[EXT] Re: [PATCH 2/2] block: remove the per-bio/request write hint.

[Jens Axboe]

On 3/15/22 9:36 AM, Luca Porzio (lporzio) wrote:
>>
>> This isn't some setup to solicit votes on who supports what. If the code isn't
>> upstream, it by definition doesn't exist to the kernel. No amount of "we're
>> also interested in this" changes that.
>>
>> What I wrote earlier still applies - whoever is interested in supporting lifetime
>> hints should submit that code upstream. The existing patchset to clean this
>> up doesn't change that process AT ALL. As mentioned, the only difference is
>> what the baseline looks like in terms of what the patchset is based on.
>>
> 
> Jens, 
> 
> Actually we might work to issue a patch and revert the patch plus add
> the code that Bean and Bart mentioned which is currently Android only.
> The reason it has not been done before is because for now it's not
> production yet but it may soon be that case.
> 
> Would this patch revert be an option and accepted as a closure for
> this discussion?

What patch revert? It's not clear to me which patch you're talking about
here. If you're talking about the "remove the per-bio/request write
hint" patch, then no, that's certainly not being reverted. See previous
replies I made and also below for why, and let's please stop beating
this dead horse.

> Another option (which I actually prefer), if I ask for a MM & Storage
> BoF discussion on storage hints where I can show you the status of
> temperature management and my studies on how this is beneficial for
> storage devices. 

As long as it's accompanied by code that implements it, then that would
be fine.

> Would this be more beneficial and maybe get some wider consensus on
> the write hints?
> 
> After that consensus reverting (or agreeing on a new approach) will be
> easier.

As I've said multiple times, whenever code is available, it'll be
reviewed and discussed. I don't like to discuss hypotheticals as too
many times in the past there's a promise made and expectations built
only for nothing to materialize. As it stands, the only in-kernel user
of the hints is gone, and that means that the support code is being
removed. We NEVER keep code in the kernel that doesn't have a user, as
it can't get tested.

Submit your patches when they are ready, it really has no bearing on the
currently queued up changes to write hints.

-- 
Jens Axboe