On 2018年04月08日 07:29, Ben Parsons wrote:
> On 7 April 2018 at 22:09, Qu Wenruo <quwenruo.bt...@gmx.com> wrote:
>> On 2018年04月07日 10:31, Ben Parsons wrote:
>>>> Pretty common hard power reset.
>>>>> looking at journalctl, there is a large stacktrace from kernel: amdgpu
>>>>> (see attached).
>>>>> then when I booted back up the pool (2 disks, 1TB + 2TB) wouldn't mount.
>>>> I'd say such corruption is pretty serious.
>>>> And what's the profile of the btrfs? If metadata is raid1, we could at
>>>> least try to recovery the superblock from the remaining disk.
>>> I am not sure what the metadata was but the two disks had no parity
>>> and just appeared as a single disk with total space of the two disks
>> Strangely, for the 2nd disk, it's sdc1, which means it has partition table.
>> While for the 1st disk, it's sda, without partition table at all.
>> Is there any possibility that you just took run partition?
>> (Or did some program uses it incorrectly?)
> I dont quite understand what you are asking.
> I was always under the impression I could run mount on either
> partition and it would mount the pool
>>> how would i got about recovering the 2nd disk? attached is
>> The 2nd disk looks good, however it's csum_type is wrong.
>> 41700 looks like garbage.
>> Despite that, incompact_flags also has garbage.
>> The good news is, the system (and metadata) profile is RAID1, so it's
>> highly possible for us to salvage (to be more accurate, rebuild) the
>> superblock for the 1st device.
>> Please dump the superblock of the 2nd device (sdc1) by the following
>> # dd if=/dev/sdc1 of=super_dump.sdc1 bs=1 count=4096 skip=64k
> See attached.
>> Unfortunately, btrfs-sb-mod tool added recently doesn't have all needed
>> fields, so I'm afraid I need to manually modify it.
>> And just in case, please paste the following output to help us verify if
>> it's really sda without offset:
>> # lsblk /dev/sda
>> # grep -obUaP "\x5F\x42\x48\x52\x66\x53\x5F\x4D"
> dd if=/dev/sdb of=toGrep.sdb bs=1 count=128M status=progress
> cat toGrep.sdb | grep -obUaP "\x5F\x42\x48\x52\x66\x53\x5F\x4D"
Well, the magic number is completely correct, and at correct location.
Would you please run "btrfs inspect dump-super -fFa /dev/sdb" again?
This time it should provide good data.
>> Above grep could be very slow since it will try to iterate the whole
>> disk. It's recommended to dump the first 128M of the disk and then grep
>> on that 128M image.
>> BTW, with superblock of sdc1 patched, you should be able to mount the fs
>> with -o ro,degraded, and salvage some data.
> Thank you so much!
> I am better off copying the data to another disk and then rebuilding the pool?
> or can I just run a scrub after the super block is fixed?
According to your latest grep output, strangely the 1st device is not
that corrupted as before.
So I think in-place recover should save you a lot of time.
> For reference here is lsblk:
> sda 8:0 0 465.8G 0 disk
> ├─sda1 8:1 0 512M 0 part /boot
> ├─sda2 8:2 0 455.3G 0 part /
> └─sda3 8:3 0 10G 0 part [SWAP]
> sdb 8:16 0 931.5G 0 disk
> -- first disk
> sdc 8:32 0 1.8T 0 disk
> └─sdc1 8:33 0 1.8T 0 part
> -- 2nd disk
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html