NAME

ipi − IPI disk interface

DESCRIPTION

All IPI disks connected to a Xylogics 7800 IPI-2 controller are accessed by the files in /dev/dsk for block mode access or /dev/rdsk for character mode access.  Each device name is of the form iXdYsZ where X is the IPI controller (0 through 4) Y is the drive number (0 through 15) on that controller, and Z is the disk partition number, vh for the disk header. 

Each Xylogics 7800 IPI-2 controller actually controls two physical IPI-2 channels. The logical disk unit numbering is set up such that even drive numbers should be on physical channel 0 and odd drive numbers should be on physical channel 1. This is a logical arrangement for performance reasons. As you add additional drives, you will then naturally balance the load onto different physical IPI-2 channels. 

FILES

/dev/{r}dsk/i?d[0-15]s? 
/dev/{r}dsk/i?d[0-15]vh

DIAGNOSTICS

Normally the ipi driver is fairly quiet. However, errors in accessing disks are logged to both the console and the file /usr/adm/messages. Errors that the driver considers cautionary will be logged as a NOTICE message, while errors that the driver considers needing more immediate attention will be logged as a WARNING message. 

The most interesting of all possible errors are those pertaining to potential media defects. The messages that may see that pertain to potential media defects are:

 Soft ECC Corrected
Auto Seek Retry Recovered
Soft Retry Recovered
 

while the messages that you may see that certainly show up media defects are:

 Hard Data ECC
Header Not Found
Header CRC Error
 

These messages will be usually accompanied with a rough indication as to relative block number (within a partition), absolute block number (from the start of the disk), and a triple of the form Cyl/Hd/Sec that gives the address of the beginning of the transfer that had an error. The aware system administrator can keep track of this error information and utilize it with ifmt(8) to deal with new media defects as they appear.

Other messages that might appear may indicate some transient condition that may not be worthy of notice. These messages are:

 Operation Timeout
VME DMA Timeout
Interrupt with No RIO?
 

Any other messages that appear, depending on whether they are flagged with the WARNING preface, usually indicate some hardware failure or configuration problem (like, the IPI cable has fallen off the drive). 

SEE ALSO

ifmt(8M)

September 02, 1992