Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 23 Sep 2004 09:26:21 +0200
From:      =?ISO-8859-1?Q?S=F8ren_Schmidt?= <sos@DeepCore.dk>
To:        Jeremy Chadwick <freebsd@jdc.parodius.com>
Cc:        freebsd-current@freebsd.org
Subject:   Re: Adaptec ICH5R-S Serial ATA RAID
Message-ID:  <41527A9D.2080700@DeepCore.dk>
In-Reply-To: <20040922220330.GA48125@parodius.com>
References:  <200409212256.i8LMu1b7032629@sage.ts.co.nz> <41517193.60604@nulis.lt> <1521.66.11.183.178.1095880306.squirrel@66.11.183.178> <20040922220330.GA48125@parodius.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Jeremy Chadwick wrote:
> As I mentioned, Soren may not have an ICH5R to test with.  If he
> doesn't, I can send him a motherboard (no CPU or RAM though) which can
> help with this department.

I do have ICH5(R) and 6300ESB boards around, but none of them uses the=20
Intel RAID metadata (not even the one thats made by Intel) they all use=20
Adaptec HostRAID which I have partial support for (unreleased yet).

> I think it's "too soon" to get this into RELENG_5, to be honest, unless=

> things happen over the next 7-8 days.  -CURRENT is an entirely differen=
t
> story.
>=20
>>From what I understand, there's been a lot of work recently on getting
> metadata support for different onboard RAID controllers.  Soren could
> comment on this...

Yes there has, but lately my efforts has concentrated on stabilizing=20
things for 5.3. There are a few issues that needs to be looked at and=20
that has higher priority (for me at least).

Now, one of the reason I havn't put more metadata formats in there is=20
that its no longer enough to use the chip PCI ID to find out what=20
metadata to look for, I know of at least 3 different formats used on SiI =

chips.
This needs some thought as we might not want to scan all disks for a=20
dozen different formats.
And if more than one format is found on a disk, which one do we use ?

Another issue is that no vendors except Promise and Highpoint has=20
released docs/info on how those formats are done, so each new format=20
need a fair amount of reverseengineering and testing to be made at least =

semireliable.

On top of that, ata-raid.c needs a serious rewrite for other reasons.=20
The way it plugs into the disks strategy routines is a hack at best. It=20
also needs to be able to take advantage of those chips that can do some=20
part of the RAID stuff (Promise, Marvell, AHCI).

So, there is plenty to do as usual, and I have to balance priorities=20
amongst it all.

Back to the matter at hand, if this patch is going into ata-raid.c it=20
will need a maintainer. Since I dont have HW with this format around  it =

wont get any testing/support from this end and will not be provided in=20
the new ata-raid.c unless someone does the integration when time comes.=20
We also need to have some kind of resolution to the problem with more=20
than one format pr device id.

And no, this is *not* RELENG_5 material, not even close.

--=20

-S=F8ren




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?41527A9D.2080700>