Trouble with LSI SAS1078

[Marius Strobl]

On Thu, Feb 09, 2012 at 01:48:32PM +0400, Slawa Olhovchenkov wrote:
> On Wed, Feb 08, 2012 at 12:18:18AM +0100, Marius Strobl wrote:
> 
> > On Wed, Feb 08, 2012 at 02:03:25AM +0400, Slawa Olhovchenkov wrote:
> > > On Tue, Feb 07, 2012 at 10:32:53PM +0100, Marius Strobl wrote:
> > > 
> > > > On Wed, Feb 08, 2012 at 12:57:50AM +0400, Slawa Olhovchenkov wrote:
> > > > > On Tue, Feb 07, 2012 at 09:42:55PM +0100, Marius Strobl wrote:
> > > > > 
> > > > > > > > > I can test you patch at this machine (currently installed on USB HDD).
> > > > > > > > 
> > > > > > > > Okay, could please give the following patch a try?
> > > > > > > > http://people.freebsd.org/~marius/mpt_sas1078.diff
> > > > > > > 
> > > > > > > On first look its working!
> > > > > > > Many strange messages in dmesg (Error 6, Unretryable error; Unexpected
> > > > > > > Bus Free; etc). Delay on load driver (kern.cam.scsi_delay may be?).
> > > > > > > 
> > > > > > > I can read first 100 secrors from da1 and da2.
> > > > > > > 
> > > > > > 
> > > > > > Some delay of maybe 2-4 seconds when mpt(4) is attaching is normal.
> > > > > 
> > > > > No, delay is 12.40s.
> > > > 
> > > > Hrm, could you please change the SAS1078 part in mpt_hard_reset() of
> > > > mpt.c to look like the following, i.e. add the printf()'s:
> > > >         if (mpt->is_1078) {
> > > >                 printf("%s: reg before=0x%x\n", __func__,
> > > >                      mpt_read(mpt, MPT_OFFSET_RESET_1078));
> > > >                 mpt_write(mpt, MPT_OFFSET_RESET_1078, 0x07);
> > > >                 DELAY(1000);
> > > >                 printf("%s: reg after=0x%x\n", __func__,
> > > >                      mpt_read(mpt, MPT_OFFSET_RESET_1078));
> > > >                 return;
> > > >         }
> > > > 
> > > > and report the values printed during attach of mpt(4) with the above
> > > > compiled in?
> > > 
> > > don't print anything. and i don't see 'hard reset' message in verbose boot.
> > > 
> > > > Okay, so apart from the lengthy delay during attach it looks good so
> > > > far. Do you have any problems using the disks?
> > > 
> > > I didn't try realy using.
> > 
> > The clutter in the verbose dmesg might be from an enclosure that
> > either only fully comes up when actually booting from a disk
> > attached to the SAS1078 or isn't properly supported by ses(4),
> > delaying the path inquiry. I'd suggest to try to boot from disk
> > and to leave out the ses(4) driver.
> 
> I am try using SAS1078:
> 
> - mptutil working, RAID1 created.
> - raid1 rebuild working
> - boot from array working
> 
> Now system boot from SAS1078 and working!
> 

Thanks for testing! I've committed the patch.

Marius