Re: HPT372 bug summary [was: RE: escalation stage 2]

From: Soeren Schmidt <sos_at_spider.deepcore.dk>
Date: Fri, 18 Jul 2003 09:44:18 +0200 (CEST)
It seems Harald Schmalzbauer wrote:
> 
> I also think the RAID configuration is stored on the disks since when I
> create a non-DOS compatible slice (starting at 0 not 63) the RAID
> configuration vanishes.

Yes the Highpoint BIOS uses sector 9 to hold the RAID config, so if you
use that for other data storage you are lost..
(blaim Highpoint sufficiently here).

> Now I assume that there are two different RAID configurations, one stored on
> disk by the controllers BIOS and anotherone which FreeBSD stores elsewhere
> (e.g: with the sil0680 I can well create slices starting at 0).

No, the info is *always* stored on disk, but when you make a FreeBSD
native ATA RAID its put at the end of disk, and the disksize is
reduced so you cannot overwrite it (so does Promise controllers)

> Now when one drive fails both configurations are marked degraded but in a
> different manner (because there is one array named "RAID1_1" and a second
> which is named "FreeBSD")
> And that's why FreeBSD panics until I delete the mirror relationship.

Well Highpoints way of dealing with broken arrays and lost disks are less
than optimal, I've tried to do my best in the ATA driver to fool the HPT
system, but its not perfect...

> Since this is my most important server I can't help you the next weeks. On
> sunday I'll buy a SIL0680 based controller because I did the same test with
> it and it's working.
> Now I'm currently setting up FreeBSD and building a kernel with DDB.

If you are not using 5.1-current the sil680 is dangerous until I get time
to backport some critical fixes...

-Søren
Received on Thu Jul 17 2003 - 22:44:23 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:15 UTC