Re: [PATCH v1 2/2] mm: remove extra ZONE_DEVICE struct page refcount

[Joao Martins <joao.m.martins@oracle.com>]

On 10/19/21 20:21, Dan Williams wrote:
> On Tue, Oct 19, 2021 at 9:02 AM Jason Gunthorpe <jgg@ziepe.ca> wrote:
>>
>> On Tue, Oct 19, 2021 at 04:13:34PM +0100, Joao Martins wrote:
>>> On 10/19/21 00:06, Jason Gunthorpe wrote:
>>>> On Mon, Oct 18, 2021 at 12:37:30PM -0700, Dan Williams wrote:
>>>>
>>>>>> device-dax uses PUD, along with TTM, they are the only places. I'm not
>>>>>> sure TTM is a real place though.
>>>>>
>>>>> I was setting device-dax aside because it can use Joao's changes to
>>>>> get compound-page support.
>>>>
>>>> Ideally, but that ideas in that patch series have been floating around
>>>> for a long time now..
>>>>
>>> The current status of the series misses a Rb on patches 6,7,10,12-14.
>>> Well, patch 8 too should now drop its tag, considering the latest
>>> discussion.
>>>
>>> If it helps moving things forward I could split my series further into:
>>>
>>> 1) the compound page introduction (patches 1-7) of my aforementioned series
>>> 2) vmemmap deduplication for memory gains (patches 9-14)
>>> 3) gup improvements (patch 8 and gup-slow improvements)
>>
>> I would split it, yes..
>>
>> I think we can see a general consensus that making compound_head/etc
>> work consistently with how THP uses it will provide value and
>> opportunity for optimization going forward.
>>

I'll go do that. Meanwhile, I'll wait a couple days for Dan to review the
dax subsystem patches (6 & 7), or otherwise send them over.

>>> Whats the benefit between preventing longterm at start
>>> versus only after mounting the filesystem? Or is the intended future purpose
>>> to pass more context into an holder potential future callback e.g. nack longterm
>>> pins on a page basis?
>>
>> I understood Dan's remark that the device-dax path allows
>> FOLL_LONGTERM and the FSDAX path does not ?
>>
>> Which, IIRC, today is signaled basd on vma properties and in all cases
>> fast-gup is denied.
> 
> Yeah, I forgot that 7af75561e171 eliminated any possibility of
> longterm-gup-fast for device-dax, let's not disturb that status quo.
> 
I am slightly confused by this comment -- the status quo is what we are
questioning here -- And we talked about changing that in the past too
(thread below), that longterm-gup-fast was an oversight that that commit
was only applicable to fsdax. [Maybe this is just my english confusion]

>>> Maybe we can start by at least not add any flags and just prevent
>>> FOLL_LONGTERM on fsdax -- which I guess was the original purpose of
>>> commit 7af75561e171 ("mm/gup: add FOLL_LONGTERM capability to GUP fast").
>>> This patch (which I can formally send) has a sketch of that (below scissors mark):
>>>
>>> https://lore.kernel.org/linux-mm/6a18179e-65f7-367d-89a9-d5162f10fef0@oracle.com/
>>
>> Yes, basically, whatever test we want for 'deny fast gup foll
>> longterm' is fine.
>>
>> Personally I'd like to see us move toward a set of flag specifying
>> each special behavior and not a collection of types that imply special
>> behaviors.
>>
>> Eg we have at least:
>>  - Block gup fast on foll_longterm
>>  - Capture the refcount ==1 and use the pgmap free hook
>>    (confusingly called page_is_devmap_managed())
>>  - Always use a swap entry
>>  - page->index/mapping are used in the usual file based way?
>>
>> Probably more things..
> 
> Yes, agree with the principle of reducing type-implied special casing.
> 
OK.

Moving from implicit devmap types to pgmap::flags is rather simple fixup.
And I suppose (respectivally) PGMAP_NO_PINF_LONGTERM, PGMAP_MANAGED_FREE_PAGE,
PGMAP_USE_SWAP_ENTRY, etc, etc.