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: <d82be232-8df4-86ff-ade5-4b9864e9cf40@huawei.com>
Date:   Wed, 15 Jun 2022 09:34:31 +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
From:   "chenxiaosong (A)" <chenxiaosong2@huawei.com>
Subject: Re: [PATCH -next,v2] NFS: report and clear ENOSPC/EFBIG/EDQUOT
 writeback error on close() file
To:     Trond Myklebust <trondmy@hammerspace.com>,
        "anna@kernel.org" <anna@kernel.org>
CC:     "linux-nfs@vger.kernel.org" <linux-nfs@vger.kernel.org>,
        "liuyongqiang13@huawei.com" <liuyongqiang13@huawei.com>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "yi.zhang@huawei.com" <yi.zhang@huawei.com>,
        "zhangxiaoxu5@huawei.com" <zhangxiaoxu5@huawei.com>
References: <20220614152817.271507-1-chenxiaosong2@huawei.com>
 <806ae30d4d53886e7d394cc9abb0b2045a314f01.camel@hammerspace.com>
In-Reply-To: <806ae30d4d53886e7d394cc9abb0b2045a314f01.camel@hammerspace.com>
Content-Type: text/plain; charset="UTF-8"; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk

在 2022/6/15 3:48, Trond Myklebust 写道:
> NACK. How many times do I have to repeat that we do NOT clear the error
> log in flush()?
> 

close(2) manpage described:

> ENOSPC, EDQUOT: On NFS, these errors are not normally reported
> against the first write which exceeds the available storage space,
> but instead against a subsequent write(2), fsync(2), or close(2).

> A  careful programmer will check the return value of close(), since
> it is quite possible that errors on a previous write(2) operation are
> reported only on the final close() that releases the open file
> description.  Failing to check the return value when closing a file
> may lead to silent loss of data.  This can especially be observed 
> with NFS and with disk quota.

write(2) manpage described:

> Since Linux 4.13, errors from write-back come with a promise that they
> may be reported by subsequent.  write(2) requests, and will be
> reported by a subsequent fsync(2) (whether or not they were also
> reported by write(2)).

Both close(2) and write(2) manpage described: report writeback error 
(not clear error), especially the write(2) manpage described: will be 
reported by a subsequent fsync(2) whether or not they were also reported 
by write(2).

If ENOSPC/EFBIG/EDQUOT writeback error can be cleared on write(), maybe 
it is better to be cleared on close() instead of saving the error for 
next open().