Re: [PATCH 2/6] mm: memory_hotplug: use more folio in do_migrate_range()

[Kefeng Wang <wangkefeng.wang@huawei.com>]

On 2024/3/27 23:58, Matthew Wilcox wrote:
> On Wed, Mar 27, 2024 at 11:10:48AM -0400, Zi Yan wrote:
>> On 27 Mar 2024, at 10:54, Matthew Wilcox wrote:
>>> How much of this is safe without a refcount on the folio?
>>
>> folio_to_pfn() should be fine, isolate_hugetlb() checks the folio
>> under hugetlb_lock, but folio_nr_pages() might return a bogus number
>> that makes pfn go beyond end_pfn and ends for loop early. The code
>> below increases the refcount, so it might be better to move this
>> part of the code after refcount is increased.
> 
> I really want to instill a little bit of fear in Kefeng.
> 
> This is all really subtle code because it's running without a refcount.
> I've mostly stayed away from it because I know that I don't know what
> I'm doing.  Kefeng has no idea that he doesn't know what he's doing.
> 
> And so we get these patches, and they're sometimes subtly wrong, and it
> takes a lot of arguing and revision and thinking and Kefeng is doing
> very little of the thinking part!
> 
> Kefeng, please stick to working on code that you understand.  Or take
> the time to learn code you don't understand before sending patches to
> it.  This kind of Leeroy Jenkins approach to development is not good.

Oh, I remember your reminder and be in awe of changes and try to think
more about changes, for this one, I only convert page to folio after
refcount increased with get_page_unless_zero(), and as replied to Zi,
PageHWPoison part need more consideration and this one only aims to
remove isolate_lru/movable_page, so don't touch the page before
get_page_unless_zero().

Thanks for your review and help all the time.