not only, to the best of my knowledge, are the interrupts not disabled at the watchdog_reboot point, but does save_and_disable_interrupts even disable the watchdog ?.. I mean, the watchdog interrupt isn't part of the 26 'user' IRQs, the ones connected to the NVIC. And it preempts all but hard faults.
Am I misunderstanding some thing about the watchdog ? Do I need to enable something ? And I don't mean enable the watchdog in the traditional sense to reboot if the device freezes up; I mean, prerequisites for watchdog_reboot(0, 0, 0); to reboot instantly. I'm just calling watchdog_reboot out of the blue here, no set-up done whatsoever. And it gets buffered, somehow. Thanks for the suggestions should have any.
Statistics: Posted by JBernard — Sat Mar 16, 2024 10:24 pm