Re: [PATCH v21 1/4] mm: add VM_DROPPABLE for designating always lazily freeable mappings

[Jason A. Donenfeld]

On Wed, Jul 10, 2024 at 05:05:06AM +0200, David Hildenbrand wrote:
> On 09.07.24 04:17, Jason A. Donenfeld wrote:
> > Hi David,
> > 
> > On Mon, Jul 08, 2024 at 10:21:09PM +0200, David Hildenbrand wrote:
> >> BTW, I was just trying to understand how MADV_FREE + MAP_DROPPABLE would
> >> behave without any swap space around.
> >>
> >> Did you experiment with that?
> > 
> > You mean on a system without any swap configured? That's actually my
> > primary test environment for this. It behaves as expected: when ram
> > fills up and the scanner is trying to reclaim what it can,
> > folio_test_swapbacked(folio) is false, and the memory gets freed. After,
> > reads fault in a zero page. So it's working as expected.
> 
> Okay, just to be clear: no swap/zram/zswap. The reclaim code regarding 
> not scanning anonymous memory without swap was a bit confusing.

Right, no swap, as boring a system as can be. I've experimented with
that behavior on my swap-less 64GB thinkpad, as well as on little
special purpose VMs, where I hacked the VM_DROPPABLE test code into the
wireguard test suite.

Jason