[lug] Adding 2nd SCSI card

[D. Stimits]

Gary Hodges wrote:
> "D. Stimits" wrote:
> 
>>Gary Hodges wrote:
>>
>>>In the spirit of "short and sweet", what am I missing about getting the
>>>machine to boot from the LSI card?  At this point I'd like to just try
>>>it out without the Adaptec card.
>>
>>Items to check would be that you built support as a module, and included
>>the module in the mkinitrd; that your version of mkinitrd (or anything
>>related in kernel docs subdirectory Documentation/, file Changes, is up
>>to date). 
> 
> 
> mkinitrd was the ticket.  Thanks much for the suggestion.  A couple of
> us spent a good bit of time poking around late yesterday and it sure
> seemed like everything (mostly) was what it should be and where it
> should be.  In the end we made a new image file without making any other
> changes and it worked like a champ.  We made a few test reads and writes
> to the RAID tower and it worked fine.  We are suspecting the Adaptec
> card _may_ be the problem, but were going to hold off on making that
> judgment for a while still.  If it is the card we've had some bad luck
> with the hardware in that machine.  We replaced a bad 1 GB Corsair DIMM
> already.
> 
> While I'm on the subject I'd like to run a couple related things by the
> group.  Here is a listing of /proc/interrupts
> 
>            CPU0       CPU1       
>   0:    2862985    2880910    IO-APIC-edge  timer
>   1:       1080        979    IO-APIC-edge  keyboard
>   2:          0          0          XT-PIC  cascade
>   5:     100414      97914   IO-APIC-level  aic7xxx
>   8:          2          0    IO-APIC-edge  rtc
>   9:      31187      29965   IO-APIC-level  sym53c8xx
>  10:      60226      60544   IO-APIC-level  usb-ohci, eth0
>  11:         15         15   IO-APIC-level  sym53c8xx
>  12:      18940      17286    IO-APIC-edge  PS/2 Mouse
> NMI:          0          0 
> LOC:    5743845    5743825 
> ERR:          0
> MIS:          0
> 
> Here is modules.conf
> 
> alias parport_lowlevel parport_pc
> alias eth0 3c59x
> alias scsi_hostadapter aic7xxx
> alias usb-controller usb-ohci
> alias scsi_hostadapter1 sym53c8xx
> alias scsi_hostadapter2 sym53c8xx
> 
> The LSI Logic card is a single bus LVD card.  Why is it listed twice? 

Do you have 2 adapters of this type? I can't say with certainty, but I 
strongly suspect it is simply reference counting. According to the 
modules.conf, you have 3 scsi adapters, 2 are sym53c8xx, 1 is aic7xxx. 
If you don't have 2 sym53c8xx, perhaps somewhere in your setup you are 
referring to the same adapter with two names.

> If I do a "more" on /proc/scsi/aic7xxx/0 I get a lot of useful
> information about the Adaptec card.  If I do the same thing for the LSI
> card I get very little.  One last item...  When a "more
> /proc/scsi/aic7xxx/0" is done I see a message about a Corrupted Serial
> EEPROM.  It doesn't sound normal and I'm thinking that might be related
> to the problem we had.  Below is a result of both the "more
> scsi/aic7xxx/0" and "more scsi/sym53c8xx/1" commands.

I would wonder if the EEPROM has been flashed, or is in need of an 
update. My aic7xxx does not show the corrupted EEPROM in there. As far 
as sparser messages go on the other, I couldn't say...perhaps all such 
data appears only when a drive is mounted through the adapter, and 
nothing is running on it at the moment...or perhaps it is just voluntary 
data that isn't in the controller (verbosity of PCI bios data in PCI 
cards seems to vary widely). You might examine "lspci -vvv" output to 
see what it says, I believe it derives its information entirely from PCI 
bios, whereas it is possible the files you are using depend in part on 
the driver reporting code. The aic7xxx I seem to recall has a kernel 
compile option for verbose reporting (or maybe that was for scsi in 
general, I don't recall).

D. Stimits, stimits AT attbi.com