mfi(4) support for MegaRAID Fury cards

[Doug Ambrisko]

On Fri, Dec 27, 2013 at 05:04:55PM -0500, Mark Johnston wrote:
| Hello,
| 
| The patch here adds mfi(4) support for my LSI 9341-4i controller, which
| has device ID 0x5f:
| 
| http://people.freebsd.org/~markj/patches/mfi_fury.diff
| 
| This diff was mostly obtained by going through the mrsas(4) code
| specific to Invader (DID 0x5d) and Fury (DID 0x5f) controllers. The main
| change is to add an end-of-list marker to scatter-gather DMA lists
| before handing them to the firmware. Without this, large writes to an
| mfi(4) volume result in a firmware crash loop, and the system needs to
| be reset. The diff adds code for both Invader and Fury cards, as this is
| what's done in mrsas(4); I haven't tested with an Invader card though,
| as I don't have access to one. With this patch, I'm able to boot FreeBSD
| 8.2 off of a RAID 1 volume on my 9341-4i.
| 
| Would anyone be able to review or test this patch? I'm particularly
| interested if anyone could try it out with an Invader or Fury card
| (there shouldn't be any differences in driver behaviour with other
| cards).

The patch looks good.  I can test it out on a Invader card that I have.
I don't have a Fury card.  I was holding off waiting to see how we
should resolve the mrsas(4) driver from LSI conflict.  We have been
looking at what needs to be done to get mrsas(4) into FreeBSD.  I
posted a change to FreeBSD SCSI list to add a tunable to reduce
the probe priority of mfi(4) for ThunderBolt and later cards.  This
way they can both be in the GENERIC kernel etc. and not have an
issue.  We'll need to do some minor updates to your patch to work
with that since I added another flag in the ident area.

Sorry for the delay getting back to you.

Thanks,

Doug A.