[gentoo-dev] dd: reading `/dev/hda7': Input/output error (not a hardware problem) - kernel problem?

[Bastian Fuchs]

Hi!
I have some problem to read partitions with dd, but I think it's not a 
hardware problem. (No badblocks or something else.)

I have two disks hda and hdb. Its possible to read hda1 and hdb1 with dd 
without an error. But on all other partitions I get an "Input/output" error.

---------------------------------------------------------------------------------
vulpec ~ # fdisk -s /dev/hda1
88326
vulpec ~ # dd if=/dev/hda1 of=/dev/null bs=1k skip=88000
326+0 records in
326+0 records out

hda1 has 88326 blocks and dd reads 88326. OK.
---------------------------------------------------------------------------------

---------------------------------------------------------------------------------
vulpec ~ # fdisk -s /dev/hda7
4200966
vulpec ~ # dd if=/dev/hda7 of=/dev/null skip=4200000 bs=1k
dd: reading `/dev/hda7': Input/output error
964+0 records in
964+0 records out
vulpec ~ # dmesg | tail
...
attempt to access beyond end of device
03:07: rw=0, want=4200968, limit=4200966

hda7 has 4002966 blocks, but only 4002964 blocks a read. Why to blocks are 
missing? And dmesg says, that dd wants to access 2 blocks beyond the end of 
device?
---------------------------------------------------------------------------------

Again:
This problem only occures with all partitions (ext2, ext3, ReiserFS) except 
the first. The partition table is ok and there are no bad blocks. I find some 
postings with google, which reports the same problem, but mostly with cdroms, 
but harddiscs too. Some people think, this is an kernel problem, but the 
mails/postings are from 2000 oder 2001.

It seems that nobody really knows, whats wrong. I hope you can help me. Maybe 
it's an bug ...
- Bastian Fuchs

PS:
vulpec ~ # uname -a
Linux vulpec.wg.j.th.de 2.4.19 #1 Mon Oct 28 02:30:04 CET 2002 i686 AMD 
Duron(tm) processor AuthenticAMD GNU/Linux

vulpec ~ # dd --version
dd (fileutils) 4.1.11