Provided by: scsitools_0.8-2_i386
scsi-config - query information from a scsi device with a nice user
scsi-config queries information from an scsi target with a nice Tcl/Tk
user interface. If you do not specify a device to query, scsi-config
calculates a list of available devices and prompts it to you.
By the nature of a graphical user interface, most things will explain
them self. Basically, scsi-config shows a list of buttons for certain
mode pages which you may press. Those buttons which you can press and
the text windows with white backgrounds can be modified by you and the
modifications send back to the device.
In the main window there is a button to instruct the device to save the
data in some non volatile memory (if it supports it). Note that this
will instruct the device to save the Read-Write Error Recovery Page ,
Disconnect-Reconnect Page , Format Device Page , Caching Page , and
Control Mode Page in its NVRAM. Usually saving even a single of those
should write them all to the NVRAM, but you never know.
You can query the current, the factory default and the values in the
NVRAM (which may not be the current parameters) from the device.
Not all combinations of button toggles or all values are valid. In
general, try to set them and see which values the drive accepts. scsi-
config rereads the device configuration immediately, s.t. you see which
values where accepted.
Also note that some disk drives are notched, and that those have an
active notch (shown in the main window) to which all your settings
apply (at least those of notched pages, which are also marked in slate
gray). You can select the active notch to which your settings apply
(and to which the values refer) in the Notch Page.
For those devices which do not feature an NVRAM (generally removable
media devices) and as a kind of backup, you can save the current
settings to a file. Actually the file will be a /bin/sh script making
the necessary scsiinfo(8) calls to set the saved parameters when
There is also a nice Overview button which will query many details
about the disk geometry and draw them in a nice picture. This looks
esp. nice for drives with many notches, that is different regions on
the disk with different tracks per sector settings. It is also very
useful for notched drives as you can immediately select the mode pages
for each notch.
SOME USAGE GUIDELINES
1. General Warning
Generally, do not modify settings you don’t understand. It is
useful to know the SCSI-II specs mentioned below. Some setting
may render the device unusable or even damage it. Usually a
power cycle resets the state (if you do not save the weird
settings in the NVRAM). Some settings affecting the assignment
of logical sectors will render the disk unusable until the next
low level format.
2. On Write Caching
As an old warning, this does also mean you should not generally
switch the write cache on. At least on those drives where you
have a choice at all. Reasons are twofold:
a) It is a priori unclear when the drive will actually
perform the writes. This is a bad thing when considering
shutdown of your machine. On the other hand, it seems
sensible to assume that the drive will immediately write
it’s cache to disk when it is idle (after all file
systems are unmounted) and due to the size of the on disk
cache this will usually only need a few seconds after the
shutdown (but the drive lamp will usually not glow, as it
is mostly connected to the host adapter (if you have a
lamp at all) and it is not participating).
There is a SCSI command to flush the caches. Linux could
call it prior to shutting down, spinning a disk down. Due
to my knowledge this is not yet done though.
b) As the writes are performed asynchronously, errors are
reported asynchronously. The disk might return an error
at some simple read instruction related to a write which
was acknowledged OK several transactions ago. This
generally confuses things and makes interpreting errors
very difficult. Some devices are known/said to not report
write errors in this mode of operation at all.
Just imagine that at the point where a file system is
unmounted, or a new removable media is detected it could
tell: Oops, BTW, there was some write error ago although
I told you it was OK already.
Thus, when you run a disk in write cache mode, keep it in mind
when weird error messages occur and give the disk time to flush
it’s buffers at shutdown. Generally it would be good if you knew
more vendor specific details on how the disk operates in the
write cache mode.
3. Reassigning Bad Blocks Automatically
One of the nice features of SCSI disk is that they allow to
remap bad blocks automatically as they are detected without any
user intervention. However, you actually have to enable this
feature! It turned out that you can not generally assume a disk
in this mode. To enable this mode or check the settings, proceed
a) Go to the Read-Write Error Recovery Page. AWRE
(Automatic Write Reallocation Enable) and ARRE (Automatic
Read Reallocation Enable) buttons enable the automatic
In the same window, you can select the maximal retries
performed. EER allows the disk to do some Early Error
Recovery which is fast (but might misdetect or miscorrect
data). A selected DCR button (Disable CoRrection Codes)
disallows the disk to use any error correction codes at
all (thus the drive will have to retry until it performs
an error free read).
Usually a sector will be reallocated after even a single
read retry or the given number of failed write retries.
When the sector cannot be recovered, it is reallocated
but the data is lost and an error is signalled.
The other buttons there apply to error reporting as well.
TB Transmits the bad Block together with the error, RC
Reads Continuous, that is, does not pause a read
operation while retrying or using error code calculations
to recover a bad block (thus may return bad data). PER
lets the disk report even recovered errors (Post ERror),
DTE (Disable Transfer on Error) even breaks a running
data transmission when an error is detected.
b) Even when the reallocation is enabled, the disk must
actually have some reserved areas where to remap the bad
blocks. The Format Page controls this. Either a given
number of Alternate Sectors Per LUN is set aside for the
whole disk or a given number of tracks is defined to be a
zone and for each zone a number of sectors or tracks is
put aside. These alternate data areas are where bad
sectors are remapped.
Note that this page is very likely to apply only to the
current notch on a notched disk device.
If there are no, or not many alternate sectors reserved
on your disk, you must change these settings.
I found that those disks which allow to modify these
settings are very often set to no reserved sectors at all
by the vendors, as this increases the disks capacity. For
the sake of stability, you should really modify these
If you decide to modify the number of alternate sectors,
i) Save the parameters to the NVRAM of the disk.
ii) Low-level format the disk drive (and not erase
the NVRAM during this operation).
to make the changes be effective.
c) Esp. when you set the disk to not report recovered errors
(or when it is in a write cache mode or something), and
just as a general guideline, keep an eye on the grown
defects list where the disk will report all the
remappings which took place as your disk ages.
The old adventurers guide line applies: Save (Backup)
Early, Save (Backup) Often. Find the right time to
replace your disk avoiding too much work recovering your
scsi-config could be enhanced by making better use of Tcl/Tk. I learned
much about it during my work on tk_scsiformat(8) and their would be
much room for fixes and enhances. On other hand, it fulfils it’s
purpose as it is quite nicely already.
As scsi-config makes heavy use of scsiinfo(8) all it’s bugs (esp. on
defect reading) apply.
/usr/lib/scsi/cache, /usr/lib/scsi/control, /usr/lib/scsi/disconnect,
/usr/lib/scsi/error, /usr/lib/scsi/format, /usr/lib/scsi/generic,
/usr/lib/scsi/inquiry, /usr/lib/scsi/notch, /usr/lib/scsi/overview,
/usr/lib/scsi/peripheral, /usr/lib/scsi/rigid, /usr/lib/scsi/save-
changes, /usr/lib/scsi/save-file, /usr/lib/scsi/tworands,
/usr/lib/scsi/verify (Tcl/Tk subroutines used by scsi-config).
scsiinfo(8), scsiformat(8), tk_scsiformat(8), fdisk(8), sd(4),
American National Standard
for information systems
SMALL COMPUTER SYSTEM INTERFACE - 2
MARCH 9, 1990
Michael Weller <email@example.com>, Versions 1.5 & 1.7