Received-SPF: pass (google.com: domain of linux-ext4+bounces-252-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Date: Thu, 30 Nov 2023 16:03:37 +0000
From: Matthew Wilcox <willy@infradead.org>
To: Ritesh Harjani <ritesh.list@gmail.com>
Cc: Jan Kara <jack@suse.cz>, Christoph Hellwig <hch@infradead.org>,
	linux-ext4@vger.kernel.org, linux-fsdevel@vger.kernel.org
Subject: Re: [RFC 2/3] ext2: Convert ext2 regular file buffered I/O to use
 iomap
Message-ID: <ZWiyWfEuWmPc+VuO@casper.infradead.org>
References: <20231130140859.hdgvf24ystz2ghdv@quack3>
 <878r6fi6jy.fsf@doe.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <878r6fi6jy.fsf@doe.com>

On Thu, Nov 30, 2023 at 09:20:41PM +0530, Ritesh Harjani wrote:
> OR, one other approach to this can be... 
> 
> IIUC, for a new block mapping for the same folio, the folio can be made to
> get invalidated once in between right? So should this be handled in folio
> cache somehow to know if for a given folio the underlying mapping might
> have changed for which iomap should again query  ->map_blocks() ? 
> ... can that also help against unnecessary re-validations against cached
> block mappings?

That's pretty much exactly what buffer heads do -- we'd see the mapped
bit was now clear.  Maybe we could have a bit in the iomap_folio_state
that is set once a map has been returned for this folio.  On the one
hand, this is exactly against the direction of iomap; to divorce the
filesystem state from the page cache state.  On the other hand, maybe
that wasn't a great idea and deserves to be reexamined.