sane-scsi - SCSI adapter tips for scanners
This manual page contains various operating-system specific tips and tricks on how to get
scanners with a SCSI interface working.
For scanners with a SCSI interface, it may be necessary to edit the appropriate backend
configuration file before using SANE for the first time. For most systems, the configura-
tion file should list the name of the generic SCSI device that the scanner is connected to
(e.g., under Linux, /dev/sg4 or /dev/sge is such a generic SCSI device). It is customary
to create a symlink from /dev/scanner to the generic SCSI device that the scanner is con-
nected to. In this case, the configuration file simply lists the line /dev/scanner. For
a detailed description of each backend's configuration file, please refer to the relevant
backend manual page (e.g., sane-epson(5) for Epson scanners, sane-hp(5) for HP scanners,
For Linux, there is an alternate way of specifying scanner devices. This alternate way is
based on /proc/scsi/scsi and allows to identify scanners by the SCSI vendor and model
string and/or by the SCSI device address (consisting of bus number, channel number, id,
and logical unit number). The syntax for specifying a scanner in this way is:
scsi VENDOR MODEL TYPE BUS CHANNEL ID LUN
where VENDOR is the SCSI vendor string, MODEL is the SCSI model string, TYPE is type SCSI
device type string, BUS is the SCSI bus number (named "host" in /proc/scsi/scsi), CHANNEL
is the SCSI channel number, ID is the SCSI id, and LUN is the logical unit number of the
scanner device. The first two fields are strings which must be enclosed in double-quotes
if they contain any whitespace. The remaining four fields are non-negative integer num-
bers. The correct values for these fields can be found by looking at the output of the
command "cat /proc/scsi/scsi". To simplify configuration, a field's value can be replaced
with an asterisk symbol (``*''). An asterisk has the effect that any value is allowed for
that particular field. This can have the effect that a single scsi-line matches multiple
devices. When this happens, each matching device will be probed by the backend one by one
and registered if the backend thinks it is a compatible device. For example, the line
scsi MUSTEK MFS-06000CX Scanner 0 00 03 00
would attach the Mustek SCSI scanner with the following /proc/scsi/scsi entry:
Host: scsi0 Channel: 00 Id: 03 Lun: 00
Vendor: MUSTEK Model: MFS-06000CX Rev: 4.04
Type: Scanner ANSI SCSI revision: 0
Usually it's sufficient to use vendor and model strings only or even only the vendor
string. The following example
scsi MUSTEK * * * * * *
would have the effect that all SCSI devices in the system with a vendor string of MUSTEK
would be probed and recognized by the backend.
If the remainder of a scsi-string consists of asterisks only, the asterisks can be omit-
ted. For example, the following line is equivalent to the one specified previously:
On some platforms (e.g., OpenStep), SANE device names take a special form. This is
explained below in the relevant platform-specific section.
When using a SCSI scanner, ensure that the access permission for the generic SCSI device
is set appropriately. We recommend to add a group "scanner" to /etc/group which contains
all users that should have access to the scanner. The permission of the device should
then be set to allow group read and write access. For example, if the scanner is at
generic SCSI device /dev/sge, then the following two commands would set the permission
$ chgrp scanner /dev/sge
$ chmod 660 /dev/sge
When your system uses the device filesystem (devfs), you have to edit /etc/devfs/perms.
There you should search the line
REGISTER ^sg[^/]* PERMISSIONS root.root 0600
and add a new line (eg. for changing permissions of sg4):
REGISTER ^sg4 PERMISSIONS root.scanner 0660
Reported to work fine under FreeBSD 2.2.2R with the aha driver.
Reported to work fine under FreeBSD 2.2.2.
The scanner probes ok but any attempt to access it hangs the entire system.
It looks like something is disabling interrupts and then not reenabling
them, so it looks like a bug in the FreeBSD aic driver.
Works on FreeBSD 2.2.5R and 3.0 using the aic driver, provided that Plug-
and-Play support is disabled on the card. If there are no uk devices, just
do a ``sh MAKEDEV uk0'' in the /dev directory. The scanner should then be
accessible as /dev/uk0 if it was probed during boot.
Reported to work fine under FreeBSD 2.2.2R with the amd driver.
First, make sure your kernel has SCSI generic support enabled. In ``make xconfig'', this
shows up under ``SCSI support->SCSI generic support''.
To keep scanning times to a minimum, it is strongly recommended to use a large buffer size
for the generic SCSI driver. From SG driver version 2.0 on, the maximum buffer size can be
changed at program run time, and there is no restriction in size. This driver version is
part of the Linux kernels from version 2.2.7 on. If the new SG driver is available some
backends (e.g. sane-umax, sane-mustek, sane-sharp) automatically request larger scsi buf-
fers. If a backend does not automatically request a larger scsi buffer, set the environ-
ment variable SANE_SG_BUFFERSIZE to the desired buffer size in bytes. It is not recom-
mended to use more than 1 MB, because for large values the probability increases that the
SG driver cannot allocate the necessary buffer(s). For ISA cards, even 1 MB might be a too
large value. For a detailed discussion of memory issues of the SG driver, see
For Linux kernels before version 2.2.7 the size of the buffer is only 32KB. This works,
but for many cheaper scanners this causes scanning to be slower by about a factor of four
than when using a size of 127KB. Linux defines the size of this buffer by macro
SG_BIG_BUFF in header file /usr/include/scsi/sg.h. Unless a system is seriously short on
memory, it is recommended to increase this value to the maximum legal value of
128*1024-512=130560 bytes. After changing this value, it is necessary to recompile both
the kernel (or the SCSI generic module) and the SCSI backends. Keep in mind that this is
only necessary with older Linux kernels.
