Michael Schmitt wrote:
Hi Pawel,
Am Montag, den 30.06.2008, 17:07 -0700 schrieb Pawel Wojcik:
Michael,
Unfortunately, you are out of luck here. The CR 6595150 refers to the
controller that has pciclass Mass Storage/Other (018000), while your
controller has pciclass Mass Storage/RAID (010400) and therefore is
not recognized as the pci-ide controller that may use ata driver.
The code in pci_boot.c file (function is_pciide ), considers a
controller to be pci-ide controller (using pci-ide/ata drivers) only
if it has pciclass Mass Storage/IDE (01018x), Mass Storage/SATA
(01060x) and Mass Storage/Other (018000).
The code change to accommodate pci-ide "RAID" controllers is very
simple but it would give users a way to destroy their RAID data (if it
exists) by attaching controller and potentially RAID devices when
Solaris could not distinguish between RAID and non-RAID disks.
Please file new bug for Sil3512 controller (category: driver,
subcategory ata-x86) and attach your prtconf output.
I will update Suggested Fix field of this bug - but it would be to up
to someone in Open Solaris to re-build the drivers and test it.
<Homer> Doh! </Homer> :)
I think I understand the problem... I'd like to know how the linux
driver handles this, most likely it just ignores the "raid" (which I
think is absolutely fine).
As far as I know there is NO driver who uses the somewhat limited raid
functionality for a different operating system than Windows. So even IF
such an windows array is poped into a linux / solaris / bsd box on that
controller (dual boot maybe) a sensible admin would not think he could
access the array with a different OS than Windows, would he? After all,
this controller is NOT a raid controller, it is a cheap controller which
may have a few (if any) raid abilities in hardware. That said I see the
possible windows array (with the windows driver) just as a complete
software raid solution which I think windows can do out of the box with
EVERY controller. Who would think he could acces the array with a
different OS?
Anyhow, at worst, I think the driver could just check the data on the
harddisks and if some known "raid block" is found, inform the user, log
it, refuse to attach, start whining... but I would just ignore it (like
the linux driver most likely does) because... sorry... but if the admin
sees there are two disks and does not start thinking why the hell it is
not just one (like with a real raid array)....
Anyhow, I'll file a bug as you suggested.
greetings
Michael
The issue is not really access to foreign OS data on such software RAID
volumes, but rather providing RAID drivers matching the Option Rom BIOS
that allows booting from RAID volumes. That is how Linux is supporting
Intel RAID.
We will have the solution like you have described (Solaris recognizing
software RAID volumes and restricting access to them while allowing the
use of on-RAID volumes), but only for SATA controllers under SATA
framework. This solution may be extended to pci-ide/ata controllers,
but I have not scoped it yet (perhaps I will leave it for OpenSolaris
folks to provide hooks in pci-ide/ata drivers...).
-Pawel
|
_______________________________________________
driver-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/driver-discuss
