Re: [Xen-devel] [PATCH v5 1/4] xen/rcu: don't use stop_machine_run() for rcu_barrier()

[Jan Beulich <jbeulich@xxxxxxxx>]

On 13.03.2020 12:18, Jürgen Groß wrote:
> On 13.03.20 12:02, Julien Grall wrote:
>> On 12/03/2020 08:28, Juergen Gross wrote:
>>> +    for ( ;; )
>>> +    {
>>> +        if ( !atomic_read(&pending_count) && get_cpu_maps() )
>>> +        {
>>> +            n_cpus = num_online_cpus();
>>> +
>>> +            if ( atomic_cmpxchg(&pending_count, 0, n_cpus + 1) == 0 )
>>> +                break;
>>> +
>>> +            put_cpu_maps();
>>> +        }
>>> +
>>> +        process_pending_softirqs();
>>> +        cpu_relax();
>>> +    }
>>> +
>>> +    atomic_set(&cpu_count, n_cpus);
>>> +    cpumask_raise_softirq(&cpu_online_map, RCU_SOFTIRQ);
>>> +
>>> +    put_cpu_maps();
>>
>> If you put the CPU maps, wouldn't it be possible to have a CPU turned 
>> off? If not, can you add a comment in the code why this is safe?
> 
> Yes, you are right. This might be possible, even if rather
> unlikely as a cpu being removed has to be in idle already, so
> the pending softirq should be picked up rather fast.

I think that's not the main aspect. The CPU to be removed may
already be spinning in cpu_hotplug_begin() (and may in particular
also already be past the rcu_barrier() that Igor's patch is going
to put there).

Jan

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel