Re: [PATCH 0/3] mm: remove zpool

[Vlastimil Babka <vbabka@suse.cz>]

On 9/4/25 16:11, Vitaly Wool wrote:
> On 9/4/25 12:13, Vlastimil Babka wrote:
>> On 9/4/25 11:33, Vitaly Wool wrote:
>>>> With zswap using zsmalloc directly, there are no more in-tree users of
>>>> this code. Remove it.
>>>>
>>>> Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
>>>
> 
> Things keep changing, and some of the proven solutions may not be a good 
> fit moving forward. While not suggesting that we should have a handful 
> of zpool backends just for the sake of variety, I'd like to emphasize 
> that there are good reasons to have zblock (especially the Rust one), 
> and there are good reasons to keep zsmalloc. That leads to the 
> conclusion that zpool should stay.

Johannes already suggested it's possible to do that by reimplementing
zsmalloc APIs without the runtime switch layer and choosing at config time.
It would also mean zram would be able to switch.

Your argument was "zpool API is neutral and well-defined" but we don't
really care about API/KABI/etc stability inside the kernel:
Documentation/process/stable-api-nonsense.rst
So that's not a sufficient argument.