Comment 6 for bug 1754038

I finally implemented a workaround to correct the problem:

in cpus.c : qemu_start_warp_timer(), in the "if (deadline > 0) { ... }" part, I added:

        CPUState *cpu;
        CPU_FOREACH(cpu) {
            atomic_mb_set(&cpu->exit_request, 1);
        }

I do not understand more than 5% of the code I am messing up, so this hack is probably broken...

Then I tested a bit more the code with different testcases... and I found a new bug when writing a reload value smaller than the current counter (the counter will then read as 0). It is due to this piece of code in armv7m_systick.c : systick_read() :

        /* The interrupt in triggered when the timer reaches zero.
           However the counter is not reloaded until the next clock
           tick. This is a hack to return zero during the first tick. */
        if (val > s->reload) {
            val = 0;
        }

Well this is not really a prb for me with normal code, and it looks like under control, but I can open another bug if needed.