... bought a Dell Poweredge 600 box with RH 7.3 (2.4.18-4 #1)
pre-installed, and SCSI hard disks. Added a Matsushita DVD-RAM
LF-D311 atapi dvd-ram (that, along with a cd-rom, are the only
IDE devices).
Now, when I do a full backup (as opposed to a differential; I'm
using Microlite's BackupEdge, a Super-Tar), the machine
(sometimes) hangs with the error message:
"Serverworks OSB4 in impossible state.
Disable UDMA or if you are using Seagate then try switching disk
types on this controller. Please report this event to osb4-
[email protected]
OSB4: continuing might cause disk corruption."
RH support advised me to remove "ide0=ata66 ide1=ata66" from the
kernel line in /boot/grub/grub.conf, add "ide-nodma" and reboot.
No joy. Note the following kernel messages at boot-up:
Jul 18 02:31:52 gemach kernel: Kernel command line: ro
root=/dev/sda9 hda=ide-scsi ide=nodma
Jul 18 02:31:52 gemach kernel: ide_setup: hda=ide-scsi
Jul 18 02:31:54 gemach kernel: ide0: BM-DMA at 0x08b0-0x08b7,
BIOS settings: hda:pio, hdb:pio
Jul 18 02:31:55 gemach kernel: ide1: BM-DMA at 0x08b8-0x08bf,
BIOS settings: hdc:DMA, hdd:pio
Jul 18 02:31:55 gemach kernel: hda: MATSHITADVD-RAM LF-D311,
ATAPI CD/DVD-ROM drive
Jul 18 02:31:55 gemach kernel: hdc: GCR-8481B, ATAPI CD/DVD-ROM
drive
Jul 18 02:31:35 gemach rc.sysinit: Setting hard drive parameters
for hdc: succeeded
Jul 18 02:31:59 gemach kernel: hdc: ATAPI 48X CD-ROM drive, 128kB
Cache
Jul 18 02:31:59 gemach kernel: hda: driver not present
Jul 18 02:31:59 gemach kernel: hda: DMA disabled
Jul 18 02:31:59 gemach kernel: hdc: DMA disabled
Jul 18 02:31:59 gemach kernel: hdc: DMA disabled
According to the RH technician,
"Unfortunately this may be beyond a software resolution. On the
machine I last saw this error on (it was not a Dell) the chipset
would not accept commands that disabled DMA properly. By issuing
the normal commands to disable DMA, then checking in various
parts of the /proc filesystem, you would find parts of the
kernel that believe DMA to be disabled, and other parts that
believe DMA enabled- thus an 'impossible state'."
Anyone out there with tips on how to resolve this? Perhaps I can
force the kernel to think that DMA has been disabled?
Thanks in advance, and please cc: me at [email protected].
--
_________________________________________
Nachman Yaakov Ziskind, EA, LLM [email protected]
Attorney and Counselor-at-Law http://yankel.com
Economic Group Pension Services http://egps.com
Actuaries and Employee Benefit Consultants
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
fredag 18. juli 2003, 22:06, skrev Nachman Yaakov Ziskind:
> Anyone out there with tips on how to resolve this? Perhaps I can
> force the kernel to think that DMA has been disabled?
No tips, I'm afraid, but don't even *think* about doing that.
As the technician said, the problem is with the chipset not accepting commands
properly; fortunately, the kernel appears to catch the situation and avoid
(potential) severe damage.
Forcing the kernel to act as if DMA is fully disabled when in fact it isn't
sounds like a very bad idea indeed; the best you could probably hope for is
to have the machine crash without losing any data.
My suggestion is this: As the hardware is obviously broken, and disabling DMA
would cause a horrendous performance drop anyway, you should get a new
chipset. Return the one you have as broken.
If that isn't an option, for whatever reason, you might try switching to a
lower-speed DMA mode using hdparm. Something like "hdparm -Xudma0 /dev/hdx"
might help, if you're lucky.
- - Svein Ove Aas
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2 (GNU/Linux)
iD8DBQE/GG/f9OlFkai3rMARApCiAKCojeoY+nfskcM6EKFco8xktnSfjQCfTq6u
rkZjj7+DtP90zv8cHuTorrA=
=lr9Y
-----END PGP SIGNATURE-----
On Gwe, 2003-07-18 at 23:08, Svein Ove Aas wrote:
> My suggestion is this: As the hardware is obviously broken, and disabling DMA
> would cause a horrendous performance drop anyway, you should get a new
> chipset. Return the one you have as broken.
Update to 2.4.20. That will put IDE disks into MWDMA2 on the Serverworks
OSB4 and avoid the mistrigger with CD-ROM errors. The later serverworks
(CSB5, CSB6) is fine btw but can hit the CD-ROM mistrigger too
Alan Cox wrote (on Fri, Jul 18, 2003 at 11:54:46PM +0100):
> On Gwe, 2003-07-18 at 23:08, Svein Ove Aas wrote:
> > My suggestion is this: As the hardware is obviously broken, and disabling
> > DMA would cause a horrendous performance drop anyway, you should get a new
> > chipset. Return the one you have as broken.
>
> Update to 2.4.20. That will put IDE disks into MWDMA2 on the Serverworks
> OSB4 and avoid the mistrigger with CD-ROM errors. The later serverworks
> (CSB5, CSB6) is fine btw but can hit the CD-ROM mistrigger too
I just did. Happiness is a machine that reboots remotely. Emboldened by my
success, I started a backup. Works like a charm, too. I hope my luck holds up!
How sweet it is.
Thanks to all who responded.
--
_________________________________________
Nachman Yaakov Ziskind, EA, LLM [email protected]
Attorney and Counselor-at-Law http://yankel.com
Economic Group Pension Services http://egps.com
Actuaries and Employee Benefit Consultants