| 1 |
On 17/11/19 20:35, n952162 wrote: |
| 2 |
>> |
| 3 |
>> Needless to say, you would not try this on the original partition, but a |
| 4 |
>> backup image you can create with ddrescue and friends. In any case, |
| 5 |
>> running |
| 6 |
>> fsck.ext4 -n (or -E nodiscard) should not cause any fs losses, unless the |
| 7 |
>> disk/hardware is faulty. Hence working on a backup image is the safest |
| 8 |
>> option. |
| 9 |
>> |
| 10 |
> |
| 11 |
> Thanks for the tip about ddrescue. |
| 12 |
|
| 13 |
I'll throw in another tip, about dm-verity. If you haven't met it, it's |
| 14 |
part of the block-management layer. It's new so I don't know much about |
| 15 |
it but I'm personally excited about its capabilities for raid recovery |
| 16 |
and stuff like that. |
| 17 |
|
| 18 |
Basically, on top of your block device you would create a dm-verity |
| 19 |
device. I don't know whether you can, but if you can't it needs adding - |
| 20 |
you want to create a device in corrupted mode where any read access will |
| 21 |
trigger a read error. |
| 22 |
|
| 23 |
As ddrescue copies your old device across it will reset the integrity |
| 24 |
layer. So when you go to read your device image, any attempt to read a |
| 25 |
successfully recovered block will be fine, any attempt to read a block |
| 26 |
that couldn't be recovered will trigger a read error. |
| 27 |
|
| 28 |
Cheers, |
| 29 |
Wol |
Archives