Re: [PATCH WIP v1 07/20] mm/rmap_id: track if one ore multiple MMs map a partially-mappable folio

[Nadav Amit]

> On Dec 18, 2023, at 4:04 PM, David Hildenbrand <david@xxxxxxxxxx> wrote:
> 
> But adding 1 "0" bit is not sufficient for handling order-2 folios (P = 4), only for handling order-1 folios. So what the current approach does is the following (P = 4):
> 
> RMAP-ID |       | Subid |
> -----------------------------------
> 0       | 0000 | 0     | 0000 0000
> 1       | 0001 | 1     | 0000 0001
> 2       | 0010 | 5     | 0000 0101
> 3       | 0011 | 6     | 0000 0110
> 4       | 0100 | 25    | 0001 1001
> 5       | 0101 | 26    | 0001 1010
> 6       | 0110 | 30    | 0001 1110
> 7       | 0111 | 31    | 0001 1111
> 8       | 1000 | 125   | 0111 1101
> 9       | 1001 | 126   | 0111 1110
> 10      | 1010 | 130   | 1000 0010
> 11      | 1011 | 131   | 1000 0011
> 12      | 1100 | 150   | 1001 0110
> 13      | 1101 | 151   | 1001 0111
> 14      | 1110 | 155   | 1001 1011
> 15      | 1111 | 156   | 1001 1100

Yes, of course. Silly me. You want to take advantage of the counter not
saturating for orders K where K-1 is not a power of 2.

I get your point. Not sure whether it worth the complexity though…