Atomic 64 bit counter on ARM 32 bit MCU without locks

On a ARM Cortex M 32 bit, under C, I need to maintain a 64 bit counter. I want to avoid a race condition like:

start:         count = 0x00000000 ffffffffthreadA: increment count_low         count = 0x00000000 00000000threadB: increment count_low         count = 0x00000000 00000001threadA: if (!count_low) increment count_high         count = 0x00000000 00000001threadB: if (!count_low) increment count_high         count = 0x00000000 00000001

It's a simple counter which will be used only two ways: 1. Read val and 2. Increment.

In ARM assembly, this can be done without introducing any locks, via LDREX and STREX. However, I'd like to avoid assembly and use C.

Is the best way to do this simply to use atomic_uint_least64_t ? Will that work?

When I try making my counter atomic_uint_least64_t , I get linker errors undefined reference to__atomic_load_8', and when I try adding -latomic, I get cannot find "-latomic:" No such file or directory`.

I'm using arm-none-eabi-gcc on Linux to cross compile for Cortex M.