Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH v2] btrfs: fix rw device counting in
 __btrfs_free_extra_devids
To:     dsterba@suse.cz, clm@fb.com, josef@toxicpanda.com,
        dsterba@suse.com, anand.jain@oracle.com,
        linux-btrfs@vger.kernel.org, linux-kernel@vger.kernel.org,
        skhan@linuxfoundation.org, gregkh@linuxfoundation.org,
        linux-kernel-mentees@lists.linuxfoundation.org,
        syzbot+a70e2ad0879f160b9217@syzkaller.appspotmail.com
References: <20210727071303.113876-1-desmondcheongzx@gmail.com>
 <20210812103851.GC5047@twin.jikos.cz>
 <3c48eec9-590c-4974-4026-f74cafa5ac48@gmail.com>
 <20210812155032.GL5047@twin.jikos.cz>
 <1e0aafb2-9e55-5f64-d347-1765de0560c5@gmail.com>
 <20210813085137.GQ5047@twin.jikos.cz>
 <a5690ae1-28ba-a933-6473-e9c1e5480f0c@gmail.com>
 <20210813103032.GR5047@twin.jikos.cz>
 <89172356-335f-1ca3-d3a2-78fac7ef93fb@gmail.com>
 <20210819173403.GI5047@twin.jikos.cz>
From:   Desmond Cheong Zhi Xi <desmondcheongzx@gmail.com>
Message-ID: <e9c5bb00-b609-aff9-fc95-ca1c5b9c2899@gmail.com>
Date:   Fri, 20 Aug 2021 11:09:05 +0800
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.11.0
MIME-Version: 1.0
In-Reply-To: <20210819173403.GI5047@twin.jikos.cz>
Content-Type: text/plain; charset=windows-1252; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 20/8/21 1:34 am, David Sterba wrote:
> On Fri, Aug 20, 2021 at 01:11:58AM +0800, Desmond Cheong Zhi Xi wrote:
>>>>> The option #2 does not sound safe because the TGT bit is checked in
>>>>> several places where device list is queried for various reasons, even
>>>>> without a mounted filesystem.
>>>>>
>>>>> Removing the assertion makes more sense but I'm still not convinced that
>>>>> the this is expected/allowed state of a closed device.
>>>>>
>>>>
>>>> Would it be better if we cleared the REPLACE_TGT bit only when closing
>>>> the device where device->devid == BTRFS_DEV_REPLACE_DEVID?
>>>>
>>>> The first conditional in btrfs_close_one_device assumes that we can come
>>>> across such a device. If we come across it, we should properly reset it.
>>>>
>>>> If other devices has this bit set, the ASSERT will still catch it and
>>>> let us know something is wrong.
>>>
>>> That sounds great.
>>>
>>>> diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
>>>> index 70f94b75f25a..a5afebb78ecf 100644
>>>> --- a/fs/btrfs/volumes.c
>>>> +++ b/fs/btrfs/volumes.c
>>>> @@ -1130,6 +1130,9 @@ static void btrfs_close_one_device(struct btrfs_device *device)
>>>>                    fs_devices->rw_devices--;
>>>>            }
>>>>     
>>>> +       if (device->devid == BTRFS_DEV_REPLACE_DEVID)
>>>> +               clear_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state);
>>>> +
>>>>            if (test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state))
>>>>                    fs_devices->missing_devices--;
>>>
>>> I'll do a few test rounds, thanks.
>>
>> Just following up. Did that resolve the issue or is further
>> investigation needed?
> 
> The fix seems to work, I haven't seen the assertion fail anymore,
> incidentally the crash also stopped to show up on an unpatched branch.
> 

Sounds good, thanks for the update. If there's anything else I can help 
with, please let me know.