So the ideal behaviour would be: - for most processes no SWAP - for tmpfs, use RAM until a quota - for tmpfs, start using a swapfile above that quota Ch

So the ideal behaviour would be: - for most processes no SWAP - for tmpfs, use RAM until a quota - for tmpfs, start using a swapfile above that quota ChatGPT doesn't think it is achievable, though it thinks cgroup2 can achieve something similar.