[conspire] (forw) Re: Adaptec Serial ATA RAID 2410SA

Rick Moen rick at linuxmafia.com
Wed Feb 25 14:18:00 PST 2009

May be of interest, if only for the broader point near the end.

----- Forwarded message from Rick Moen <rick at linuxmafia.com> -----

Date: Wed, 25 Feb 2009 14:16:59 -0800
From: Rick Moen <rick at linuxmafia.com>
To: Mark Wilber <mw0 at sonic.net>
Subject: Re: Adaptec Serial ATA RAID 2410SA

Quoting Mark Wilber (mw0 at sonic.net):

>    Hi,
>    I noticed that on your Serial ATA on Linux page
>    ([1]http://linuxmafia.com/faq/Hardware/sata.html) you list Adaptec's
>    2410SA controller card as hardware RAID. This page from adaptec.com
>    indicates that special drivers are needed for use with linux:
>    [2]http://www.adaptec.com/en-US/speed/raid/aac/linux/aacraid-dkms-1_1_
>    5-2400_tgz.htm
>    Perhaps this is needed for support software that notifies the user of
>    a drive failure, but the description leaves me wondering if this isn't
>    a proprietary software driver for some sort of fakeraid. (I'm no
>    expert; just flagging for your consideration.) I didn't see any such
>    kernel module support for, eg., their 2420SA controller, leading me to
>    wonder if that one is true hardware raid . . .
>    Regards,
>    -Mark
> References
>    1. http://linuxmafia.com/faq/Hardware/sata.html
>    2. http://www.adaptec.com/en-US/speed/raid/aac/linux/aacraid-dkms-1_1_5-2400_tgz.htm

Thanks for the note.  I'm largely dependent on feedback from
correspondents to help improve the page, because I've not actually used
practically any of the hardware it describes.  The first couple of years
that page existed, I'd not even seen any SATA gear at all!  (That's why
there's a disclaimer at the page bottom.)

Here are some reasons to think it's hardware RAID:

1.  It has an Intel IOP302/IOP303 I/O processor chip for XOR
calculations, etc.  That's a distinctively hardware-RAID feature.

2.  Various other credible pages describe it and the IOP30x series as
hardware RAID.  

3.  It supports RAID levels 5 and 10.  A tip-off about fakeraid controllers
is talking about built-in support for only levels 0 and 1.  That's
because the ROM-assist booting isn't sophisticated to boot from anything
more complex than a spanned drive set or a mirrored drive set.

Anyway, special drivers are _not_ required.  Linux's native aacraid
drive works fine.

In general, you should not trust what manufacturers say about driver
requirements for Linux.  They will always try to push you towards their
proprietary, binary-only drivers -- because that's their standard model
for all operating systems.  It's the way they think:  They assume that
any OS user is going to want manufacturer-developed drivers coded under
NDA and released as separate downloads in binary-only format.

You might think that such drivers would tend to be better than open
source alternatives developed by outsiders without (in many cases) help,
datasheets, or sample driver code, who in many cases are volunteer
programmers -- but the opposite is in fact true.  Linux mainline kernel
drivers tend to be vastly superior to those from manufacturers.  The
binary-only manufacturer drivers tend to be unstable, brittle, and badly
written.  (The latter surprise has emerged in some cases when
long-secret drivers were suddenly opened up.  It seemed that part of
what they were hiding was incredibly sloppy programming.)

So, it is most often the case that you should do everything feasible to
make the generic open source drivers work, _before_ falling back on the
proprietary, binary-only manufacturer drivers as a last resort.

(There are a variety of special cases, not covered here.)

----- End forwarded message -----

More information about the conspire mailing list