Re: [PATCH v5 4/4] mm: Defer ZONE_DEVICE page initialization to the point where we init pgmap

From: Oscar Salvador
Date: Tue Oct 30 2018 - 04:06:07 EST

> Yes, the hotplug lock was part of the original issue. However that
> starts to drift into the area I believe Oscar was working on as a part
> of his patch set in encapsulating the move_pfn_range_to_zone and other
> calls that were contained in the hotplug lock into their own functions.

While reworking it for my patchset, I thought that we can move
out of hotplug lock.
But then I __think__ we would have to move init_currently_empty_zone() within
the span lock as zone->zone_start_pfn is being touched there.
At least that is what the zone locking rules say about it.

Since I saw that Dan was still reworking his patchset about unify HMM/devm code,
I just took one step back and I went simpler [1].
The main reason for backing off was I felt a bit demotivated due to
the lack of feedback,
and I did not want to interfer either with your work or Dan's work.
Plus I also was unsure about some other things like whether it is ok calling
kasan_add_zero_shadow/kasan_remove_zero_shadow out of the lock.
So I decided to make less changes in regard of HMM/devm.

Unfortunately, I did not get a lot of feedback there yet.
Just some reviews from David and a confirmation that fixes one of the
issues Jonathan reported [2].

> I was hoping to wait until after Dan's HMM patches and Oscar's changes
> had been sorted before I get into any further refactor of this specific
> code.

I plan to ping the series, but I wanted to give more time to people
since we are in the merge window now.