Re: Any successful installs on a Broadcom HT1000 chipset?

On Wednesday 21 November 2007 12:24:14 pm Travis Mikalson wrote:
> Freddie Cash wrote:
> > On November 21, 2007 04:26 am Barney Cordoba wrote:
> >> It would be a major
> >> black eye to the project to say "FreeBSD 7 is
> >> released, but BTW, if you have a supermicro or tyan
> >> opteron system it won't work at all".
> >
> > The HT1000 is only 1 of the Opteron chipsets available from Tyan, and
> > it's only available on their 1 socket boards (the Tomcats).  The dual-
> > and quad- socket boards either use AMD or AMD+nVidia chipsets (Socket
> > 939/940), or nVidia chipsets (Socket AM2).  Out of their entire lineup,
> > the HT1000 makes up maybe 10% of the chipsets used by Tyan.
> >
> > I'd hardly call that a "black eye".  And it's only the onboard SATA
> > controller that has issues.  Pop in a proper RAID controller, and it's no
> > longer an issue.
>
> Hate to sound like a broken record but that's exactly what I did and I
> ran into the same nasty data corruption problem on that controller, too.
> That was a little frustrating.
>
> atapci0_at_pci0:2:3:0:     class=0x010000 card=0x11ab11ab chip=0x608111ab
> rev=0x09 hdr=0x00
>      vendor     = 'Marvell Semiconductor (Was: Galileo Technology Ltd)'
>      device     = 'MV88SX6081 8-port SATA II PCI-X Controller'
>      class      = mass storage
>      subclass   = SCSI
>
> I don't want a hardware RAID or a "software RAID" controller, I just
> want normal SATA ports with no RAID so I can use ZFS :)
>
> I find myself much preferring ZFS to any other data management solution
> I have ever used, including 3ware hardware RAID which was my previous
> favorite.
>
> Another thing I haven't tried yet is a 3ware 9500 series PCI-X
> controller in my HT1000 board's PCI-X slot to see if that's also broken
> or if it's just a PCI-X + ata(4) specific problem.
>
> I'll reply with the result of that when I get to trying it out.
>
> >> Someone was complaining 2 years ago that they didn't
> >> have hardware to do the work. Is the FreeBSD project
> >> really so poorly funded that they can't get their
> >> hands on a $150 motherboard for 2 years?
> >
> > Ever considered sending one in?  :)
>
> I am considering it. Are any reputable FreeBSD developers interested in
> this problem that could benefit from having the problematic hardware in
> their posession?

I'm in the middle of retiring nearly a dozen supermicro socket AM2 boards with 
this chipset.  With the SATA controllers in native SATA mode the disks get 
scribbled on horribly.  With the SATA controllers in legacy mode FreeBSD 
detects them as ATA/33 but they seem to hold up a bit better.  Under heavy 
I/O load they will stay up for about a week before panicing, although once 
they do panic they double panic and hang solid.  I thought I was going dodge 
swapping platforms by installing 3ware 9550 RAID controllers but I all I 
managed to do was reduce the panic rate to about once per month per box, so 
there are gremlins that affect more than the onboard SATA I'm afraid.  I've 
attempted to work with at least one FreeBSD developer on a fix, but the 
inability of the boxes to complete a panic, let alone write a crash dump has 
made troubleshooting them incredibly painful.  (Sometimes they don't even 
manage to get the entire first line of the panic string on the console)  The 
fact that they will stand up for days under heavy I/O load doesn't help 
either.  Intermittant problems suck.

-- 
Thanks,

Josh Paetzel

PGP: 8A48 EF36 5E9F 4EDA 5A8C 11B4 26F9 01F1 27AF AECB