SCSI Disk Controllers and Linux

Cable and Termination Problems

The SCSI bus needs terminators at both ends to work reliably. A terminator is a set of resistors that indicate the end of the SCSI bus. One end is the controller card itself, which typically has the terminator on it. Often, a SCSI device has two SCSI connectors, so that you can daisy-chain several external SCSI devices. You are supposed to place a terminator on the last connector on the chain.

To avoid problems with overly sensitive SCSI cards, use cables that come from a reputable vendor, and use cables from the same vendor to connect all SCSI devices. The cables should be SCSI 2-compliant and should have an impedance of 132 ohms (a characteristic of the cable; all you have to do is make sure the specified value is 132 ohms).

Most SCSI problems are due to bad cables or improper termination. You should check the cables and the terminator before trying anything else.

SCSI Device at All SCSI IDs

If a SCSI device shows up at all possible SCSI IDs, you must have configured that device with the same SCSI ID as the SCSI controller (usually, 7). Change the ID of that device to another value. (Many devices have a simple switch for setting the SCSI ID; on others, you have to change a jumper.)

SCSI Device at All LUNs

If a SCSI device shows up at all possible SCSI logical unit numbers (LUNs), the device probably has errors in the firmware-the built-in code in the device's SCSI interface. To verify these errors, use the following command line during boot:

max_scsi_luns=1

If the device works with this option, you can add it to the list of blacklisted SCSI devices in the array of structures named device_list in the file /usr/src/linux-2.4/drivers/ scsi/scsi_scan.c. The definition of that structure and some parts of the array are as follows:

struct dev_info{
    const char * vendor;
    const char * model;
    const char * revision; /* Latest revision known to be bad. Not used yet */
    unsigned flags;
};
/*
 * This is what was previously known as the blacklist.  The concept
 * has been expanded so that we can specify other types of things we
 * need to be aware of.
 */
static struct dev_info device_list[] =
{
/* The following devices are known not to tolerate a lun != 0 scan for
 * one reason or another.  Some will respond to all luns, others will
 * lock up.
 */
{"Aashima","IMAGERY 2400SP","1.03",BLIST_NOLUN},/* Locks up if polled for lun !=  0 */
{"CHINON","CD-ROM CDS-431","H42", BLIST_NOLUN}, /* Locks up if polled for lun !=
 0 */
{"CHINON","CD-ROM CDS-535","Q14", BLIST_NOLUN}, /* Locks up if polled for lun !=
 0 */
{"DENON","DRD-25X","V", BLIST_NOLUN},      /* Locks up if probed for lun != 0 */
{"HITACHI","DK312C","CM81", BLIST_NOLUN},  /* Responds to all lun - dtg */
{"HITACHI","DK314C","CR21" , BLIST_NOLUN}, /* responds to all lun */
{"IMS", "CDD521/10","2.06", BLIST_NOLUN},  /* Locks-up when LUN>0 polled. */
{"MAXTOR","XT-3280","PR02", BLIST_NOLUN},  /* Locks-up when LUN>0 polled. */
... lines deleted ...
        {"HITACHI", "DF600", "*", BLIST_SPARSELUN},
        {"IBM", "ProFibre 4000R", "*", BLIST_SPARSELUN | BLIST_LARGELUN},
/*
 * Must be at end of list...
 */
{NULL, NULL, NULL}
};

From this list, you can get an idea of the types of SCSI devices that have the problem of showing up at all LUNs. If you have the same problem with a SCSI device, you can add that device's name to this list before the last line before rebuilding the kernel. You may not want to do this, however, if you are not familiar with the C programming language.

Sense Errors on an Error-Free SCSI Device

The cause of this problem usually is bad cables or improper termination. Check all cables, and make sure the SCSI bus is terminated at both ends.

Networking Kernel Problems with a SCSI Device

If a Linux kernel with networking support does not work with SCSI devices, the problem may be the autoprobe function of the networking drivers. The autoprobe capability is meant to detect the type of networking hardware automatically; the network drivers read from and write to specific I/O addresses during autoprobing. If an I/O address happens to be the same as that a SCSI device uses, the system may have a problem. In this case, you have to check the I/O address, IRQ, and DMA values of the network cards and of the SCSI controller and make sure no conflicts exist. Most SCSI controllers (and even network adapters) allow you to configure these parameters (I/O address, IRQ, and DMA) through setup software that comes with the adapter.

Device Is Detected but Not Accessible

If the kernel detects a SCSI device (as reported in the boot messages, which you can see by typing the command dmesg | more) but you cannot access the device, the device file may be missing from the /dev directory.

cd /dev
./MAKEDEV st0

man MAKEDEV

SCSI Lockup

If the SCSI system locks up, check the SCSI controller card, using any diagnostic software that came with the card (usually, the diagnostic software runs under DOS). Look for conflicts in I/O address, IRQ, or DMA with other cards. Some sound cards, for example, use a 16-bit DMA channel in addition to an 8-bit DMA; make sure you have not inadvertently used the same 16-bit DMA for the SCSI card.

The Linux SCSI driver for some SCSI cards supports only one outstanding SCSI command at a time. With such a SCSI card, if a device such as a tape drive is busy rewinding, the system may not be able to access other SCSI devices (such as a hard disk or a CD-ROM drive) that are daisy-chained with that tape drive. A solution to this problem is to add a second SCSI controller to take care of the tape drives.

SCSI Devices Not Found

If the Linux kernel does not detect your SCSI devices at startup, you get the following message when Linux boots:

scsi : 0 hosts
scsi : detected total

If you see this message, but you know that the SCSI devices are there (and that they work under DOS), the problem may be the lack of a BIOS on the SCSI controller; the autoprobe routines that detect SCSI devices rely on the BIOS.

This problem occurs for the following SCSI cards:

• Adaptec 152x, 151x, AIC-6260, and AIC-6360

• Future Domain 1680, TMC-950, and TMC-8xx

• Trantor T128, T128F, and T228F

• Seagate ST01 and ST02

• Western Digital 7000

  Insider Insight

  Even if a SCSI controller has a BIOS, jumpers often are available for disabling the BIOS. If you have disabled the BIOS for some reason, you may want to reenable it (read the documentation of your SCSI controller for directions) so that Linux can detect the SCSI devices automatically.

For a SCSI card, such as the Adaptec 151x, that does not have any BIOS, use the following command line during boot to force detection of the card:

aha152x=0x340,11,7,1