The task dispatch method for the esp_timer could stall even if it is
armed if ther ISR dispatch alarm triggers close to the task dispatch.
This MR fixes a bug where the esp_timer cleared the incorrect cached
array timer index and subsequently the timer task is never woken up.
Closes https://github.com/espressif/esp-idf/issues/18808
print_timer_info advanced the dump cursor by snprintf's return value. When a timer line was truncated, snprintf returned the full would-be length, which could move the cursor past the heap buffer and wrap the remaining size before the next write. Add a bounded append helper that clamps truncation to the end of the buffer while preserving the NUL terminator.
On real hardware, ESP_TIMER_ISR callbacks run in a hardware interrupt
that preempts any FreeRTOS task. On the Linux simulator, there are no
real ISRs — the alarm is detected by a native pthread but was only
forwarded to the FreeRTOS timer_task via xTaskNotifyGive(). This meant
ISR-dispatch callbacks could be starved by higher-priority FreeRTOS
tasks, breaking components like the task watchdog that rely on
ISR-dispatch timers to detect scheduling starvation.
Move ISR-dispatch timer processing into the alarm pthread itself,
mirroring the hardware ISR path. The FreeRTOS timer_task is only
notified when no ISR-dispatch timer consumed the alarm. This is safe
because the Linux FreeRTOS port already handles vPortEnterCritical()
calls from non-FreeRTOS threads (bumps nesting counter without
blocking on scheduled-task checks).
ESP32-S31 has a separate functional clock gate for the systimer
(reg_systimer_clk_en in HP_SYS_CLKRST) that defaults to off. Without
it, the counter snapshot never completes, hanging execution inside
esp_timer_init_nonos.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Previously it would return ESP_ERR_INVALID_STATE, which meant that if called from
user-code before the system tries to initialize the timer then esp-idf would
fail to boot.
This could happen if a user wanted to use esp-timer from a cpp constructor.
Closes https://github.com/espressif/esp-idf/issues/9679
This commit replaces the use of portNUM_PROCESSORS and configNUM_CORES
macros in all of ESP-IDF. These macros are needed to realize an SMP
scenario by fetching the number of active cores FreeRTOS is running on.
Instead, a new Kconfig option, CONFIG_FREERTOS_NUMBER_OF_CORES, has been
added as a proxy for the FreeRTOS config option, configNUMBER_OF_CORES.
This new commit is now used to realize an SMP scenario in various places
in ESP-IDF.
[Sudeep Mohanty: Added new Kconfig option CONFIG_FREERTOS_NUMBER_OF_CORES]
Signed-off-by: Sudeep Mohanty <sudeep.mohanty@espressif.com>
CONFIG_ESP_TIMER_ISR_AFFINITY can be equal to -1, whereas
ESP_SYSTEM_INIT_FN takes an uint16_t argument. To avoid overflow,
move the choice of init mask into source code and set the value
explicitly.
Similar to how the secondary init functions were already registered
via ESP_SYSTEM_INIT_FN, do the same for the core init functions.
This MR doesn't actually move the init functions into respective
components yet. This has to be carefully done in follow-up MRs.
Inadequate locking in the esp_timer component allowed corruption
of the s_timers linked list:
1. timer_armed(timer) returns false
2. another task arms the timer and adds it to s_timers
3. the list is locked
4. the timer is inserted into s_timers again
The last step results in a loop in the s_timers list, which causes
an infinite loop when iterated. This change always locks the
list before checking if the timer is already armed avoiding
the data race.
