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;
Message-ID: <9d3ad717-e3c2-69c8-4ef0-3f5f24c4ae1c@huawei.com>
Date:   Thu, 30 Mar 2023 10:56:59 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:91.0) Gecko/20100101
 Thunderbird/91.3.2
Subject: Re: [PATCH 0/2] Add some msg for io error
To:     Theodore Ts'o <tytso@mit.edu>
CC:     <linux-ext4@vger.kernel.org>, <linfeilong@huawei.com>,
        <louhongxiang@huawei.com>, <liuzhiqiang26@huawei.com>
References: <20230325065652.2111384-1-zhanchengbin1@huawei.com>
 <20230326143128.GA436186@mit.edu>
From:   zhanchengbin <zhanchengbin1@huawei.com>
In-Reply-To: <20230326143128.GA436186@mit.edu>
Content-Type: text/plain; charset="UTF-8"; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk


On 2023/3/26 22:31, Theodore Ts'o wrote:
> On Sat, Mar 25, 2023 at 02:56:50PM +0800, zhanchengbin wrote:
>> If there is an EIO during the process of fsck, the user can be notified of it.
> 
> Can you identify a code path where the user is *not* getting notified
> while e2fsck is running without this patch series?
> 
> The unix_io.c module calls fsync() through unix_flush() only.  When
> unix_write_byte() calls flush_cached blocks(), if the read or write
> system call fails, the error will be returned to the caller of
> flush_cached_byte(), and the unix_write_byte() will return the error
> back to the caller (in this case, e2fsck).
> 

io_channel_flush and io_channel_write_byte do have return values, but
they may not necessarily be checked at their calling points. As in the
following path:

e2fsck_run_ext3_journal
  ext2fs_flush // Ignore errors.
   ext2fs_flush2
    io_channel_flush
  ext2fs_mmp_stop // Ignore errors.
   ext2fs_mmp_write
    io_channel_flush

ext2fs_flush // Many calls ignore errors.
  ext2fs_flush2
   write_primary_superblock
    io_channel_write_byte

Thanks,
  - bin.

> So in both cases, e2fsck checks the error return from ext2fs_flush()
> (which is the only place where write_byte gets called) and
> io_channel->flush(), and so the user will get some kind of error
> report already.
> 
> The error message might not identify exactly what I/O failed, but the
> "Error sync" message that this commit series provides is not going to
> be much better.
> 
> 						- Ted
> 
> .
>