Re: Path forward for Virtualized Swap?

[Shakeel Butt]

On Fri, Sep 25, 2026 at 09:15:17PM +0800, Kairui Song wrote:
[...]

Thanks Kairui for very clear description of what you want and looking for.
After reading your email, I think you want memsw support in v2. Currently you
somehow are implementing memsw using memory.current and memory.swap.current.
(I am assuming you need to subtract swapcache to avoid double counting.)
To conclude, you want to use vswap (the aim is it to default) but you still want
your way to implementing memsw keep working.

Please correct me if I missed something.

I am wondering if you really want memsw, why not propose memsw. Someone from
tencent proposed memsw [1] previously, so why not continue that effort? I am
pretty sure folks from Google (people who actually work on prodkernel there)
will be happy to collaborate with you on that.

[1] https://lore.kernel.org/20250319064148.774406-1-jingxiangzeng.cas@xxxxxxxxx/

Setting aside the memsw conversation, I think we should go back to the actual
vswap conversation and list down what we agree on and where we still need
discussion and decision. (I will take a stab at it and start a separate thread
but still on this email and will use AI for the summary.)