A common issue with SCSI scanners is what to do when you booted the system while the scan-
ner was turned off? In such a case, the scanner won't be recognized by the kernel and
SANE won't be able to access it. Fortunately, Linux provides a simple mechanism to probe
a SCSI device on demand. Suppose you have a scanner connected to SCSI bus 2 and the scan-
ner has a SCSI id of 5. When the system is up and running and the scanner is turned on,
you can issue the command:
echo "scsi add-single-device 2 0 5 0" > /proc/scsi/scsi
and the kernel will probe and recognize your scanner (this needs to be done as root).
It's also possible to dynamically remove a SCSI device by using the ``remove-single-
device'' command. For details, please refer to to the SCSI-2.4-HOWTO.
Scanners are known to work with the following SCSI adapters under Linux. This list isn't
complete, usually any SCSI adapter supported by Linux should work.
Acard/Advance SCSI adapters
Some versions of the kernel driver (atp870u.c) cut the inquiry information.
Therefore the scanner can't be detected correctly. See http://www.meier-
geinitz.de/sane/trouble.html#acard for a solution.
Reported to work fine with Linux since v2.0. If you encounter kernel freezes
or other unexpected behaviour get the latest Linux kernel (2.2.17 seems to
work) or reduce SCSI buffer size to 32 kB.
Reported to work fine with Linux v2.0.
To configure the BusLogic card, you may need to follow these instructions
(contributed by Jeremy <firstname.lastname@example.org>): During boot, when your Bus-
Logic adapter is being initialized, press Ctrl-B to enter your BusLogic
adapter setup. Choose the address which your BusLogic containing your scan-
ner is located. Choose ``SCSI Device Configuration''. Choose ``Scan SCSI
Bus''. Choose whatever SCSI id that contains your scanner and then choose
``View/Modify SCSI configuration''. Change ``Negotiation'' to ``async'' and
change ``Disconnect'' to ``off''. Press Esc, save, and Esc again until you
are asked to reboot.
NCR/Symbios 53c400/53c400a or Domex DTC3181E/L/LE (DTCT436/436P) ISA SCSI card
This card is supplied by Mustek (and other vendors). It's supported since
Linux 2.2. The SCSI cards are supported by the module g_NCR5380. It's nec-
essary to tell the kernel the io port and type of card. Example for a
53c400a: ``modprobe g_NCR5380 ncr_addr=0x280 ncr_53c400a=1''. Once the ker-
nel detects the card, it should work all right. However, while it should
work, do not expect good performance out of this card---it has no interrupt
line and therefore while a scan is in progress, the system becomes almost
unusable. You may change the values of the USLEEP macros in driv-
ers/scsi/g_NCR5380.c. Some documentation is in this file and NCR5380.c.
For some scanners it may be necssary to disable disconnect/reconnect. To
achieve this use the option ncr53c8xx="disc:n". Some people reported that
their scanner only worked with the 53c7,8xx driver, not the ncr53c8xx. Try
both if you have trouble.
For Linux kernels before 2.0.33 it may be necessary to increase the SCSI
timeout. The default timeout for the Linux kernels before 2.0.33 is 10 sec-
onds, which is way too low when scanning large area. If you get messages of
the form ``restart (ncr dead ?)'' in your /var/log/messages file or on the
system console, it's an indication that the timeout is too short. In this
case, find the line ``if (np->latetime>10)'' in file ncr53c8xx.c (normally
in directory /usr/src/linux/drivers/scsi) and change the constant 10 to,
say, 60 (one minute). Then rebuild the kernel/module and try again.
The driver can be downloaded from http://www.garloff.de/kurt/linux/dc395/.
For some older scanners it may be necessary to disable all the more advanced
features by using e.g. modprobe dc395x_trm dc395x_trm=7,5,1,32.
Version 1.11 of the Tekram driver seems to work fine mostly, except that the
scan does not terminate properly (it causes a SCSI timeout after 10 min-
utes). The generic AM53C974 also seems to work fine and does not suffer
from the timeout problems.
Solaris, OpenStep and NeXTStep
Under Solaris, OpenStep and NeXTStep, the generic SCSI device name refers to a SCSI bus,
not to an individual device. For example, /dev/sg0 refers to the first SCSI bus. To tell
SANE which device to use, append the character 'a'+target-id to the special device name.
For example, the SCSI device connected to the first SCSI controller and with target-id 0
would be called /dev/sg0a, and the device with target-id 1 on that same bus would be
called /dev/sg0b, and so on.
If the library was compiled with debug support enabled, this environment variable
controls the debug level for the generic SCSI I/O subsystem. E.g., a value of 128
requests all debug output to be printed by the backend. A value of 255 also prints
kernel messages from the SCSI subsystem (where available). Smaller levels reduce
sets the timeout value for SCSI commands in seconds. Overriding the default value
of 120 seconds should only be necessary for very slow scanners.
sane(7), sane-find-scanner(1), sane-"backendname"(5), sane-usb(5)
13 Apr 2002 sane-scsi(5)