Re: SATA/IDE Dual Mode w/Intel 945 Chipset or HOW TO LIQUIFY a flash IDE chip under 2.6.18

Previous thread: Fix the reproducible oops in scsi by OGAWA Hirofumi on Tuesday, January 9, 2007 - 2:04 pm. (3 messages)

Next thread: RE: [PATCH 4/4] x86_64 ioapic: Improve the heuristics for when check_timer fails. by Lu, Yinghai on Tuesday, January 9, 2007 - 3:15 pm. (2 messages)
From: Jeff V. Merkey
Date: Tuesday, January 9, 2007 - 1:46 pm

I just finished pulling out a melted IDE flash drive out of a Shuttle 
motherboard with the intel 945 chipset which claims to support
SATA and IDE drives concurrently under Linux 2.6.18.

The chip worked for about 30 seconds before liquifying in the chassis.  
I note that the 945 chipset in the shuttle PC had some serious
issues recognizing 2 x SATA devices and a IDE device concurrently.   Are 
there known problems with the Linux drivers
with these newer chipsets.

One other disturbing issue was the IDE flash drive was configured (and 
recognized) as /dev/hda during bootup, but when
it got to the root mountint, even with root=/dev/hda set, it still kept 
thinking the drive was at scsi (ATA) device (08,13)
and kept crashing with VFS cannot find root FS errors.

Jeff
-

From: Jeff V. Merkey
Date: Tuesday, January 9, 2007 - 1:51 pm

Jeff V. Merkey wrote:

root=/dev/hda2 is what was passed to the kernel from grub.


-

From: Auke Kok
Date: Tuesday, January 9, 2007 - 2:39 pm

it sounds like someone switched the BIOS IDE setting from ide-compatible/legacy to AHCI 
or similar, a not uncommon option in the sata controllers on those boards.

None of that would explain the melting of anything of course.

Cheers,

Auke
-

From: Jeff V. Merkey
Date: Tuesday, January 9, 2007 - 3:11 pm

Would be if pin 20 were powered for some reason (which it should NOT 
be).     


-

From: Lennart Sorensen
Date: Tuesday, January 9, 2007 - 3:59 pm

Had the drive ever been used in any other machine?  Had any ide device
ever been used in this machine before?  It really sounds like a hardware
problem, since I can't think of anything software could do to make that
kind of current go through the flash drive.

I remember seeing the controller chip on a 730MB quantum scsi drive
start to glow red many years ago, just before the drive stopped
responding to the system (and I turned off the power).  Hardware does
fail.  It almost never has anything to do with software.

--
Len Sorensen
-

From: Jeff V. Merkey
Date: Tuesday, January 9, 2007 - 4:35 pm

Yes. external cabled CDROM Drive seems to work.


-

From: Jeff Garzik
Date: Tuesday, January 9, 2007 - 4:21 pm

We have two sets of ATA drivers now, and Intel motherboards support 
bazillion annoying IDE modes, so you will need to provide more info than 
this.

Is the motherboard in combined mode?  native mode?  AHCI or RAID mode?
What driver set did you pick?  is drivers/ide built in, modular, or 
disabled?  is drivers/ata built in, modular, or disabled?

The cannot-find-root-FS errors are definitely caused by driver and/or 
initrd misconfiguration.  The melted flash, I dunno, maybe you managed 
to get two drivers fighting over the same hardware.

	Jeff



-

From: Jeff V. Merkey
Date: Tuesday, January 9, 2007 - 4:34 pm

No.  Seems related to the chipset problems.  If I say "root=/dev/hda2" I 
have better not be getting errors claiming device 08:13 could not mount 
as root.  memory corruption?

The melted flash seems power related (like pin 20 was live for some 
reason on a standard IDE).


-

From: Jeff Garzik
Date: Tuesday, January 9, 2007 - 5:25 pm

Combined mode is a technical term.  Judging from your answers, you are 

Judging from your answers, you are not in AHCI mode.

Side note:  You should use AHCI if available.  Emulating a PATA 
interface for SATA devices is error prone [in the silicon].  AHCI is 

If the kernel cannot mount the requested root= disk, it tries the 
default that is encoded into the vmlinuz image at build time, which is 

Probably, otherwise we would have many more reports like this than just 
yours.

	Jeff


-

From: Jeff V. Merkey
Date: Tuesday, January 9, 2007 - 5:13 pm

AHCI It is.   


