Again, not really this patch, but since I had to look at this code .... On Mon, May 08, 2023 at 09:43:39PM +0200, Thomas Gleixner wrote: > @@ -1048,60 +1066,89 @@ static int do_boot_cpu(int apicid, int c /* * AP might wait on cpu_callout_mask in cpu_init() with * cpu_initialized_mask set if previous attempt to online * it timed-out. Clear cpu_initialized_mask so that after * INIT/SIPI it could start with a clean state. */ cpumask_clear_cpu(cpu, cpu_initialized_mask); smp_mb(); ^^^ that barrier is weird too, cpumask_clear_cpu() is an atomic op and implies much the same (this is x86 after all). If you want to be super explicit about it write: smp_mb__after_atomic(); (which is a no-op) but then it still very much requires a comment as to what exactly it orders against what. /* * Wake up a CPU in difference cases: * - Use a method from the APIC driver if one defined, with wakeup * straight to 64-bit mode preferred over wakeup to RM. * Otherwise, > * - Use an INIT boot APIC message > */ > if (apic->wakeup_secondary_cpu_64) > + return apic->wakeup_secondary_cpu_64(apicid, start_ip); > else if (apic->wakeup_secondary_cpu) > + return apic->wakeup_secondary_cpu(apicid, start_ip); > > + return wakeup_secondary_cpu_via_init(apicid, start_ip); > +}
