Re: [PATCH] rust: slab: add basic slab module

[Danilo Krummrich]

On Thu Sep 25, 2025 at 8:15 PM CEST, Danilo Krummrich wrote:
> On Thu Sep 25, 2025 at 8:05 PM CEST, Elijah wrote:
>> On 9/25/2025 10:43 AM, Danilo Krummrich wrote:
>>
>>  > Anyways, I'd also like to hear some more opinions, especially 
>> regarding (4), as
>>  > mentioned.
>>
>> ok I just looked it over. I think (4) isn't really viable, the cache 
>> pointer storage overhead is probably too much (in addition to some 
>> complex issues).
>
> Probably, given that kmemcache is meant to be used when we have a lot of (small)
> allocations.
>
> However, I'd like to hear from the MM folks, if there may be options in the
> existing infrastructure, or if it's really a dead end.
>
> What are the other complex issues you are referring to?
>
>> I don't really like having a macro (so (3) basically) 
>> and I think there are issues with non-static caches, I don't know 
>> specifically though. obviously (1) doesn't work. (2) is exactly what I 
>> was thinking specifically for rbtree. IMO the only useful options are 
>> (2) and (3). I would say maybe implement (2) first and if you want 
>> macro-generated ZST allocators they can wrap whatever that Box ends up 
>> being.
>
> I'm not sold on duplicating the Box code just yet. What issues are you referring
> to?

Also, if we argue that (4) has a problem with scalability, then (2) has the
problem as well.