Issue with PATA (IDE) RAID on SiI0680 chip

Giampiero Gabbiani giampiero_gabbiani at tin.it
Fri Oct 28 06:40:41 UTC 2005


Hi Richard,

> Hey all,
>  I have a SYBA RAID card which uses the Silicon Image 680 (or sometimes
> referred to as CMD680) medley chipset. I came across dmraid whilst
> searching for a working solution for my fakeraid setup, and noticed that
> the website reports that the Silicon Image Medley chipsets are supported by
> dmraid. The problems is, I've tried several boot disks which include

As for I know, the difference from mdadm (software raid) and dmraid (ataraid) 
is that the former manages the raid paradigm enterely from the kernel code, 
while the second can access to the metadata set by the BIOS of the 
controller.

> support for dmraid (such as the standard gentoo livecd with boot param
> dodmraid, the gen2dmraid boot cd, the ahorn5 boot cd and fedora core 4's
> livecd), and none of these seem so be able to detect my raid array.
>  Running a dmraid -ay on any of the cds yields a "No Software RAID disks",
> however the kernel always detects the card perfectly (the two disks can be
> accessed directly through /dev/hde and /dev/hdg).

I don't know about these distros but, with mandrake/mandriva, I experimented 
similar problems. This was due to the fact that dmraid needs the device 
mapper module in order to work, but, in mandrake, this was compiled as module 
AND was NOT present in the initrd.

If this is your case, there are two alternatives:

- recompile the kernel with the device mapper support set NOT as a module (I 
made so...)

- build an initrd with the device mapper module present

Before doing so try the following :

1) enabling the kernel device-mapper
 
 [root at aphroditi kernel]# modprobe dm-mod
 [root at aphroditi kernel]# tail -f /var/log/messages
 <snip>
 Oct 10 19:32:50 aphroditi kernel: device-mapper: 4.4.0-ioctl (2005-01-12)
 initialised: dm-devel at redhat.com
 <snip>

2) scanning ATARAID devices:
 
 [root at aphroditi kernel]# dmraid -r
 /dev/hdc: sil, "sil_afbgbgcccjcab", mirror, ok, 240119680 sectors, data@ 0
 /dev/hde: sil, "sil_afbgbgcccjcab", mirror, ok, 240119680 sectors, data@ 0
 
 [root at aphroditi kernel]# dmraid -s
 *** Set
 name : sil_afbgbgcccjcab
 size : 240119680
 stride : 0
 type : mirror
 status : ok
 subsets: 0
 devs : 2
 spares : 0

3) enabling the found devices
 
 [root at aphroditi kernel]# dmraid -ay
 
 after this the RAID device is mapped as /dev/mapper/sil_afbgbgcccjcab
 
4) after this, format the partition in not already done, and mount the device.

>  All of this leads me to believe that the Silicon Medley PATA Raid chipset
> is not currently supported by dmraid, and this support is limited to the
> data raid chipset (3112, I think). Also, my motherboard has Nforce3 PATA NV

Just take a look at the code of dmraid, you'll find SilI0680 specific code 
(i.e. it's true, sili0680 it's supported...).

> Raid onboard as well, and dmraid does not work with that fakeraid solution
> either.
>  Does this mean that no PATA solution for fakeraid exists in linux?
>  I appreciate your help and would love to help out with getting PATA raid
> chipsets included in dmraid's support list. Please let me know how I can
> help out, or if a solution already exists, what it is.
>  Kind Regards,
> --
> Richard Powell (rspowell at gmail.com)

**FINAL CONSIDERATIONS**

Even if dmraid could see my SilI0680 controller, finally I went to mdadm 
(software raid) for the following reasons:

1) both ataraid and mdadm  use the CPU in order to manage the controller card, 
that is: there is no CPU usage benefit from using dmraid or mdadm;

2) I couldn't find any daemon, in ataraid, able to perform the necessary 
synchronization work in my RAID1 set.
The only solution found was periodically starting windows and let it do the 
'dirty job'.... 

Hope this could help.
Best Regards.
-- 
Giampiero Gabbiani




More information about the Ataraid-list mailing list