Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Thu, 30 Jul 2020 04:19:34 +0100
From:   Matthew Wilcox <willy@infradead.org>
To:     Yu Kuai <yukuai3@huawei.com>
Cc:     hch@infradead.org, darrick.wong@oracle.com,
        linux-xfs@vger.kernel.org, linux-fsdevel@vger.kernel.org,
        linux-kernel@vger.kernel.org, yi.zhang@huawei.com
Subject: Re: [RFC PATCH] iomap: add support to track dirty state of sub pages
Message-ID: <20200730031934.GA23808@casper.infradead.org>
References: <20200730011901.2840886-1-yukuai3@huawei.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20200730011901.2840886-1-yukuai3@huawei.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, Jul 30, 2020 at 09:19:01AM +0800, Yu Kuai wrote:
> +++ b/fs/iomap/buffered-io.c
> @@ -29,7 +29,9 @@ struct iomap_page {
>  	atomic_t		read_count;
>  	atomic_t		write_count;
>  	spinlock_t		uptodate_lock;
> +	spinlock_t		dirty_lock;

No need for a separate spinlock.  Just rename uptodate_lock.  Maybe
'bitmap_lock'.

>  	DECLARE_BITMAP(uptodate, PAGE_SIZE / 512);
> +	DECLARE_BITMAP(dirty, PAGE_SIZE / 512);

This is inefficient and poses difficulties for the THP patchset.
Maybe let the discussion on removing the ->uptodate array finish
before posting another patch for review?

> +static void
> +iomap_iop_set_or_clear_range_dirty(
> +	struct page *page,
> +	unsigned int off,
> +	unsigned int len,
> +	bool is_set)

Please follow normal kernel programming style.  This isn't XFS.
Also 'set or clear' with a bool to indicate which to do is horrible
style.  Separate functions!

> @@ -1386,7 +1432,8 @@ iomap_writepage_map(struct iomap_writepage_ctx *wpc,
>  	for (i = 0, file_offset = page_offset(page);
>  	     i < (PAGE_SIZE >> inode->i_blkbits) && file_offset < end_offset;
>  	     i++, file_offset += len) {
> -		if (iop && !test_bit(i, iop->uptodate))
> +		if (iop && (!test_bit(i, iop->uptodate) ||
> +		    !test_bit(i, iop->dirty)))
>  			continue;

Surely we don't need to test ->uptodate here at all.  Why would we write
back a block which isn't dirty?