Bug 5928

Most recent kernel where this bug did not occur:
Distribution:LFS
Hardware Environment:Exabyte IDE tape drive
Software Environment:
Problem Description:
I'm afraid I found another bug, but with ide-tape this time.

I try to give you a description :

I have 3 tar files :
tapetar.tar :           10k
midletapetar.tar :      30k
bigtapetar.tar :        950k

With st sg and ide-scsi, I wrote twice the 3 files consecutively :
root@mybox $ dd if=tapetar.tar if=/dev/nst0 bs=1k
root@mybox $ dd if=midletapetar.tar if=/dev/nst0 bs=1k
root@mybox $ dd if=bigtapetar.tar if=/dev/nst0 bs=1k
*2

As I mentionned above, I succeeded to read them.
But then I unloaded st and loaded ide-tape instead :
root@mybox $ rmmod st sg ide-scsi
root@mybox $ modprobe ide-tape

And then I performed a read :
root@mybox $ mt -f /dev/nht0 tell
At block 0.
root@mybox $ dd if=/dev/nht0 of=outtar.tar bs=1k
1980+0 records in
1980+0 records out

Well it did read all the files I wrote on the tape putting them all in the same output tar file.
I don't know if it's a normal behaviour, however then I tried to write on the tape then to read :


root@mybox $ mt -f /dev/nht0 tell
At block 0.
root@mybox $ dd if=bigtar.tar of=/dev/ht0 bs=1k
950+0 records in
950+0 records out
root@mybox $ mt -f /dev/nht0 tell
At block 0.
root@mybox $ mt -f /dev/nht0 asf 1
root@mybox $ mt -f /dev/nht0 tell
At block 11.

It seems that the write did silently fail, or did not change the file marks, well it looks wrong.



Steps to reproduce: