Re: [PATCH v6 01/22] mm/swap: decouple swap cache from physical swap infrastructure

[Nhat Pham]

On Mon, May 11, 2026 at 3:46 PM Yosry Ahmed <yosry@xxxxxxxxxx> wrote:
>
> On Tue, May 05, 2026 at 08:38:30AM -0700, Nhat Pham wrote:
> > When we virtualize the swap space, we will manage swap cache at the
> > virtual swap layer. To prepare for this, decouple swap cache from
> > physical swap infrastructure.
> >
> > We will also remove all the swap cache related helpers of swap table. We
> > will keep the rest of the swap table infrastructure, which will be
> > repurposed to serve as the rmap (physical -> virtual swap mapping)
> > later.
>
> I didn't look through the entire series, but let me ask the same
> high-level question I asked before. Instead of moving things out of the
> swap table, why not reuse the swap table as the representation of the
> virtual swap space? Seems like most/all metadata is already moved there
> in a nice concise format.

The honest answer is I wasn't sure it would work, so I was hacking
quietly a prototype on my own time :)

I finally got something that survives stress-ng and constant
memory.reclaim thrown at it though. I figured I should send it out to
get feedback before digging myself deeper into that hole:

https://lore.kernel.org/all/20260528212955.1912856-1-nphamcs@xxxxxxxxx/

There is still a small problem left (the metadata duplication issue
that Johannes brought up). It is potentially fixable, but I haven't
actually tried it out yet, so I don't want to overstate here. But take
a look at it and let me know how you feel about this alternative
approach!