-

From: Lennart Sorensen
Date: Wednesday, January 10, 2007 - 7:39 am

I would have thought 'sata 3.0 + IDE' sounded a lot like combined mode,

I tried setting my sister's new machine to AHCI mode (Asus P5B with 965
chipset), but I eventually gave up since it also needed windows xp on it
and I can't for the life of me find an AHCI driver for windows that
would install.

--
Len Sorensen
-

From: Jeff Garzik
Date: Wednesday, January 10, 2007 - 7:58 am

Enhanced mode means separate SATA and PATA.

(I recommend avoiding the "IDE" acronym, it is largely meaningless and 

Um, ok?

We're talking about Linux here.  Linux regularly supports hardware 
before Windows does.  This is nothing new.

	Jeff


-

From: Lennart Sorensen
Date: Wednesday, January 10, 2007 - 8:15 am

That is certainly true.  I just found it odd that intel wouldn't have an
ahci driver available.  But then again if ahci is standard I guess they
would expect microsoft to provide the driver instead, which they

--
Len Sorensen
-

From: Prakash Punnoor
Date: Wednesday, January 10, 2007 - 10:29 am

You can install the Intel Matrix driver after "adjusting" the inf file...

=2D-=20
(=B0=3D                 =3D=B0)
//\ Prakash Punnoor /\\
V_/                 \_V
From: Lennart Sorensen
Date: Wednesday, January 10, 2007 - 10:47 am

Hmm, I guess a good question is: Why should I have to edit the inf file?
Is it an issue of them making it only install if your hardware is
already set to ahci mode?  But how am I supposed to boot and install the
driver until I have the driver installed then.  Well I might try that
next time I go there.  How stupid of intel.

--
Len Sorensen
-

From: Jeff V. Merkey
Date: Wednesday, January 10, 2007 - 10:25 am

No doubt part of the Wintel (intel + Microsoft) strategy to perpetually 
break non-windows platforms with new incompatible
hardware like the switch over from the e1000 MT adapters to e1000 GT 
which are not backward compatible with the older chipsets.

I still have not seen the GT adapter work correctly off windows.



-

From: Lennart Sorensen
Date: Wednesday, January 10, 2007 - 10:58 am

But isn't AHCI a new standard intel helped develop?  Why would they want
to make it hard to use intel hardware using a standard interface intel
helped create?  It makes no sense.  Linux doesn't care if the sata is
set to the old PATA compatible interface, or the new AHCI mode.  Windows
simply can't boot in AHCI mode and refuses to install a driver for
AHCI mode when it is not already in AHCI mode.

--
Len Sorensen
-

From: Jeff Garzik
Date: Wednesday, January 10, 2007 - 11:17 am

I presume you mean breaking /windows/ platforms?

As I noted, Linux often supports the hardware from the "big" hardware 
vendors before Windows does.

They use Linux as a "rabbit" to push Microsoft into supporting 
something, with the "Linux supports it already" argument.

	Jeff


-

From: Prakash Punnoor
Date: Wednesday, January 10, 2007 - 10:56 am

Intel wants you to buy hw with ICH8R. ICH8 isn't get the advanced features =
for=20
free....

To get the driver going: Put your hd to the jmicron, install driver, put hd=
=20
back to ich8...


=2D-=20
(=B0=3D                 =3D=B0)
//\ Prakash Punnoor /\\
V_/                 \_V
From: Lennart Sorensen
Date: Wednesday, January 10, 2007 - 11:00 am

But the BIOS has AHCI mode as an option.  I don't want their fake raid,

Hmm, could try that, assuming the jmicron controller doesn't mind.  Of
course the jmicron can also be set to ahci mode (not that I have an ahci
driver for it either under windows).

--
Len Sorensen
-

From: Jeff Garzik
Date: Wednesday, January 10, 2007 - 11:19 am

What advanced features do you claim are missing from ICH8?

The 'R' indicates software RAID, provided by BIOS and a software driver. 
  Which uses the standard AHCI programming interface.  ICH8 provides 
AHCI, just like ICH8R does.

	Jeff



-

From: Prakash Punnoor
Date: Wednesday, January 10, 2007 - 11:33 am

I don't claim anything. Intel does. I know that the chip is basically the=20

Yes, I know...

=2D-=20
(=B0=3D                 =3D=B0)
//\ Prakash Punnoor /\\
V_/                 \_V
Previous thread: Fix the reproducible oops in scsi by OGAWA Hirofumi on Tuesday, January 9, 2007 - 2:04 pm. (3 messages)

Next thread: RE: [PATCH 4/4] x86_64 ioapic: Improve the heuristics for when check_timer fails. by Lu, Yinghai on Tuesday, January 9, 2007 - 3:15 pm. (2 messages)