Doug,

I am running an IBM Netfinity 5000 with dual PIII-500's and 256mb of 
memory. It has a dual channel aic7xxx chipset on the motherboard and an 
IBM ServeRAID controller in one of the PCI slots. I was using only the 
built-in 10/100 Ethernet controlled (an AMD PC-net chipset) and had 
everything working as expected.

The only hard drive in the system is the RAID array on the IBM 3L 
controller, running GA version 1.0 of the ips driver. There is a DLT-IV 
(35/70) tape drive attached to the external interface on channel A of 
the dual channel aic7xxx. The CD-ROM is IDE. The way the scsi "hosts" 
break down is:
SCSI0 is channel B, the internal channel of the aic7xxx. There are no 
devices attached.
SCSI1 is channel A, the external channel of the aic7xxx, with the DLT 
35/70 drive at address 5 (I have also tried 2,3,4 and 6). The drive is 
detected and named st0.
SCSI2 is the ServeRAID 3L with the RAID array pseudo disk at sd0 and 
the RAID controller at sg0.

The way the irq's break down is:

15 - SCSI0
14 - IDE
13 - numeric coprocessor
12 - PS/2 mouse
11 - shared by SCSI1 and SCSI2
10 - ethernet
9  - ???
8  - ???
7  - printer
6  - floppy
5  - not used
4  - ttyS0 - a modem used for paging
3  - ttyS1 - the UPS monitor
2  - cascade
1  - keyboard
0  - clock/timer?

At this point everything worked fine.

Then I had to add a token-ring adapter because this box is going to 
have to talk to both our token ring and ethernet networks, but out 
token-ring network is gradually being isolated to a UNIX only network 
with no routing/gateways between the two.

Linux, as near as I can tell, does not yet support PCI token ring 
adapters, at least not any I have here. Therefore I am using an IBM ISA 
Token Ring Auto 16/4 Adapter (we have lots lying around) which is 
limited to IRQ's 3,9,10,11. On IRQ 9 the card was not even seen by the 
driver, so I had only one option for IRQ for the token ring card, and 
that was IRQ 3. This meant I had to disable the internal serial port B 
and add a serial card on which one of the serial ports could be moved 
to the unused IRQ 5.

I have experimented with two BIOS settings.

1. leave IRQ 5 in the PNP range. When I do this, the system boots fine, 
but my serial port communication through ttyS1 to the UPS is very 
unreliable. In fact only about 10% of the time will it establish 
communications with the UPS at all, and even then it will lose contact 
for periods of a few seconds to several minutes.

2. assign IRQ 5 to ISA Legacy resources. When I do this, the boot 
appears to hang discovering SCSI devices. If I leave it long enough, 
the tape drive is discovered at every possible SCSI address on the bus 
and then the process begins to loop.

What is the interaction of the device discovery process and IRQ's? What 
driver settings might I be able to use to circumvent this process? Or 
do I not have any options here?

I am currently running RedHat 6.0 with patches through July 28th. So I 
am running kernel level 2.2.5-22 and whatever aic7xxx driver came with 
that kernel. I believe the original 6.0 kernel had an aic7xxx driver 
that would often not work on this system at all.

Werner Kliewer
Manitoba Public Insurance
Werner Kliewer
Manitoba Public Insurance



-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]

Reply via email to