[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <ZoQvwkcL3uWONfzV@casper.infradead.org>
Date: Tue, 2 Jul 2024 17:50:10 +0100
From: Matthew Wilcox <willy@...radead.org>
To: Christoph Hellwig <hch@....de>
Cc: "Pankaj Raghav (Samsung)" <kernel@...kajraghav.com>,
david@...morbit.com, chandan.babu@...cle.com, djwong@...nel.org,
brauner@...nel.org, akpm@...ux-foundation.org,
linux-kernel@...r.kernel.org, yang@...amperecomputing.com,
linux-mm@...ck.org, john.g.garry@...cle.com,
linux-fsdevel@...r.kernel.org, hare@...e.de, p.raghav@...sung.com,
mcgrof@...nel.org, gost.dev@...sung.com, cl@...amperecomputing.com,
linux-xfs@...r.kernel.org, Zi Yan <zi.yan@...t.com>
Subject: Re: [PATCH v8 06/10] iomap: fix iomap_dio_zero() for fs bs > system
page size
On Tue, Jul 02, 2024 at 02:02:50PM +0200, Christoph Hellwig wrote:
> On Tue, Jul 02, 2024 at 10:15:56AM +0000, Pankaj Raghav (Samsung) wrote:
> > Willy suggested we could use raw pages as we don't need the metadata
> > from using a folio. [0]
>
> Ok, that feels weird but I'll defer to his opinion in that case.
Let me see if I can make you feel less weird about it, since I think
this is something that people should have a clear feeling about.
In the Glorious Future, when we've separated pages and folios from each
other, folios are conceptually memory that gets mapped to userspace.
They have refcounts, mapcounts, a pointer to a file's mapping or an anon
vma's anon_vma, an index within that object, an LRU list, a dirty flag,
a lock bit, and so on.
We don't need any of that here. We might choose to use a special memdesc
for accounting purposes, but there's no need to allocate a folio for it.
For now, leaving it as a plain allocation of pages seems like the smartest
option, and we can revisit in the future.
Powered by blists - more mailing lists