ciss(4) HP Smart Array P840

[Scott Long]

> On Nov 19, 2015, at 12:44 PM, Jonathon Reinhart <jonathon.reinhart at gmail.com> wrote:
> 
> On Thu, Nov 19, 2015 at 12:24 PM, Sean Bruno <sbruno at freebsd.org> wrote:
>> On 11/19/15 09:03, Jonathon Reinhart wrote:
>>> Hello all,
>>> 
>>> TL;DR: Why isn't the P840 supported by ciss(4)?
>>> 
>> I think I was the last person actively maintaining this driver for an
>> employer.  HP doesn't have staff to do this support any longer and I
>> no longer work/have this hardware.
>> 
>> That's pretty much it as far as I can tell with this product line.
>> Someone has to put in the effort to get farther along with the hardware.
>> 
>>> 
>>> I tried to setup FreeNAS 9.3 (running 9.3-RELEASE-p25) on an HP
>>> DL380 Gen9, with the P840 RAID controller. I didn't realize that
>>> the P840 is not officially supported by the ciss(4) driver. In my
>>> troubleshooting, I ended up in the same place as documented on this
>>> forums.freenas.org post
>>> <https://forums.freenas.org/index.php?threads/hp-gen9-server-w-p840-hba-mode-no-drives-visible.28620/>:
>>> 
>>> 
>>>> It's interesting that camcontrol devlist shows the drives, but
>>>> there are no da* or ada* designations for them. I don't know what
>>>> that means, but I'd speculate it's something driver-related.
>>> 
>>> Indeed:
>>> 
>>> # camcontrol devlist -v scbus0 on ciss0 bus 0: scbus1 on ciss0 bus
>>> 32: <ATA ST1000DM003-1ER1 CC46>        at scbus1 target 4 lun 0
>>> (pass0) <ATA ST1000DM003-1ER1 CC46>        at scbus1 target 5 lun 0
>>> (pass1) scbus2 on ciss0 bus 33: ...
>>> 
>>> But there are no device names listed there, after passX, or in
>>> /dev.
>>> 
>> 
>> ciss(4) has "scsi-like" features which mean that behavior like this is
>> kind of expected.
>> 
>>> I see in the ciss source
>>> (https://github.com/freebsd/freebsd/blame/master/sys/dev/ciss/ciss.c#L320)
>>> 
>>> 
>> that the VID/DID for this card (0x103C, 0x3239) is listed.
>>> 
>> 
>> Yup.  I was asked to add that almost 3 years ago by HP before their
>> team and my employment ended.
>> 
>>> I'm new to FreeBSD, but I'm confused as to how the driver supports
>>> this card enough to read the drive model numbers, but not enough
>>> to actually use the drives. Is there something I'm missing, or is
>>> there some other technical reason this controller can't be used?
>>> 
>>> Thank you for your time,
>>> 
>>> Jonathon Reinhart
>> 
>> 
>> I'm more than willing to help out here, you might want to push on HP
>> if possible to get someone to reach out to me with details on what
>> needs to be modified to support this controller.  I no longer have
>> valid contacts to get this information.  :-(
>> 
>> sean
>> 
>> bcc jonathon
> 
> Thanks for the quick reply, Sean.
> 
> I took a quick look at the Linux hpsa driver (which obsoletes cciss)
> to see if they treat these new cards any differently, and they don't
> appear to do so, at all. AFAICT every supported board uses the same
> "SA5_access" method. Which makes me again wonder why, given the
> FreeBSD driver has the right VID/DID, it doesn't *just work*. I'd
> probably have to turn on CISS debugging to see if there's any useful
> info.


Since you can see the volumes in camcontrol, the low-level driver and communications protocol is working.  What’s odd at the moment is that the volumes are returning an inquiry response that CAM doesn’t fully understand; that’s why you get the pass devices but no da devices.  Can you use camcontrol to do an inquiry request and send me the results?  Something like the following:

sudo camcontrol cmd -n pass -u 0 -v -c "12 00 00 00 ff 00" -i 255 - | hd

Thanks,
Scott