Received-SPF: pass (google.com: domain of linux-nfs-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <ff6de37e-899a-8b23-d1a9-bf11fde10e5d@huawei.com>
Date:   Wed, 15 Jun 2022 20:43:36 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:91.0) Gecko/20100101
 Thunderbird/91.2.1
To:     Trond Myklebust <trondmy@hammerspace.com>
CC:     "linux-nfs@vger.kernel.org" <linux-nfs@vger.kernel.org>,
        "zhangxiaoxu5@huawei.com" <zhangxiaoxu5@huawei.com>
From:   "chenxiaosong (A)" <chenxiaosong2@huawei.com>
Subject: Question about reporting unexpected wb err in nfs write()/flush()
Content-Type: text/plain; charset="UTF-8"; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hi Trond:

OK, I understand, I will not clear wb err in close(). I will not change 
rules that apply to all filesystems.

But if old wb err (such as -ERESTARTSYS or -EINTR, etc.) exist in struct 
address_space->wb_err, nfs_file_write() will always return the 
unexpected error. filemap_check_wb_err() will return the old error even 
if there is no new writeback error between filemap_sample_wb_err() and 
filemap_check_wb_err().
```c
    since = filemap_sample_wb_err() = 0 // never be error
      errseq_sample
        if (!(old & ERRSEQ_SEEN)) // nobody see the error
          return 0;
    nfs_wb_all // no new error
    error = filemap_check_wb_err(..., since) != 0 // unexpected error
```

I will fix this by store old err before filemap_sample_wb_err(), and 
restore old err after filemap_check_wb_err():
```c
   // Store the wb err
   old_err = file_check_and_advance_wb_err(file)
   since = filemap_sample_wb_err()
   nfs_wb_all // detect new wb err
   error = filemap_check_wb_err(..., since)
   // Restore old wb err if no new err is detected
   if (!error && old_err)
   errseq_set(&file->f_mapping->wb_err, old_err);
```

Is my fixes reasonable ?