Re: [Xen-devel] [PATCH v3 1/6] xen, cpupool: correct error handling when removing cpu from cpupool

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 04.03.16 at 10:54, <JGross@xxxxxxxx> wrote:
> On 04/03/16 10:42, Jan Beulich wrote:
>>>>> On 03.03.16 at 17:48, <JGross@xxxxxxxx> wrote:
>>>      if ( !ret )
>>>      {
>>>          ret = schedule_cpu_switch(cpu, NULL);
>>>          if ( ret )
>>> -        {
>>>              cpumask_clear_cpu(cpu, &cpupool_free_cpus);
>>> -            goto out;
>>> +        else
>>> +        {
>>> +            cpupool_moving_cpu = -1;
>>> +            cpupool_put(cpupool_cpu_moving);
>>> +            cpupool_cpu_moving = NULL;
>>>          }
>>> -        cpupool_moving_cpu = -1;
>>> -        cpupool_put(cpupool_cpu_moving);
>>> -        cpupool_cpu_moving = NULL;
>>>      }
>>>  
>>>      for_each_domain_in_cpupool(d, c)
>> 
>> So this now also eliminates the bypass of the loop the beginning of
>> which we can still see here - is that really intended, or wouldn't
>> this better be made conditional upon !ret?
> 
> Did you look at the commit message? It is explaining exactly this
> behaviour.

Argh - I had looked, even twice (once before writing the
comment, and once before hitting the send button), but
nevertheless managed to not connect things together.
I'm sorry.

And I think this aspect (and only it) makes this a backport
candidate.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel