Problems with stable/9 and LSI controlers?

[Kenneth D. Merry]

On Tue, Sep 10, 2013 at 12:35:02 -0700, Dennis Glatting wrote:
> On Tue, 10 Sep 2013, Kenneth D. Merry wrote:
> >On Tue, Sep 10, 2013 at 12:11:38 -0700, Dennis Glatting wrote:
> >>On Tue, 10 Sep 2013, Kenneth D. Merry wrote:
> >>
> >>>On Tue, Sep 10, 2013 at 11:21:02 -0700, Dennis Glatting wrote:
> >>>>
> >>>>Today I updated several systems against stable/9 that had not bee 
> >>>>updated
> >>>>in ~five months. Two of the systems are serviced by LSI 2008 controllers
> >>>>with a RAID1 array housing the operating system.
> >>>>
> >>>>These systems no longer properly boot.
> >>>>
> >>>>Specifically, they begin the boot process but bomb when they try to 
> >>>>mount
> >>>>root. Neither system sees the GPT partitions of the boot disk at the
> >>>>"mountroot>" prompt. On one system I replaced the hard disks,
> >>>>repartitioned, and reinstalled but the problem continued.
> >>>>
> >>>>I don't see any mention of geom or controller update in UPDATING but I
> >>>>know they have been updated because I saw their revisions fly actross my
> >>>>screen during the svn update. Also, I am running IR vers 16 on my
> >>>>controllers, which is the latest version of LSI firmware.
> >>>>
> >>>>Is there some new action I have to take to sucessfully boot LSI disks
> >>>>against the updated kernel?
> >>>
> >>>Can you try the mpslsi driver from LSI's web site and see if that works?
> >>>
> >>>If it does, then it is a problem we've introduced locally.
> >>>
> >>
> >>I'll give it a shot but just to be clear, I have been running the 16
> >>firmware for months.
> >
> >The issue is very likely in the driver, not the firmware.  I'm just
> >trying to figure out whether it broke in LSI's Phase 16 or in our
> >additional changes in FreeBSD.
> >
> >In any case, I've reproduced the problem, and hopefully I'll know soon
> >where the issue is.
> >
> >So don't worry about trying LSI's driver.  Just go back to the previous
> >version of stable/9 that worked for you until we fix it.
> >
> >This only affects RAID volumes, not regular disks.
> >
> 
> Thanks.
> 
> I downloaded the mpslsi.ko driver from LSI under r255451 and the system 
> now boots.

Okay, try the attached patch with the new driver and see whether it fixes
the problem for you.

The issue was that with the new rescanning code, we weren't rescanning RAID
volumes that got created.

It fixes the problem for me with a RAID-1 volume.

There is still the issue of all of the verbose probe errors that crop up
when an Integrated RAID volume is enabled, but it isn't obvious at the
moment whether we can easily fix that.

Ken
-- 
Kenneth Merry
ken at FreeBSD.ORG
-------------- next part --------------
==== //depot/users/kenm/FreeBSD-stable2/9/sys/dev/mps/mps_sas_lsi.c#10 - /usr/home/kenm/perforce4/kenm/FreeBSD-stable2/9/sys/dev/mps/mps_sas_lsi.c ====
*** /tmp/tmp.91039.2	Tue Sep 10 14:13:13 2013
--- /usr/home/kenm/perforce4/kenm/FreeBSD-stable2/9/sys/dev/mps/mps_sas_lsi.c	Tue Sep 10 14:07:08 2013
***************
*** 892,898 ****
--- 892,901 ----
  		free(lun, M_MPT2);
  	}
  	SLIST_INIT(&targ->luns);
+ #if ((__FreeBSD_version >= 1000000) && (__FreeBSD_version < 1000039)) || \
+     (__FreeBSD_version < 902502)
  	if ((sassc->flags & MPSSAS_IN_STARTUP) == 0)
+ #endif
  		mpssas_rescan_target(sc, targ);
  	mps_dprint(sc, MPS_MAPPING, "RAID target id %d added (WWID = 0x%jx)\n",
  	    targ->tid, wwid);