Re: [RFC] iomap: fix race between readahead and direct write

To: Gao Xiang <gaoxiang25@xxxxxxxxxx>
Subject: Re: [RFC] iomap: fix race between readahead and direct write
From: "yukuai (C)" <yukuai3@xxxxxxxxxx>
Date: Sun, 19 Jan 2020 11:15:18 +0800
Cc: Matthew Wilcox <willy@xxxxxxxxxxxxx>, <hch@xxxxxxxxxxxxx>, <darrick.wong@xxxxxxxxxx>, <linux-xfs@xxxxxxxxxxxxxxx>, <linux-fsdevel@xxxxxxxxxxxxxxx>, <linux-kernel@xxxxxxxxxxxxxxx>, <houtao1@xxxxxxxxxx>, <zhengbin13@xxxxxxxxxx>, <yi.zhang@xxxxxxxxxx>
In-reply-to: <20200119030123.GA223124@architecture4>
User-agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:60.0) Gecko/20100101 Thunderbird/60.8.0



On 2020/1/19 11:01, Gao Xiang wrote:

IMO, if use_list == true, it will call read_pages -> .readpages
and go just like the current implementation.


The problem is that in the current implementatin, iomap_next_page()
add page to page cache and it is replaced with readahead_page()

Thanks!
Yu Kuai

References:
- [RFC] iomap: fix race between readahead and direct write
  - From: yu kuai
- Re: [RFC] iomap: fix race between readahead and direct write
  - From: Matthew Wilcox
- Re: [RFC] iomap: fix race between readahead and direct write
  - From: yukuai (C)
- Re: [RFC] iomap: fix race between readahead and direct write
  - From: Matthew Wilcox
- Re: [RFC] iomap: fix race between readahead and direct write
  - From: yukuai (C)
- Re: [RFC] iomap: fix race between readahead and direct write
  - From: Gao Xiang