[Coco] BUG with rbsuper and tc3 drivers SOLVED!

[Tormod Volden]

On Tue, Feb 14, 2017 at 1:51 AM, Tim Fadden wrote:
> On 2/13/2017 5:36 PM, Allen Huffman wrote:
>>>
>>> On Feb 13, 2017, at 6:23 PM, Tim Fadden <t.fadden at cox.net> wrote:
>>>
>>> And in hind site, you may have to create entirely new descriptors for any
>>> existing dirves/partitions.  The ones I had did not work.
>>
>> What made existing descriptors incompatible with the update?
>>
>>                 — A
>>
>>
> ha ha ha,  I was so happy to get it figured out I did not try!  Now that I
> have things working again I will see if I can find out.
>
> During my testing, any time I would access my custom drivers, the system
> would hang. Even dmode!  I am guessing one of the settings I had on them was
> compatible with the old drivers, but not the new. The older stock ones
> worked  with the new drivers. When I get everything sorted out, i will try
> to see just what that was.

dmode doesn't call the driver so that sounds strange. Can you please
send these descriptors to me. At least dmode issues I can reproduce
without having SCSI disks :)

What remains unclear to me is whether the DRV field has to be unique
per controller, for every target and partition. I get a vague
impression of this after reading the SuperDriver manual, and I also
see rbsuper use the DRV number for its cache matching logic.

I wonder if reading a sector from target 4 just after writing to the
same sector number to target 3 would return what was written to target
3, unless also the DRV numbers are different? Especially since rbsuper
is unaware of DNS being used for target number. Or likewise reading
from one partition after writing to the same logical sector on another
partition, because although it is rbsuper that adds the partition
offsets, it does so after checking for cache hit and only just before
calling the low-level driver on cache miss.

Tormod