Re: [PATCH] mm: hugetlb: simplify per-node sysfs creation and removal

[Muchun Song]

> On Aug 19, 2022, at 15:00, Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> 
> On Fri, Aug 19, 2022 at 02:44:13PM +0800, Muchun Song wrote:
>> 
>> 
>>> On Aug 19, 2022, at 14:32, Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>>> 
>>> On Fri, Aug 19, 2022 at 01:21:37PM +0800, Muchun Song wrote:
>>>> The following commit offload per-node sysfs creation and removal to a kworker and
>>>> did not say why it is needed.  And it also said "I don't know that this is
>>>> absolutely required".  It seems like the author was not sure as well.  Since it
>>>> only complicates the code, this patch will revert the changes to simplify the code.
>>>> 
>>>> 39da08cb074c ("hugetlb: offload per node attribute registrations")
>>> 
>>> Any specific reason why you did not cc: the original author of this
>>> commit, or anyone else on the patch?
>> 
>> OK. Cc Lee Schermerhorn.
> 
> He can't see the patch here, so there is no context.  Please resend the

Unluckily, his email is out of date, I cannot resend it to him.

> whole thing.  You also didn't copy the people who signed off on it (i.e.
> Andi), any reason why?

I can resend this to Andi. Why I didn’t send to them this time is because
I just follow the output of scripts/get_maintainer.pl. I will resend a new
one.

> 
>>>> We could use memory hotplug notifier to do per-node sysfs creation and removal
>>>> instead of inserting those operations to node registration and unregistration.
>>>> Then, it can reduce the code coupling between node.c and hugetlb.c.  Also, it can
>>>> simplify the code.
>>> 
>>> I do not think we had memory hotplug notifier back in 2009 when this
>>> commit was first written.
>> 
>> Maybe not. Commit 39da08cb074c is merger in 2009. However, hotplug notifier mechanism
>> is merged in 2006. The document is updated in 2007 (see commit 10020ca246c5).
>> 
>>> 
>>> How did you test this?  Did you use a HUGETLBFS system and verify that
>>> everything still works properly?  You are deleting a lot of code (always
>>> nice), but making sure everything is still operating the same is a good
>>> thing.
>> 
>> I really did the test (through a VM), it works properly.
> 
> How about on real hardware?  On a HUGE system with real hardware?  On a
> small system?

I have test a VM with 128 GB memory, it works properly. I cannot test it on
a real hardware since I didn’t have a real hardware with hotplug memory. I
am not sure if there is some ways to emulate a hotplug memory node.

Thanks.

> 
> thanks,
> 
> greg k-h