[LINUX:34]

[FOREJIN@example.com]

Hello,

Recently I've been having some very unusual problems on my Linux
system.  The problem is when doing some standard disk i/o things (cp, cat,
tar, etc) I get the message:

command:  filename: I/O error

For example if I was cat'ing a file foo it would say:

cat:  foo: I/O error

Unfortunately, it does not give any more information than that.  Like
what type or I/O error or anything like that.  Also it has some other
strange characteristics.  The error seems to be physical, and it seems to
be limited to one directory, though it may crop up in other directories as
well.  The can best be exaplained with the following examples (which are
what really happened).

#cp /dosd/incoming/linux/foo .
#cat foo > /dev/null
#cp /dosd/incoming/linux/foo bar
#cat bar > /dev/null
#sync
#cat bar > /dev/null
cat:  bar: I/O error
#cat foo > /dev/null
#

So, as you can see it seems to be a physical thing on the disk as it
appeared after the sync.  But the strange thing is shown below.  This
follows exactly after the above:

#rm foo
#rm bar
#cp /dosd/incoming/linux/foo ..
#sync
#cat ../foo > /dev/null
#cp /dosd/incoming/linux/foo ../bar
#sync
#cat ../bar > /dev/null
#

So, it appeared it was somehow a problem in the current directory only,
and not its parent.  I then thought that if it was a physical error the
second set of events might have not used the same sectors as the first, so
I repeated the second set copying to "foo1", "foo2", . . ., "foon" until
the filesystem was full.  I then cat'd them all to /dev/null and not a
single one gave me the error.  Mind you all of this occured in the parent
directory, not the problem directory.  Just for my sanity I removed all of
those foo's, and repeated the experiment in the problem directory and I got
the error again.

My last set of tests was to create another directoy in the parent
(good) directory, and move everything from the bad one to there.  At first
that seemed fine, as I didn't get the error for a while, but eventually I
did.  So, following the specs on my system, are my questions:

My Linux System:

Kernel 1.1.45 (no additional patches)
(this error also occurred under 1.1.0 and 1.1.9)
IBM ThinkPad 750
340 MB HD (my Linux partition is 150MB, swap is 20M, and DOS is the rest)

My questions:

1.  Does anyone have any idea why this is occurring?
2.  How can I fix it?
3.  Why is it's behavior so erratic and strange?
4.  Is there a Linux command to do a low level check of the hard drive
for errors, without harming the data already on it (I've fsck with no
results)?
5.  Does anyone have any other suggestion, comments, or questions?


Thank you,

Rainer