This is (C) Marc SCHAEFER 1998. Some work has been done within DILOG SA
You can contact me at schaefer@dilog.ch. This is released under the GPL.
Any warranty is expressly denied. Backup your data. Do not use it
on disks with data on them (part of the benchmark is writing ...).

This only works on the first SCSI device of the system, because this
has been coded very badly. I think that the direct I/O approach taken
by Steven TWEEDIE is probably a lot better, and probably gives similar
result as this benchmark.

NB: filesystems benchmark are really better. However, sometimes you want to
    show customers how a disk drive behaves depending on the LBA address.
    That was the goal of the software. In the scripts/ directory, you will
    find a script converting from the data that linear generates into a
    postscript graph showing speed over the whole disk for both read and
    write. Well, the ideal benchmark *is your specific application*.
    No benchmark can replace your application's behaviour and needs.

DILOG SA is a RAID manufacturer. More information at http://www.dilog.ch
You can see a few disk versus DILOG products benchmarks in the results/
directory. AGAIN, what interest us in this benchmark is the speed
depending on the LBA, not the real speed.

Installation instructions:

- type make
- compile your kernel with support for user misc modules (CONFIG_UMISC).
  This cannot be compiled as a module. This software can, but needs
  UMISC compiled-in first. You also need to run /sbin/lilo and reboot.
  If you use standard kernels from distributions such as SuSE or 
  RedHat, you probably have UMISC built-in already.
- then
     /sbin/insmod benchio.o
     mknod /dev/benchio c 10 147
- launch the benchmark (linear NUMBER_OF_LBA). You can get the number of
  LBA from dmesg when you insert sd_mod.
