new isp test code/24XX

[mjacob at freebsd.org]

The 24XX support is going in, yes. That's the reason for the major 
rototilling.

And, yes, target mode is part of it. The work is actually being paid for 
by a client of mine who needs Linux target mode. I suggested the SCST 
package, but since they'd been using my driver in production for just 
about 3 years now they wanted to stick with it, so this work has been 
being done to accomodate them, and the platform-specific FreeBSD part 
should be relatively small.

The current status of the newisp branch:

a) General support: the new fabric/loop code seems to be settling down 
to be relatively stable. I still need to implement some policy knobs 
that actually specify the amount of time to wait before declaring a 
device dead (path failover is part of this effort).

b) The 24XX initiator mode support seems to be pretty stable at this 
time. The 24XX target mode support isn't completed yet. I'm now finally 
getting CMND IUs coming but still need to connect the dots to process 
them.

c) Regression testing:

  + I've validated that at least parallel SCSI/isp still works on sparc64
    on a 1040c. I have no 1280s or 12160s to test with any more.

  + I still need to check FC *and* the 24XX on sparc64

  + I still need to check out SBus/sparc

  + I've tested 2200s and 23XX cards. I don't have any 2100s any more to 
test with.

d) Plans

  + Once I check SBus out, I'll probably push newisp into -current to get 
some broader exposure. This is likely to happen by the end of the month.

If anyone wants to try this stuff earlier, please feel free and take a 
child of the newisp branch (it's locked). I keep it pretty in sync with 
-current.


> Hello Matt,
>
> Two questions:
> - is the 2462 supported by this driver?
> - what about target mode? :)
>
> Thanks for your valuable work.
>
> On 08/26/06 21:01, Matt Jacob wrote:
>> In order to accomodate the 4Gb QLogic card, the isp driver has had to be
>> restructured in a pretty large way. The 24XX cards have the same rough
>> architecture as all the other QLogic cards, but have massively changed
>> inside (16 -> 32 bit registers; IOCBs that have same names and type
>> fields but different layouts).
>> 
>> Between this and the desire to make some changes that will make failover
>> and N-port virtualization in target mode easier to handle, loop and fabric
>> evaluation have also been rewritten (which also better accomodates the
>> 2K port login f/w which is now standard even for 23XX cards). This is
>> a big change and needs some serious test time to get right.
>> 
>> The changes were so much that rather than impose this even on
>> -current, some settle time in the FreeBSD Perforce sandbox seemed
>> appropriate.
>> 
>> If anyone would like to beta test these changes, please let me know.
>> The more feedback now, the more painless the change later. The
>> perforce depot is //depot/client/newisp.
>> 
>> -matt
>> 
>> 
>> _______________________________________________
>> freebsd-scsi at freebsd.org mailing list
>> http://lists.freebsd.org/mailman/listinfo/freebsd-scsi
>> To unsubscribe, send any mail to "freebsd-scsi-unsubscribe at freebsd.org"
>> 
>
>
> -- 
> Attila Nagy                                   e-mail: Attila.Nagy at fsn.hu
> Free Software Network (FSN.HU)                 phone: +3630 306 6758
> http://www.fsn.hu/
>
>