Re: [PATCH 1/4] vmstat: remove zone->lock from walk_zones_in_node

[KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>]

> > Thanks lots comments. 
> > hmm.. I'd like to clarily your point. My point is memory-hotplug don't take zone lock,
> > then zone lock doesn't protect anything. so we have two option
> > 
> > 1) Add zone lock to memroy-hotplug
> > 2) Remove zone lock from zoneinfo
> > 
> > I thought (2) is sufficient. Do you mean you prefer to (1)? Or you prefer to ignore rarely event
> > (of cource, memory hotplug is rarely)?
> > 
> 
> I think (2) will make zoneinfo harder to use for examining all the counters
> properly as I explained above. I haven't looked at memory-hotplug in a
> while but IIRC, fields like present_pages should be protected by a lock on
> the pgdat and a seq lock on the zone. If this is not true at the moment,
> it is a problem.
> 
> For the free lists, memory hotplug should be taking the zone->lock properly as
> the final stage of onlining memory is to walk the sections being hot-added,
> init the memmap and then __free_page() each page individually - i.e. the
> normal free path.
> 
> So, if memory hotplug is not protected by proper locking, it's not intentional.

ok, I drop this patch. thanks.



--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>