Query on skip_onerr field in struct cpuhp_step

Query on skip_onerr field in struct cpuhp_step

[Mukesh Ojha]

Hi All,

This is about one of field in struct cpuhp_step

* @skip_onerr: Do not invoke the functions on error rollback
  *              Will go away once the notifiers are gone
         bool                    skip_onerr;

Why this field was introduced, can anyone explain the history behind ? 
As i don't see it being set from anywhere

If it is specifically was dependent on notifiers, did we missed to 
remove it as the notifiers are gone or the usecase still there?
And If it is still valid, in what scenario i would not want my callback 
to be called in rollback path?

Thanks,
Mukesh

Re: Query on skip_onerr field in struct cpuhp_step

[Thomas Gleixner]

[-- Attachment #1: Type: text/plain, Size: 597 bytes --]

On Tue, 21 Aug 2018, Mukesh Ojha wrote:
> Hi All,
> 
> This is about one of field in struct cpuhp_step
> 
> * @skip_onerr: Do not invoke the functions on error rollback
>  *              Will go away once the notifiers are gone
>         bool                    skip_onerr;
> 
> Why this field was introduced, can anyone explain the history behind ? As i
> don't see it being set from anywhere
> 
> If it is specifically was dependent on notifiers, did we missed to remove it
> as the notifiers are gone or the usecase still there?

As the comment says....

Re: Query on skip_onerr field in struct cpuhp_step

[Mukesh Ojha]

On 8/21/2018 7:27 PM, Thomas Gleixner wrote:
> On Tue, 21 Aug 2018, Mukesh Ojha wrote:
>> Hi All,
>>
>> This is about one of field in struct cpuhp_step
>>
>> * @skip_onerr: Do not invoke the functions on error rollback
>>   *              Will go away once the notifiers are gone
>>          bool                    skip_onerr;
>>
>> Why this field was introduced, can anyone explain the history behind ? As i
>> don't see it being set from anywhere
>>
>> If it is specifically was dependent on notifiers, did we missed to remove it
>> as the notifiers are gone or the usecase still there?
> As the comment says....

Thanks for your reply
Sorry, for further question but i did not get the comment ..did you 
mean, we can remove this from the structure as notifier are
not their now. ?

Also if you are aware with the history, can you tell , why would i want 
to avoid a call in undo_cpu_up path?

Thanks,
Mukesh

Re: Query on skip_onerr field in struct cpuhp_step

[Thomas Gleixner]

On Tue, 21 Aug 2018, Mukesh Ojha wrote:
> On 8/21/2018 7:27 PM, Thomas Gleixner wrote:
> > > If it is specifically was dependent on notifiers, did we missed to remove
> > > it as the notifiers are gone or the usecase still there?
> > As the comment says....
> 
> Thanks for your reply
> Sorry, for further question but i did not get the comment ..did you mean, we
> can remove this from the structure as notifier are not their now. ?

Yes, that's why the comment for that member says: Remove after conversion.

> Also if you are aware with the history, can you tell , why would i want to
> avoid a call in undo_cpu_up path?

Because the notifiers worked completely differently. Go back in the
git history where the initial conversion from notifiers to the state
machine happened.

Thanks,

	tglx