1 |
On Thu, Dec 18, 2014 at 3:24 PM, <meino.cramer@×××.de> wrote: |
2 |
> Hi, |
3 |
> |
4 |
> with |
5 |
> |
6 |
> sysctl vm.block_dump=1 |
7 |
> |
8 |
> one can enable the logging of IO to the harddisk/flashmem/... |
9 |
> into dmesg. |
10 |
> The logs report the block number of the file in question... |
11 |
> but not the filename itsself. |
12 |
> |
13 |
> Is there any other way as examine each single file of the |
14 |
> filesystem to find the file to which a certain block number |
15 |
> is assigned? |
16 |
> |
17 |
> Thank you very mcuh for any help! |
18 |
> Best regards, |
19 |
> Meino |
20 |
> |
21 |
> |
22 |
> |
23 |
|
24 |
That depends entirely on the filesystem being used. In the case of |
25 |
ext2/3/4, I believe /sbin/debugfs will do the trick with its icheck |
26 |
command to get the inode, and once you have the inode, you can get the |
27 |
filename via find. What I'm not 100% certain of is whether the block |
28 |
numbers involved map 1:1 with physical sectors, and how that plays |
29 |
with the 512B vs 4KB sectors, etc. With NTFS it's a hair quicker with |
30 |
ntfscluster and ntfsinfo doing the trick fairly trivially (I use a |
31 |
tool centered around that combo to identify files lost when I recover |
32 |
peoples windows drives with ddrescue). |
33 |
|
34 |
-- |
35 |
Poison [BLX] |
36 |
Joshua M. Murphy |