1 |
On 9/8/07, Volker Armin Hemmann <volker.armin.hemmann@××××××××××××.de> |
2 |
wrote: |
3 |
> |
4 |
> On Samstag, 8. September 2007, Jorge Peixoto de Morais Neto wrote: |
5 |
> > http://forums.gentoo.org/viewtopic-p-4225196.html#4225196 |
6 |
> > |
7 |
> > Hi. The forums being down, can you give me help by mail on the topic |
8 |
> > http://forums.gentoo.org/viewtopic-p-4225196.html#4225196, since I can't |
9 |
> > use my Gentoo installation until the problem is solved? |
10 |
> > |
11 |
> > The most important thing to me is to know the answer to the two |
12 |
> questions: |
13 |
> > |
14 |
> > 1) How can I know if other files were corrupted? |
15 |
> |
16 |
> well, you can try a script that compares the md5sums of the files |
17 |
> installed |
18 |
> with the m5sums of the files on the harddisk. But everything that wasn#t |
19 |
> installed with portage can't be checked that way. |
20 |
|
21 |
There is a equery command for that... equery check if memory serves. But I |
22 |
issued this command some days ago and it always reports some files as |
23 |
different... so I guess it is normal for one or two files out of 1000 in a |
24 |
package to be modified without Portage knowing... but I want to know about |
25 |
the packages that were modified *because of the corruption.*, not the ones |
26 |
that were modified because of other reasons... but, perhaps this is the only |
27 |
option: issue equery check and |
28 |
1) if the checksum match I know it was not corrupted |
29 |
2) if it does not match than it may or may not be because of the corruption |
30 |
|
31 |
Too bad it does not apply to files not managed by Portage. |
32 |
Hum, perhaps I should have made checksums of my personal data? Obviously, |
33 |
nothing substitutes a backup, but for data that is not worth backing up |
34 |
(because it is huge - thus costly to back up - and I can withstand a small |
35 |
chance of losing said data, since I can obtain it again; a rip of a DVD |
36 |
movie for example) I could at least save checksums, so if the file gets |
37 |
corrupted, at least I'll know... |
38 |
|
39 |
|
40 |
> 2) Do you think I should just use the computer, after reemerging the |
41 |
> > packages that provide the corrupted files? |
42 |
> |
43 |
> yes |
44 |
|
45 |
Do you think that there is any plausible chance that using the partition |
46 |
might cause further damage? |
47 |
|
48 |
|
49 |
> The background is: a corruption ocurred in my reiserfs partition, possibly |
50 |
> > due to hardware problems; I performed reiserfsck and recovered virtually |
51 |
> > all files, but at least some files in /bin are corrupted. In fact, they |
52 |
> > cannot be executed, and executing the "file" command on them tells that |
53 |
> > they are just "data" instead of recognising them as an executable. See: |
54 |
> > |
55 |
> > |
56 |
> > jorge@jorge:/media/hda2/bin$ for file in *; do file "${file}" | grep -q |
57 |
> > data && wc -c "${file}"; done |
58 |
> > 13772 basename |
59 |
> > 13500 chroot |
60 |
> > 27048 cut |
61 |
> > 79420 dir |
62 |
> > 13836 dirname |
63 |
> > 59084 du |
64 |
> > 13436 env |
65 |
> > 23336 expr |
66 |
> > 24516 head |
67 |
> > 14644 mkfifo |
68 |
> > 18988 readlink |
69 |
> > 34100 rm |
70 |
> > 14684 rmdir |
71 |
> > 17740 seq |
72 |
> > 14772 sleep |
73 |
> > 65168 sort |
74 |
> > 37380 stty |
75 |
> > 12584 sync |
76 |
> > 35852 tail |
77 |
> > 34328 touch |
78 |
> > 27492 tr |
79 |
> > 12200 true |
80 |
> > 12800 tty |
81 |
> > 15316 uname |
82 |
> > 79420 vdir |
83 |
> > 23108 wc |
84 |
> > 12876 yes |
85 |
> > |
86 |
> > Aren't all these files part of coreutils? |
87 |
> |
88 |
> not all, but most of them. |
89 |
> |
90 |
> > |
91 |
> > For more details, see the forum thread. |
92 |
> > I would appreciate any help. Thanks. |
93 |
> |
94 |
> It doesn't look so bad. You can try moving the corrupted stuff to a backup |
95 |
> dir, and create symlinks to busybox. Busybox should be installed on your |
96 |
> system. |
97 |
|
98 |
Yes it is. |
99 |
|
100 |
for example |
101 |
> ln -s busybox rm |
102 |
|
103 |
Gentoo should have an automated way to do this. For me, it looks like there |
104 |
should be an eselect option for "activating" busybox. |
105 |
|
106 |
|
107 |
> I hope you learned your lessons! |
108 |
> |
109 |
> Lesson 1: /home on its own partition. |
110 |
|
111 |
I read somewhere that most of the time when a disk fails it will take all of |
112 |
its partitions with it, so putting /home is its own partition does not help. |
113 |
|
114 |
Perhaps that person was wrong... at least in my case, I clearly had a |
115 |
logical failure in the partition, with no physical failure in the disk, so |
116 |
if I had multiple partitions, maybe only one would have problems. |
117 |
|
118 |
|
119 |
> Lesson 2: backups. |
120 |
|
121 |
Hehe. Yes I know. Fortunately it seems I was very lucky this time... |