Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH] ubifs: Don't discard nodes in recovery when ecc err
 detected
To:     Richard Weinberger <richard.weinberger@gmail.com>
CC:     Richard Weinberger <richard@nod.at>,
        Sascha Hauer <s.hauer@pengutronix.de>,
        "zhangyi (F)" <yi.zhang@huawei.com>,
        <linux-mtd@lists.infradead.org>,
        LKML <linux-kernel@vger.kernel.org>
References: <1582293853-136727-1-git-send-email-chengzhihao1@huawei.com>
 <CAFLxGvyJdWcXQt3H2aknTuGhCJpV5YvAbW_wuHfs3m+KcNSjtw@mail.gmail.com>
From:   Zhihao Cheng <chengzhihao1@huawei.com>
Message-ID: <58b11ca2-6b91-52b3-bc75-d44abb202cfb@huawei.com>
Date:   Mon, 2 Mar 2020 11:58:35 +0800
User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:68.0) Gecko/20100101
 Thunderbird/68.5.0
MIME-Version: 1.0
In-Reply-To: <CAFLxGvyJdWcXQt3H2aknTuGhCJpV5YvAbW_wuHfs3m+KcNSjtw@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"; format=flowed
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

在 2020/3/2 4:46, Richard Weinberger 写道:
> Zhihao Cheng,
>
> On Fri, Feb 21, 2020 at 2:57 PM Zhihao Cheng <chengzhihao1@huawei.com> wrote:
>> The following process will lead TNC to find no corresponding inode node
>> (Reproduce method see Link):
> Please help me to understand what exactly is going on.
>
>>    1. Garbage collection.
>>       1) move valid inode nodes from leb A to leb B
>>          (The leb number of B has been written as GC type bud node in log)
>>       2) unmap leb A, and corresponding peb is erased
>>          (GCed inode nodes exist only on leb B)
> At this point all valid nodes are written to LEB B, right?
Yes.
>
>>    2. Poweroff. A node near the end of the LEB is corrupted before power
>>       on, which is uncorrectable error of ECC.
> If writing nodes to B has finished, these pages should be stable.
> How can a power-cut affect the pages where these valid nodes sit?
I mean, the uncorrectable ECC error is caused by hardware which may lead 
to corrupted nodes detected in UBIFS. I found uncorretable ECC errors on 
my NAND, in the environment of high temperature and humidity.

At present, UBIFS ignores all EBADMSG errors, so the corrupted node is 
only considered in being caused by unfinished writing. I think UBIFS 
should consider the corrupted area caused by ECC errors in process 
ubifs_recover_leb(). no_more_nodes() will skip a read-write unit. Maybe 
the corrupted area is skipped.