[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: File corruption



Firstly I think this is more suitable for debian-user, so I'm BCCing -devel 
and CCing -user.  Please CC me on any follow-ups.

On Wed, 17 Oct 2001 15:27, Steven Mooij wrote:
> Hello, I'm not a debian developer (yet), but I hope I am allowed to post a
> message here. It is about a potential bug, but I can't send it with the
> BTS, because I don't know which package to file it against.

Firstly what you are describing isn't a debian bug so is not suitable for the 
BTS.  It's either a kernel issue or a hardware issue.

> On Wed, 17 Oct 2001 00:58:59 Russell Coker wrote:
> > Also there's this new "tainted kernel" thing which is a huge new feature,
> > and
> > there are rumours about Ext3 FS corruption and data loss..
>
> Maybe my problem has something to do with these rumours, maybe not: When I
> copy large amounts of data sometimes some bytes are changed. This must
> sound vague: sometimes, some bytes, but unfortunately this is the case, I
> haven't figured out a reproducable test-situation.

Firstly create a file with a variety of data that's significantly larger than 
ram.  Eg:
tar cvf /tmp/junk /usr

Then run md5sum repeatedly on it in the following fashion:
while /bin/true ; do md5sum /tmp/junk ; done > /tmp/out

Leave that running overnight and then run the following in the morning:
uniq < /tmp/out

If it gives more than one line of output then your hard drives are not 
repeatedly returning the same data to read requests.

There is a potential problem that this error may only occur on certain parts 
of the hard drive.  Booting with init=/bin/sh and running the following could 
be useful in that case:
while /bin/true ; do md5sum /dev/md1 ; done

Doing that all night shouldn't exhaust the kernel scroll-back buffer.

Then if that doesn't show anything then you have to try repeatedly copying a 
file and using md5sum to check the result (should be the same after every 
copy).

> 2628DB Delft

If your machine is portable I could meet you at a HCC or NLLGG meeting and 
check it out.  Send private mail if that interests you.

-- 
http://www.coker.com.au/bonnie++/     Bonnie++ hard drive benchmark
http://www.coker.com.au/postal/       Postal SMTP/POP benchmark
http://www.coker.com.au/projects.html Projects I am working on
http://www.coker.com.au/~russell/     My home page



Reply to: