yangsong8-a1 opened a new pull request, #21850: URL: https://github.com/apache/nuttx/pull/21850
## Summary Add a watchdog lower half for the ARM CMSDK APB watchdog (`CONFIG_WATCHDOG_CMSDK`), registered with `cmsdk_wdt_initialize()` as `/dev/watchdogN`. The first expiry raises the watchdog interrupt; the second one resets the system. With `WDIOC_CAPTURE`, the reset stage is disabled and the handler is called on each first-stage expiry. ## Impact New driver, off by default. No board enables it yet. The watchdog interrupt must be a maskable IRQ: the driver keeps INTEN set and relies on the interrupt staying masked until a capture handler is attached. On a system where it is routed to NMI (as on the MPS2/MPS3 boards) the interrupt handler reloads the counter, so the reset stage is never reached. ## Testing Ubuntu 22.04, arm-none-eabi-gcc 10.3, QEMU 6.2. | Config | Result | |---|---| | mps2-an500:nsh + WATCHDOG + WATCHDOG_CMSDK (Make and CMake) | builds, 0 warnings | | mps3-an547:nsh + WATCHDOG + WATCHDOG_CMSDK | builds, 0 warnings | | QEMU mps2-an500, watchdog at 0x40008000, test program, `WDIOC_CAPTURE` mode | see the log below | | reset stage | not run on QEMU: mps2 wires the watchdog interrupt to NMI, which cannot be masked, so the reset stage is never reached there. Tested on my own product and EVB board | QEMU mps2-an500, timeout 500 ms with a capture handler: ``` nsh> hello w hello w: register ret=0 nsh> hello c open fd=3 settimeout(500) ret=0 capture ret=0 start ret=0 status flags=0x5 timeout=500 timeleft=242 early-warning interrupts in 3.2 s: 8 stop ret=0 interrupts 1 s after stop: 0 ``` 8 instead of 6 expiries: on QEMU mps2-an500 the guest SysTick time runs slower than the host clock that drives the watchdog model. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
