Re: [PATCH 2/2] mm: Make swap_readpage() for SWP_FS_OPS use ->direct_IO() not ->readpage()

[Theodore Ts'o]

On Thu, Aug 12, 2021 at 07:14:37PM +0100, Matthew Wilcox wrote:
> 
> Well ... this would actually allow the filesystem to break COWs and
> allocate new blocks for holes.  Maybe you don't want to be doing that
> in a low-memory situation though ;-)

I'm not sure the benefits are worth the costs.  You'd have to handle
ENOSPC errors, and it would require some kind of metadata journal
transaction, which could potentially block for any number of reasons
(not just due to memory allocations, but because you're waiting for a
journal commit to complete).  As you say, doing that in a low-memory
situation seems to be unneeded complexity.

					- Ted