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:   Tue, 11 Apr 2023 20:51:24 +0530
Message-Id: <87wn2izbpn.fsf@doe.com>
From:   Ritesh Harjani (IBM) <ritesh.list@gmail.com>
To:     Christoph Hellwig <hch@infradead.org>
Cc:     linux-fsdevel@vger.kernel.org, linux-ext4@vger.kernel.org,
        Jan Kara <jack@suse.cz>, Christoph Hellwig <hch@infradead.org>,
        "Darrick J . Wong" <djwong@kernel.org>,
        Ojaswin Mujoo <ojaswin@linux.ibm.com>
Subject: Re: [RFCv2 5/8] ext2: Move direct-io to use iomap
In-Reply-To: <ZDT0JFmwg/9ijdcv@infradead.org>
Precedence: bulk

Christoph Hellwig <hch@infradead.org> writes:

> On Tue, Apr 11, 2023 at 10:51:53AM +0530, Ritesh Harjani (IBM) wrote:
>> +extern void ext2_write_failed(struct address_space *mapping, loff_t to);
>
> No need for the extern.
>

Sure will drop it.

>> +	/* handle case for partial write and for fallback to buffered write */
>> +	if (ret >= 0 && iov_iter_count(from)) {
>> +		loff_t pos, endbyte;
>> +		ssize_t status;
>> +		int ret2;
>> +
>> +		iocb->ki_flags &= ~IOCB_DIRECT;
>> +		pos = iocb->ki_pos;
>> +		status = generic_perform_write(iocb, from);
>> +		if (unlikely(status < 0)) {
>> +			ret = status;
>> +			goto out_unlock;
>> +		}
>> +
>> +		iocb->ki_pos += status;
>> +		ret += status;
>> +		endbyte = pos + status - 1;
>> +		ret2 = filemap_write_and_wait_range(inode->i_mapping, pos,
>> +						    endbyte);
>> +		if (!ret2)
>> +			invalidate_mapping_pages(inode->i_mapping,
>> +						 pos >> PAGE_SHIFT,
>> +						 endbyte >> PAGE_SHIFT);
>> +		if (ret > 0)
>> +			generic_write_sync(iocb, ret);
>> +	}
>
> Nit, but to me it would seem cleaner if all the fallback handling
> was moved into a separate helper function.  Or in fact by not
> using generic_file_write_iter even for buffered I/O and at doing
> the pre-I/O checks and the final generic_write_sync in common code in
> ext2 for direct and buffered I/O.
>

Make sense. However, since we are on the path to modify ext2 buffered-io
code as well to move to iomap interface, I wouldn't bother too much as
of now for this code as, all of this is going to go away anyways.


>> +	/*
>> +	 * For writes that could fill holes inside i_size on a
>> +	 * DIO_SKIP_HOLES filesystem we forbid block creations: only
>> +	 * overwrites are permitted.
>> +	 */
>> +	if ((flags & IOMAP_DIRECT) && (first_block << blkbits < i_size_read(inode)))
>> +		create = 0;
>
> No need for braes around the < operation, but I think you might need
> them around the shift.

left-shift has a higher precedence. But let me make it more clear in
next rev.

>
> Also an overly long line here.
>

Sure, will see to it.

>> +	if ((flags & IOMAP_WRITE) && (offset + length > i_size_read(inode)))
>
> No need for the second set of inner braces here either.

It's just avoids any confusion this way.

-ritesh