1 |
Alan McKinnon píše v Út 23. 01. 2007 v 21:55 +0100: |
2 |
> On Tuesday 23 January 2007 19:47, Hans-Werner Hilse wrote: |
3 |
> |
4 |
> > Did you reboot between changing the partition layout and creating |
5 |
> > that new partition (and moving data)? Otherwise the kernel wouldn't |
6 |
> > be aware of the new partition layout. Well, if everything you wrote |
7 |
> > is correct, that data should have ended up on that former Windows |
8 |
> > partition and that partition should now be an ext3 one. But if you |
9 |
> > just didn't care and mounted the old linux partition (sdb2 at that |
10 |
> > point in time before the new partition layout), copied data and you |
11 |
> > _then_ rebooted -- then you would have written your data to a |
12 |
> > partition that was only a reminiscence in the kernel's structures and |
13 |
> > not |
14 |
> > corresponding to what cfdisk wrote to the HD. That would be an |
15 |
> > explanation why the next boot failed. |
16 |
> > |
17 |
> > > When I do "mount /dev/sdb1 /mnt/zaloha" at /mnt/zaloha I see that |
18 |
> > > old Windows NTFS partition that I already deleted (There are |
19 |
> > > "Program Files", "WINDOWS", ...). I don't understand why (somewhere |
20 |
> > > I read that ext3 start writing at the middle of the disk space to |
21 |
> > > prevent defragmentation). |
22 |
> > |
23 |
> > Deleting the partition is something that only affects the boot |
24 |
> > sector. Ext3 should in fact have overwritten this with it's first |
25 |
> > superblock. So the mkext2fs you issued did definitively hit the wrong |
26 |
> > partition. |
27 |
> > |
28 |
> > So my suggestion is: try "gpart -w ext2,1.5 /dev/sdb" to find your |
29 |
> > partition (even better: write back the backup you've made from the |
30 |
> > old partition table. Errrm...) |
31 |
> |
32 |
> Some background here to elaborate on what Hans has said: |
33 |
> |
34 |
> It looks like when you moved the data onto the new partition, it got |
35 |
> written somewhere on the disk. However, the kernel's idea of how the |
36 |
> partitions are laid out at that time and what fdisk just wrote to the |
37 |
> disk probably don't agree and the kernel had got it wrong.... This does |
38 |
> happen when you delete two or more partitions and create one large one. |
39 |
> |
40 |
> That's the bad news. The good news is that unless you did something to |
41 |
> wipe the disk clean, the data is there somewhere and you need to find |
42 |
> it. Hans' gpart command will search the disk looking for the sequence |
43 |
> of data that is found at the start of a filesystem, and will then make |
44 |
> a smart estimate as to what the partition ought to look like. |
45 |
> |
46 |
> The next good news is that you can create and delete partitions many |
47 |
> times and still get the data back intact as long as you don't overwrite |
48 |
> it. fdisk updates the partition table right at the start of the disk |
49 |
> and does nothing else so you can always undo these changes. Until you |
50 |
> are happy that everything is back it will be smart to mount this |
51 |
> partition read-only so it can't be changed: |
52 |
> |
53 |
> mount -o rw /dev/sdb1 /path/to/mount/point |
54 |
> |
55 |
> You say in your original mail that after moving the data "everything was |
56 |
> fine". What exactly do you mean by that: |
57 |
> |
58 |
> 1. The command ended without failure so you assume it moved stuff |
59 |
> correctly, or |
60 |
> 2. You proved the move was done by mounting the partition and all your |
61 |
> files were there, or |
62 |
> 3. Some other reason? |
63 |
> |
64 |
> alan |
65 |
> |
66 |
|
67 |
|
68 |
|
69 |
-- |
70 |
gentoo-user@g.o mailing list |