OS/FreeRTOS: fix tick drift caused by wrong elapsed time calculation

When WFI wakes up but MTIME has not yet incremented, xModifiableIdleTime
equals XLastLoadValue. The previous '>' comparison would fall through to
the else branch and compute a bogus wraparound value (0xFFFFFFFF cycles),
causing vTaskStepTick() to advance the system tick by a huge amount.

Change '>' to '>=' so that equal values correctly yield zero elapsed time.

Signed-off-by: Huaqi Fang <578567190@qq.com>

Author: fanghuaqi

OS/FreeRTOS/Source/portable/port.c

@@ -681,7 +681,11 @@ __attribute__((weak)) void vPortSuppressTicksAndSleep(TickType_t xExpectedIdleTi
             periods (not the ulReload value which accounted for part
             ticks). */
             xModifiableIdleTime = SysTimer_GetLoadValue();
-            if (xModifiableIdleTime > XLastLoadValue) {
+            /* Use >= to handle the case where WFI wakes up but MTIME has not
+               yet incremented (xModifiableIdleTime == XLastLoadValue), in which
+               case elapsed time is 0 instead of wrapping around via the else
+               branch which would produce a bogus large tick count. */
+            if (xModifiableIdleTime >= XLastLoadValue) {
                 ulCompletedSysTickDecrements = (xModifiableIdleTime - XLastLoadValue);
             } else {
                 ulCompletedSysTickDecrements = (xModifiableIdleTime + portMAX_BIT_NUMBER - XLastLoadValue);

doc/source/changelog.rst

@@ -169,6 +169,7 @@ This is release version of ``0.9.0`` of Nuclei SDK.
   - Fix FreeRTOS port macro interrupt masking implementation in ``portmacro.h`` to ensure pending interrupts are properly handled after enabling interrupts, preventing assertion failures from executing code before interrupt response
   - Fix RT-Thread and UCOSII interrupt masking implementation in ``cpuport.c`` and ``os_cpu_port.h`` to add memory barriers after CSR operations, ensuring pending interrupts are properly handled after enabling interrupts
   - Fix FreeRTOS tickless sleep mode by removing unnecessary ``__FENCE_I()`` call in ``vPortSuppressTicksAndSleep`` function to avoid i-cache misses and reduce interrupt latency. The ``__RWMB()`` memory barrier is sufficient for proper interrupt enable propagation
+  - Fix FreeRTOS tickless sleep elapsed time calculation in ``vPortSuppressTicksAndSleep`` function by changing ``>`` to ``>=`` comparison, which prevents bogus tick count when WFI wakes up but MTIME has not yet incremented, avoiding incorrect system tick advancement
   - Upgrade FreeRTOS to v11.2.0 and update porting layer to match new API requirements (critical nesting macros and recursive lock functions now accept core ID parameter for SMP systems)
   - Fix ThreadX ``tx_port.h`` interrupt disable/restore macros (``TX_DISABLE``/``TX_RESTORE``) by adding ``volatile`` keyword and ``memory`` clobber to inline assembly to prevent compiler reordering and ensure proper memory barrier semantics
   - Add memory barriers to ThreadX UP and SMP ports after interrupt control operations (such as ``__enable_irq()`` and ``CSR_MSTATUS`` access) to ensure proper synchronization and prevent potential compiler/CPU reordering issues