Well if the only thing that is different is the OS and the localization of the file, and you already tried different Linux versions to rule out the OS (and you stated that you have problems on Windows also, even if they are from a different kind), then it has to be the localization of the file (reminds me of a famous sentence from Sherlock Holmes). May be your disk has bad sectors on the ext4 partition (and may be in unused aprts of other partitions), but not in the used parts of your NTFS partition.
A disk check will make this clear. If it is not the disk I would assume that it is the motherboard or the CPU, there may be different symptoms of the same issue, because they are different OSes and act differently.

Well, I probably did not described it clearly.
I did not run check-sum test from OS windows.
I am checking ntfs files under linux OSes (installed NTFS-3G 2011.4.12) (6 different ubuntu instances on 4 different hard disks).
This weird behavior is not only on one hard drive but on FOUR DIFFERENT hard disks (and 5 different ubuntu linux instances).
Two of hard drives are connected through SATA III/II.
One hard disk is connected through USB 3.0 port but disk supports only usb 2.0.
One ext4 partition is on USB flash drive (motherboard USB 3.0 and flash-drive USB 2.0).

There is very low probability that all four drives have the same problem - especially if they are using different hardware connectivity (usb or sata) and have different hardware equipment (3 are "classic" magnetic hard disks (not SSD) , 1 is flash-memory usb stick).

There is as well low probability that all of four ext4 partitions located on 4 different hard disks are wrong the same way.

And all ntfs (8 partitions/6 hard disks) are correct. (Connected by sata or usb - internal or external HDDs.)

Thats the reason why I do not understand. Because if it would have been problem only on one disk - thats I would understand - it could be hardware problem.

But all of 4 disks have the same ext4 issue but not ntfs. So, I cannot assume that only ext4 partitions are wrong because of hardware failure and ntfs are correct. Or assume that I have ext4 partitions on wrong places and ntfs are on right places :-) That is really sounds to me as extremely low probability :-)

And why is this happening only on files bigger than 3.9 Gigabytes ?

I really made all those tests with so many combinations.
ntfs partitions have been tested with "chkdsk /r" parameter many times within last months. - no errors.
ext4 partitions have been tested without -c parameter ,not mounted, many times like "sudo fsck -t ext4 -v -y -f /dev/????" - no errors.

solved
 

Hi,

I would like to thank all of you , who helped me to solved this problem.
I have found out that one out of four memory modules is wrong.

cheers,
Martin

Hi,

I have the same problem.
My linux system is a ubuntu 11.10 (Kernel 3.0) and my hw specs are:
AsusP8P67-deluxe, 16GB RAM (2133MHz), i7-2600K runnung from 4GHz upto 5.1GHz.

My problem is to read a text file of size 3GB.
A simple c program reads blockwise, sometimes i get errors.
Also md5sum delivers a different of the 3GB file value after i read with my c program.
I get this error on a raptor hdd 74GB (ext4), as well on a SSD disk(ext4) (corsair force gt 120GB).
I also get this error on my 500GB samsumg data disk (ext3) if i run the same program.
On the 500GB samsung disk I get one "Input/output error" when running md5sum.

Hmm, .. very strange.
A kernel bug??
The next thing what I try is to format the hdd with some other fs.

Best,
Michael

I would rather assume that your issue is an unstable overclock. try if you get the same errors when running the system with stock settings.

Maybe, but today I tried to connect my 2 hdds on the Marvell 6GBit SATA boards (instead of the Intel P67 SATA port) and based on my first tests it works!
Let's see if this solves the problem also in the future.
Currently I run on 4.8GHz.

Michael

This lets me assume that you didn't test your overclock for stability. Have you at least an eye on the temps or do you fry your system slowly?

I tested it for stability, it is linpack stable over a long time (with intel 12.0 compiler suite..). Also memtest has no errors in an overnight run.
Temps are o.k. - I have an excellent watercooler.
If I have time, I will try to reconnect the hdd's on the P67's SATA boards and try to repoduce the file read error with default bios settings.

Best,
Michael