HP NetRAID-4M Configuration and Upgrade Guide (Release 5)

Chapter 9 Issues and Problem Resolution

Mfg diagnostic scsi_x.exe encounters the problem by trying to do a bus scan on all

possible host adapter numbers. When it does it at a nonexistent HA# after the inquire

command is executed it enters an infinite loop, polling and waiting to see a

command done status, but this never occurs.

The driver should treat the inquire command to non-existing HA# as if no device

found and returned the same command done status. This is what all other DOS ASPI

drivers from other vendors do. Adaptec needs to resolve this problem.

Impact to customer: Some DOS based diagnostic programs cannot access the

control-A BIOS setting.

Resolution: The root cause for this problem was that the card was not checking for

an invalid adapter return status. If the adapter did not exist the system still went

ahead and assumed that everything was OK and this caused a problem in the code

flow. To fix this problem, the code now includes the invalid adapter status check.

8 Retry interval on broadcasting a SMART event set to 500min!

Problem: The Interval Time is being change from 0x0BB5 (5min) to 0x493E0

(500min) in Page 1C, after first SMART event. The controller is issuing a Mode

Select to the drive when the drive is re -inserted after setting the Test Bit.

Impact to customer: Low. SMART reliability reporting has a very long polling

cycle.

Resolution: Reset interval time to 0x0BB5 (5min).

9 Multiple drive failures

Problem: It appears that the SCSI pass through commands are NEVER assigned a

window to the host memory. On further investigation it appears that from 0 to 3 G

is covered with the 2 fixed windows of the 21554 chip and then the LookUp entries

are used for commands above 3 G. Normal code flow obtains a lookup entry

correctly, but the pass thru commands bypass this code. Therefore the command will

fail, unless a normal command created an entry (which stays around until needed

again) that covered the target address, thereby creating a random and inconsistent

failure occurrence.

Impact to customer: Serious. RAID can fail and data will need to be restored from

current back-up.

Resolution: Pass-thru code was fixed to assign proper SG map window.

10 Regression: NetWare fails unload certification test