On Tue, 2022-02-01 at 13:56 +0100, Borislav Petkov wrote:
> On Tue, Feb 01, 2022 at 12:39:17PM +0000, David Woodhouse wrote:
> > In the top of my git tree, you can see a half-baked 'parallel part 2'
> > commit which introduces a new x86/cpu:wait-init cpuhp state that would
> > invoke do_wait_cpu_initialized() for each CPU in turn, which *would*
> > release them all into load_ucode_bsp() at the same time and have
> > precisely the problem you're describing.
> 
> The load_ucode_bsp() is the variant that runs on the boot CPU but
> yeah...

Right. Brain not fully online today.  Sorry.

> > Hm, not sure I see how that's protecting itself from someone
> > simultaneously echoing 1 > /sys/devices/system/cpu/cpu${SIBLING}/online
> 
> So
> 
> echo 1 > ../online
> 
> means onlining the sibling.
> 
> But reload_store() grabs the CPU hotplug lock *first* and *then* runs
> check_online_cpus() to see if all CPUs are online. It doesn't do the
> update if even one CPU is missing. You can't offline any CPU for the
> duration of the update...
> 
> So I guess you'd need to explain in more detail what protection hole
> you're seeing because I might be missing something here.

No, I'd just missed cpus_read_lock() because I was looking for
something else. My fault; it looks fine. Thanks.