Re: [RFC PATCH 0/5] Rust block device driver API and null block driver

From: Andreas Hindborg
Date: Sun Mar 17 2024 - 03:10:16 EST


Matthew Wilcox <willy@xxxxxxxxxxxxx> writes:

> On Wed, Mar 13, 2024 at 12:05:07PM +0100, Andreas Hindborg wrote:
>> - Adopted the folio abstraction where applicable
>
> I don't think this was the correct move. The memory you're managing
> with folios is only used for storing the data being stored in the blkdev.
> It's not mapped to userspace, it doesn't make use of the flags (locked,
> uptodate, writeback, etc), it doesn't need an LRU list, a mapping,
> an index, a refcount or memcg.
>
> I think you should use pages instead of folios to keep a handle on
> this memory. Apologies for the confusion; we're in the middle of a
> years-long transition from the overly complex and overused struct page
> to splitting it into different data types for different purposes.

Ok, thanks. I'll swap it back.

I was under the impression that using folios also give the advantage
that handles are always head pages. No need to worry about head/tail
pages. If the driver moves to use higher order pages for larger block
sizes, would it then make sense with folios, or are page still
preferred?

> More detail on this here, if you're interested:
> https://kernelnewbies.org/MatthewWilcox/Memdescs

Thanks, that is useful.

Best regards,
Andreas