Re: [patch v3 08/36] x86/smpboot: Split up native_cpu_up() into separate phases and document them

[Thomas Gleixner <tglx@xxxxxxxxxxxxx>]

On Tue, May 09 2023 at 12:31, Peter Zijlstra wrote:
> On Mon, May 08, 2023 at 09:43:39PM +0200, Thomas Gleixner wrote:
>> +    /*
>> +     * Wait for the AP to mark itself online, so the core caller
>> +     * can drop sparse_irq_lock.
>> +     */
>> +    while (!cpu_online(cpu))
>> +            schedule();
>> +}
>
> These schedule() loops make me itch... this is basically Ye Olde yield()
> loop with all it's known 'benefits'.
>
> Now, I don't think it's horribly broken, we're explicitly waiting on
> another CPU and can't have priority inversions, but yuck!
>
> It could all be somewhat cleaned up with wait_var_event{_timeout}() and
> wake_up_var(), but I'm really not sure that's worth it. But at least it
> requires a comment to justify.

All of them are gone with the later patches and the control CPU will
just go straight to wait for the completion in the core code.