I have been trying with multiple tutorials to implement a simple spin lock in pure c (i.e. with no OS specific stuff).

This is the last i have come up with

myFile.c

static atomic_flag mg_lockAllBuffers = ATOMIC_FLAG_INIT;


    /**
     * @brief Busy blocks infinite until the buffer is unlocked
     * @param pHandle - Pointer to the buffer
     */
    static inline void waitAndLockConcurrentAccess();
    
    /**
     * @brief unlocks the buffer
     * @param pHandle - Pointer to the buffer
     */
    static inline void unlockConcurrentAccess();
    
    static inline void waitAndLockConcurrentAccess()
    {
      // Atomic load
      int dummy = 0;
      while (atomic_flag_test_and_set(&mg_lockAllBuffers))
      {
        // busy wait
        dummy++;
      }
    }
    
    
    static inline void unlockConcurrentAccess()
    {
      atomic_flag_clear(&mg_lockAllBuffers);
    }

However GCCs thread sanitiser is complaining about a data race within waitAndLockConcurrentAccess(). But I do not know why.

Read of size 8 at 0x7f3e05b45ed0 by thread T2:
Previous write of size 8 at 0x7f3e05b45ed0 by thread T1:

Could you help we out here?

thx guys