Received-SPF: pass (google.com: domain of linux-ext4-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 12 Apr 2023 19:32:26 +0530
Message-Id: <87ttxlyz9p.fsf@doe.com>
From:   Ritesh Harjani (IBM) <ritesh.list@gmail.com>
To:     Christoph Hellwig <hch@infradead.org>,
        Matthew Wilcox <willy@infradead.org>
Cc:     Christoph Hellwig <hch@infradead.org>,
        linux-fsdevel@vger.kernel.org, linux-ext4@vger.kernel.org,
        Jan Kara <jack@suse.cz>,
        "Darrick J . Wong" <djwong@kernel.org>,
        Ojaswin Mujoo <ojaswin@linux.ibm.com>
Subject: Re: [RFCv2 2/8] libfs: Add __generic_file_fsync_nolock implementation
In-Reply-To: <ZDaZR+zHcpUyNOND@infradead.org>
Precedence: bulk

Christoph Hellwig <hch@infradead.org> writes:

> On Tue, Apr 11, 2023 at 01:33:17PM +0100, Matthew Wilcox wrote:
>> On Mon, Apr 10, 2023 at 10:27:10PM -0700, Christoph Hellwig wrote:
>> > On Tue, Apr 11, 2023 at 10:51:50AM +0530, Ritesh Harjani (IBM) wrote:
>> > > +/**
>> > > + * __generic_file_fsync_nolock - generic fsync implementation for simple
>> > > + * filesystems with no inode lock
>> >
>> > No reallz need for the __ prefix in the name.
>>
>> It kind of makes sense though.
>>
>> generic_file_fsync does the flush
>> __generic_file_fsync doesn't do the flush
>> __generic_file_fsync_nolock doesn't do the flush and doesn't lock/unlock
>
> Indeed.  Part of it is that the naming is a bit horrible.
> Maybe it should move to buffer.c and be called generic_buffer_fsync,
> or generic_block_fsync which still wouldn't be perfect but match the
> buffer.c naming scheme.
>

Eventually it anyways needs some work to see if we can kill the lock
variant all together. I didn't do that in this series which is
focused on ext2 conversion of iomap.
So, if it's not that bad, I would like to keep both function
definitions at one place so that it can be worked out later.

>>
>> > > +extern int __generic_file_fsync_nolock(struct file *, loff_t, loff_t, int);
>> >
>> > No need for the extern.  And at least I personally prefer to spell out
>> > the parameter names to make the prototype much more readable.
>>
>> Agreed, although I make an exception for the 'struct file *'.  Naming that
>> parameter adds no value, but a plain int is just obscene.
>>
>> int __generic_file_fsync_nolock(struct file *, loff_t start, loff_t end,
>> 		bool datasync);
>
> While I agree that it's not needed for the file, leaving it out is a bit
> silly.
>

Sure. Will fix it.

>> (yes, the other variants don't use a bool for datasync, but they should)
>
> .. including the ->fsync prototype to make it work ..

Sure, this work should go as a seperate series.

-ritesh