[dm-crypt] Data Corruption - hard drive or DM-crypt?

[dm-crypt] Data Corruption - hard drive or DM-crypt?

[Matt Rosales]

[-- Attachment #1: Type: text/plain, Size: 974 bytes --]

Hi All,
I have been experiencing data corruption on my DM-Crypt / LUKS
system running Jaunty. Most noticeably it happens with my media (music
dominantly), in the form of read-write errors when copying or moving files,
and as corrupted songs that worked fine before. It is
entirely plausible that other parts of my system are corrupt as well, but
the media is the most noticeable case. This happened before when I was
running Hardy as well, and so when I upgraded to Jaunty I did a fresh
install, wiping the partitions before hand. The data corruption may be due
to a periodic hanging of the system at shut down, which I have not
discovered the cause of, but which apparently prevents the crypted disks
from unmounting, as I get errors upon the following boot- however fsck
doesn't return anything of note! So my question is: How do I determine if I
have a bad hard disk, bad memory, or merely an improper shutdown problem?

Thanks for the advice,

Matt
GPG Key ID: 113828CC

[-- Attachment #2: Type: text/html, Size: 1064 bytes --]

Re: [dm-crypt] Data Corruption - hard drive or DM-crypt?

[Milan Broz]

Matt Rosales wrote:
> I have been experiencing data corruption on my DM-Crypt / LUKS
> system running Jaunty. Most noticeably it happens with my media (music
> dominantly), in the form of read-write errors when copying or moving
> files, and as corrupted songs that worked fine before. It is

If it is really IO errors, the problem is below dm-crypt, your
disk or other hw is probably faulty. Dm-crypt returns IO error only if it
receives failed operation from underlying device.

Also, if your disk or memory is flaky and randomly flips some bits (without IO error),
this cause at least (because of encryption) lost of encryption block
or the whole sector. (Without encryption only flipped bit is wrong and this
can be unnoticed).

Run memory test (memetest+), run some disk-level test (badblocks with write
test - see man badblocks), check syslog for previous errors etc.

Milan

Re: [dm-crypt] Data Corruption - hard drive or DM-crypt?

[Arno Wagner]

On Sun, Sep 13, 2009 at 02:13:45PM +0000, Matt Rosales wrote:
> Hi All,
> I have been experiencing data corruption on my DM-Crypt / LUKS
> system running Jaunty. Most noticeably it happens with my media (music
> dominantly), in the form of read-write errors when copying or moving files,
> and as corrupted songs that worked fine before. It is
> entirely plausible that other parts of my system are corrupt as well, but
> the media is the most noticeable case. This happened before when I was
> running Hardy as well, and so when I upgraded to Jaunty I did a fresh
> install, wiping the partitions before hand. The data corruption may be due
> to a periodic hanging of the system at shut down, which I have not
> discovered the cause of, but which apparently prevents the crypted disks
> from unmounting, as I get errors upon the following boot- however fsck
> doesn't return anything of note! So my question is: How do I determine if I
> have a bad hard disk, bad memory, or merely an improper shutdown problem?

The first step is to determine the precise nature of the corruption.
Is it flipped bits (->RAM), defects in sector size (512B) or blocks
(4kB) (-> disk) or files cut that you just before copied there 
(->shutdown)?

For RAM you can also run memtest86+ for a day or so.
For the disks you can run s long smart selftest and look at
the attributes, both with the smartmontools.

Arno

-- 
Arno Wagner, Dr. sc. techn., Dipl. Inform., CISSP -- Email: arno@wagner.name 
GnuPG:  ID: 1E25338F  FP: 0C30 5782 9D93 F785 E79C  0296 797F 6B50 1E25 338F
----
Cuddly UI's are the manifestation of wishful thinking. -- Dylan Evans

If it's in the news, don't worry about it.  The very definition of 
"news" is "something that hardly ever happens." -- Bruce Schneier