Re: [Ipoverscsi-developer] Adaptec

[<ski...@su...>]

On Mon, Nov 19, 2001 at 02:52:19PM +0000, Mark Kilmartin wrote:
> OK I had a better look through the AIC7xxx code this weekend.
> 
> I believe the commands it needs to support are INQUIRY and WRITE10 is
> this correct?
> Is there more?
> 

You need some more. WDTR (Wide data transfer request) for 16 bits
comunications. SDTR (Sync data transfer request) Syncrouns transfer, for
more than 10MB/s transfer rate. IDENTIFY so that you can choose you LUN.
This last is very easy. You just accept it. It normaly appears before
WDTR and SDTR. This messages are sent by the other controler to
negotiate the speed of comunications. But you also need COMMAND COMPLETE
to indicate if the command was done correctly.

> It seems to support the INQUIRY command but I'm not sure if it supports
> it enough for the purposes.

The only thing that INQUIRY is needed is to say to the other party what
this board is. I think it's the first command to implement. When you can
do a IDENTIFY all the others are more or less the same.

You will need the SCSI standart to see how to do this messages. We got a
copy in the internet, but it wasn't easy. If you whant, i can send you
a copy. 

> It does not seem to support the WRITE10 command though :(

WRITE10 is the way we use to send data. It's the main command. If you
got WRITE10 and IDENTIFY done, it will work, but slowly (5/10 MB/s). All
the other commands are not essenstial. If you make IDENTIFY work,
WRITE10 is easy. 

> 
> What I now need is any advice on how to get it to support the WRITE10
> command.

In our case, the SYM53C8XX has a integrated processor. We can program it
using a special language called SCRIPTS. What we made was to change the
flow of the program so that it would make a WAIT SELECT. This command
waits for a controler to select our target. Then we decoded the command
and made a hardware interrupt so that to read the information. As I
said, the best way is to do INQUIRY and then make WRITE10. I will check
the code to see if I can help anymore.

> 
> MArk
> 
> 
> 
Pedro Semeano