This commit reverts the revert on the new task watchdog API. It also fixes the following bug which caused the reversion. - sdkconfig TASK_WDT_TIMEOUT_S has been reverted from the unit of ms back to the unit of seconds. Fixes bug where projects using the new API without rebuilding sdkconfig would cause the old default value of 5 to be interpreted in ms. This commit also adds the following features to the task watchdog - Updated idle hook registration to be compatible with dual core hooks - Updated dual core hooks to support deregistration for cpu - Legacy mode has been removed and esp_task_wdt_feed() is now replaced by esp_task_wdt_reset(). esp_task_wdt_feed() is deprecated - Idle hooks to reset are now registered/deregistered when the idle tasks are added/deleted from the Task Watchdog instead of at Task Watchdog init/deinit - Updated example
103 lines
4.7 KiB
ReStructuredText
103 lines
4.7 KiB
ReStructuredText
Watchdogs
|
|
=========
|
|
|
|
Overview
|
|
--------
|
|
|
|
The ESP-IDF has support for two types of watchdogs: The Interrupt Watchdog Timer
|
|
and the Task Watchdog Timer (TWDT). The Interrupt Watchdog Timer and the TWDT
|
|
can both be enabled using ``make menuconfig``, however the TWDT can also be
|
|
enabled during runtime. The Interrupt Watchdog is responsible for detecting
|
|
instances where FreeRTOS task switching is blocked for a prolonged period of
|
|
time. The TWDT is responsible for detecting instances of tasks running without
|
|
yielding for a prolonged period.
|
|
|
|
Interrupt watchdog
|
|
^^^^^^^^^^^^^^^^^^
|
|
|
|
The interrupt watchdog makes sure the FreeRTOS task switching interrupt isn't blocked for a long time. This
|
|
is bad because no other tasks, including potentially important ones like the WiFi task and the idle task,
|
|
can't get any CPU runtime. A blocked task switching interrupt can happen because a program runs into an
|
|
infinite loop with interrupts disabled or hangs in an interrupt.
|
|
|
|
The default action of the interrupt watchdog is to invoke the panic handler. causing a register dump and an opportunity
|
|
for the programmer to find out, using either OpenOCD or gdbstub, what bit of code is stuck with interrupts
|
|
disabled. Depending on the configuration of the panic handler, it can also blindly reset the CPU, which may be
|
|
preferred in a production environment.
|
|
|
|
The interrupt watchdog is built around the hardware watchdog in timer group 1. If this watchdog for some reason
|
|
cannot execute the NMI handler that invokes the panic handler (e.g. because IRAM is overwritten by garbage),
|
|
it will hard-reset the SOC.
|
|
|
|
Task Watchdog Timer
|
|
^^^^^^^^^^^^^^^^^^^
|
|
|
|
The Task Watchdog Timer (TWDT) is responsible for detecting instances of tasks
|
|
running for a prolonged period of time without yielding. This is a symptom of
|
|
CPU starvation and is usually caused by a higher priority task looping without
|
|
yielding to a lower-priority task thus starving the lower priority task from
|
|
CPU time. This can be an indicator of poorly written code that spinloops on a
|
|
peripheral, or a task that is stuck in an infinite loop.
|
|
|
|
By default the TWDT will watch the Idle Tasks of each CPU, however any task can
|
|
elect to be watched by the TWDT. Each watched task must 'reset' the TWDT
|
|
periodically to indicate that they have been allocated CPU time. If a task does
|
|
not reset within the TWDT timeout period, a warning will be printed with
|
|
information about which tasks failed to reset the TWDT in time and which
|
|
tasks are currently running on the ESP32 CPUs and.
|
|
|
|
The TWDT is built around the Hardware Watchdog Timer in Timer Group 0. The TWDT
|
|
can be initialized by calling :cpp:func:`esp_task_wdt_init` which will configure
|
|
the hardware timer. A task can then subscribe to the TWDT using
|
|
:cpp:func:`esp_task_wdt_add` in order to be watched. Each subscribed task must
|
|
periodically call :cpp:func:`esp_task_wdt_reset` to reset the TWDT. Failure by
|
|
any subscribed tasks to periodically call :cpp:func:`esp_task_wdt_reset`
|
|
indicates that one or more tasks have been starved of CPU time or are stuck in a
|
|
loop somewhere.
|
|
|
|
A watched task can be unsubscribed from the TWDT using
|
|
:cpp:func:`esp_task_wdt_delete()`. A task that has been unsubscribed should no
|
|
longer call :cpp:func:`esp_task_wdt_reset`. Once all tasks have unsubscribed
|
|
form the TWDT, the TWDT can be deinitialized by calling
|
|
:cpp:func:`esp_task_wdt_deinit()`.
|
|
|
|
By default :ref:`CONFIG_TASK_WDT` in ``make menuconfig`` will be enabled causing
|
|
the TWDT to be initialized automatically during startup. Likewise
|
|
:ref:`CONFIG_TASK_WDT_CHECK_IDLE_TASK_CPU0` and
|
|
:ref:`CONFIG_TASK_WDT_CHECK_IDLE_TASK_CPU1` are also enabled by default causing
|
|
the two Idle Tasks to be subscribed to the TWDT during startup.
|
|
|
|
JTAG and watchdogs
|
|
^^^^^^^^^^^^^^^^^^
|
|
|
|
While debugging using OpenOCD, the CPUs will be halted every time a breakpoint
|
|
is reached. However if the watchdog timers continue to run when a breakpoint is
|
|
encountered, they will eventually trigger a reset making it very difficult to
|
|
debug code. Therefore OpenOCD will disable the hardware timers of both the
|
|
interrupt and task watchdogs at every breakpoint. Moreover, OpenOCD will not
|
|
reenable them upon leaving the breakpoint. This means that interrupt watchdog
|
|
and task watchdog functionality will essentially be disabled. No warnings or
|
|
panics from either watchdogs will be generated when the ESP32 is connected to
|
|
OpenOCD via JTAG.
|
|
|
|
|
|
Interrupt Watchdog API Reference
|
|
--------------------------------
|
|
|
|
Header File
|
|
^^^^^^^^^^^
|
|
|
|
* :component_file:`esp32/include/esp_int_wdt.h`
|
|
|
|
|
|
Functions
|
|
---------
|
|
|
|
.. doxygenfunction:: esp_int_wdt_init
|
|
|
|
Task Watchdog API Reference
|
|
----------------------------
|
|
|
|
A full example using the Task Watchdog is available in esp-idf: :example:`system/task_watchdog`
|
|
|
|
.. include:: /_build/inc/esp_task_wdt.inc
|