LSI-MegaRAID 150-4 BTX Halted on 5.4, 5.5, 6.1

[Carroll Kong]

> -----Original Message-----
> From: John Baldwin [mailto:jhb at freebsd.org] 
> Sent: Tuesday, August 22, 2006 3:07 PM
> To: freebsd-hackers at freebsd.org
> Cc: Carroll Kong
> Subject: Re: LSI-MegaRAID 150-4 BTX Halted on 5.4, 5.5, 6.1
> 
> On Tuesday 22 August 2006 00:12, Carroll Kong wrote:
> > I am trying to install FreeBSD on a new Intel Server 
> SE7230NH1-E using 
> > a PCI-X riser card on the Intel S1475 chassis, pentium D 
> 3.2 gig dual 
> > core proc 940.
> > 
> > Whenever I try to install using the CDROM for 5.4, 5.5, and 
> 6.1... BTX 
> > halts immediately.
> > 
> > The second I remove the card, the system boots up fine.  In fact, I 
> > was able to install 6.1 on one of the SATA disks on there.  
> However, 
> > once I put the card back, BTX Halts.
> > 
> > BTX Halts even if I remove all logical drives on the array 
> (making it 
> > completely empty and it does not show up as a disk at all 
> in the BTX Bios).
> > I even disabled the card's BIOS mode, and it still halts.
> > 
> > Since FreeBSD 5.4 supports the LSI Megaraid 150-4, I 
> suspect it might 
> > be the riser card doing interesting things.
> > 
> > I highly doubt hardware is the issue since I was able to install 
> > CentOS without a hitch (eek, I really don't want to use it 
> though... 
> > unless Vmware can run a freebsd box from it).  Of course, the 
> > possibility of Linux ignoring potentially critical errors 
> is another 
> > possibility.  :)
> > 
> > Just a wild guess here since I have no real hardware 
> programming experience.
> > I really think it is the riser card probably doing some 
> different alignment.
> > I cannot test the card without the riser (it's a weird board that 
> > needs the riser card to 'automatically' mix to the right 
> modes I think?).
> > 
> > Here is the BTX dump.  It is copied verbatim from a screen shot.  
> > (hopefully I wrote it out exactly)
> > 
> > int=0000000d err=00000013 efl=00030402 eip=0000554d
> > eax=00000204 ebx=00000000 ecx=00000001 edx=00001421
> > esi=00000008 edi=00000008 ebp=00000000 esp=0000040c cs=f000 ds=3ec9 
> > es=44b0 fs=0000 gs=0000 ss=9e4c
> > cs:eip=e6 e4 e4 71 c3 53 b7 00-eb 08 53 b7 01 eb 03 53
> >        b7 02 9c fa 8a d8 8a c4-e8 e3 ff 80 ff 00 75 04
> > ss:esp=36 54 4a 91 00 00 96 02-b4 11 05 00 44 1d 05 00
> >        f8 48 09 00 84 9c 00 00-00 00 00 00 b4 11 05 00
> > 
> > So, I CAN boot into FreeBSD 6.1 if I remove the card.  If I need to 
> > recompile something, it should be doable.  Thanks in advance guys!
> 
> Hmmm.  This is quite odd.  The instruction is one that should 
> be allowed:
> 
> 00000000  E6E4              out 0xe4,al
> 00000002  E471              in al,0x71
> 00000004  C3                ret
> 
> My guess is that somehow the TSS has been corrupted.  Many 
> years ago Mike Smith was running into weirdness with a RAID 
> BIOS (not amr(4) I don't think, maybe mlx(4)) that was 
> somehow corrupting the TSS.  I don't know if he ever managed 
> to solve it.
> 
> --
> John Baldwin

Forgive me for my ignorance, but what is a TSS?  Also, I told the card to
disable it's BIOS during bootup, and it still zonked.  I even removed all
the logical drives (so there was no RAID array) just to see if it would
boot, it would still zonk!  Only if I physically removed it, would it work.
I figured, at that point why would BTX even care about the RAID controller?
It's not being asked to boot from it anymore.

Oddly enough, CentOS is working great on it now.  I have 2 Vmware guests
running... you guessed it -- FreeBSD 6.1.  However, if this TSS sounds like
a hardware issue, I will probably have to do more burn-in tests to ensure it
is not just the linux driver "ignoring" something.

I also wonder if this would have been easier if I just went with the LSI
Megaraid SCSI controller instead.  I went with the SATA version since it was
supposedly well supported by FreeBSD 5.4.  Although, I am more likely to
blame the riser card for confusing FreeBSD, I do not believe I have a
hardware issue like the other gentlemen.  (I have reinstalled the OSes and
guest OSes multiple times).



- Carroll Kong