MARS could help here (for some parts)

I think there are different main trunks to follow:

1) Short term: SMR drives need to be established in their market niche (regardless how big this niche might be at the beginning).

If SMR cannot gain their own market niche, the following trunks would be pointless. So I think this is prio #1.

Here MARS could help at the block layer, needing only minimal modifications itself (quick win).

Altering higher layers will need some time anyway (e.g. major developments at the FS layer are measured in decades nowadays).

2) Medium term: adapt block layer components (not limited to MARS) to the specifics of SMR.

This is easier than 3) because of lower complexity.

Please talk with me, I have some ideas about this.

Best would be a discussion at some Linux conference with all block layer people, changing the title to "Block Layer Support for SMR devices".

3) Long term: FS layer adaptations, as already started the discussion here.

Best after SMR would have really established their niche. This would be a much better motivation for long-lasting tedious work ;)