Multiple (ICH3) IDE-controllers in a system

Multiple (ICH3) IDE-controllers in a system

[Shai Fultheim]

Hi,

I am working on an IO-robust custom-made PC server, which runs the Linux
kernel.  The machine is working very nicely with 2.4 kernels, but we
recently found out that in 2.6 it couldn't see all IDE controllers.

In Linux 2.6.x pcibios_fixups table (in arch/i386/pci/fixup.c), use the
pci_fixup_ide_trash() quirk for Intel's ICH3 (my case specifically
8086:248b).  That quirk wasn't in use for ICH3 by the 2.4.x kernels.  The
result of the change is that the system, which has multiple ICH3's can't use
any of the IDE controllers beside the one on the first ICH3 (again, that
worked in 2.4 kernels).

Is there a real reason to call that quirk?  If there is not, why are you
calling that quirk?  If there is, can I suggest that the quirk will handle
only the first ICH3 (i.e. add check in the quirk that it is called for the
first time only).  This is better than "loosing" all these IDE controllers
in the case their BARs set right.


Blow references for bare 2.6.5 kernel:

Code walkup starts with pci_scan_single_device() at
http://lxr.linux.no/source/drivers/pci/probe.c?v=2.6.5#L590 which calls to
pci_fixup_device() (line 601) which in turn use the quirks table at
http://lxr.linux.no/source/arch/i386/pci/fixup.c?v=2.6.5#L190 (my chipset is
in line 248).  The quirks table for ICH3 use pci_fixup_ide_trash()at
http://lxr.linux.no/source/arch/i386/pci/fixup.c?v=2.6.5#L90 - this reset
the BARs for the device.  Resetting the all (4) BARs of (ICH3) IDE
controllers will cause them to use the defaults BARs (0x170, 0x1f0) in
ide_hwif_configure() at
http://lxr.linux.no/source/drivers/ide/setup-pci.c?v=2.6.5#L420.  This will
fail with any subsequent (ICH3) IDE controllers (two devices can't use the
same ports).  

I'm not sure if ICH3 needs to reset its BARs at all, but if it is, I suggest
making sure pci_fixup_ide_trash() reset BARs only for first time being
called.  In that way subsequent IDE controllers will use the BIOS BARs
(which we set fine, and worked GREAT at 2.4.x), as said above, this is
better than "loosing" all the other IDE controllers in the case their BARs
set right.

Thanks in advance,
Shai.


 
_________________________
Shai Fultheim
FT Consulting
 
Mobile: +1 (408) 480-1612
Fax:    +1 (501) 647-4113
E-Mail: shai@ftcon.com
Web:    www.ftcon.com

Re: Multiple (ICH3) IDE-controllers in a system

[Bartlomiej Zolnierkiewicz]

Hi Vojtech,

Do I correctly assume that these fixups for Intel chipsets are from you?

http://linus.bkbits.net:8080/linux-2.5/cset@3cfbacdfzHvfqp0Sa45QXt9pNn3LNg?nav=index.html|src/|src/arch|src/arch/i386|src/arch/i386/pci|related/arch/i386/pci/fixup.c
http://linus.bkbits.net:8080/linux-2.5/cset@3cfcec0fOJreGFyCWkPeT7EWiydYFw?nav=index.html|src/|src/arch|src/arch/i386|src/arch/i386/pci|related/arch/i386/pci/fixup.c

Care to explain why 'trash' fixup is needed in 2.6 but not in 2.4?

Cheers,
Bartlomiej

On Wednesday 05 of May 2004 00:13, Shai Fultheim wrote:
> Hi,
>
> I am working on an IO-robust custom-made PC server, which runs the Linux
> kernel.  The machine is working very nicely with 2.4 kernels, but we
> recently found out that in 2.6 it couldn't see all IDE controllers.
>
> In Linux 2.6.x pcibios_fixups table (in arch/i386/pci/fixup.c), use the
> pci_fixup_ide_trash() quirk for Intel's ICH3 (my case specifically
> 8086:248b).  That quirk wasn't in use for ICH3 by the 2.4.x kernels.  The
> result of the change is that the system, which has multiple ICH3's can't
> use any of the IDE controllers beside the one on the first ICH3 (again,
> that worked in 2.4 kernels).
>
> Is there a real reason to call that quirk?  If there is not, why are you
> calling that quirk?  If there is, can I suggest that the quirk will handle
> only the first ICH3 (i.e. add check in the quirk that it is called for the
> first time only).  This is better than "loosing" all these IDE controllers
> in the case their BARs set right.
>
>
> Blow references for bare 2.6.5 kernel:
>
> Code walkup starts with pci_scan_single_device() at
> http://lxr.linux.no/source/drivers/pci/probe.c?v=2.6.5#L590 which calls to
> pci_fixup_device() (line 601) which in turn use the quirks table at
> http://lxr.linux.no/source/arch/i386/pci/fixup.c?v=2.6.5#L190 (my chipset
> is in line 248).  The quirks table for ICH3 use pci_fixup_ide_trash()at
> http://lxr.linux.no/source/arch/i386/pci/fixup.c?v=2.6.5#L90 - this reset
> the BARs for the device.  Resetting the all (4) BARs of (ICH3) IDE
> controllers will cause them to use the defaults BARs (0x170, 0x1f0) in
> ide_hwif_configure() at
> http://lxr.linux.no/source/drivers/ide/setup-pci.c?v=2.6.5#L420.  This will
> fail with any subsequent (ICH3) IDE controllers (two devices can't use the
> same ports). 
>
> I'm not sure if ICH3 needs to reset its BARs at all, but if it is, I
> suggest making sure pci_fixup_ide_trash() reset BARs only for first time
> being called.  In that way subsequent IDE controllers will use the BIOS
> BARs (which we set fine, and worked GREAT at 2.4.x), as said above, this is
> better than "loosing" all the other IDE controllers in the case their BARs
> set right.
>
> Thanks in advance,
> Shai.

Re: Multiple (ICH3) IDE-controllers in a system

[Vojtech Pavlik]

On Wed, May 05, 2004 at 05:16:43PM +0200, Bartlomiej Zolnierkiewicz wrote:
> 
> Hi Vojtech,
> 
> Do I correctly assume that these fixups for Intel chipsets are from you?

Yes.

> http://linus.bkbits.net:8080/linux-2.5/cset@3cfbacdfzHvfqp0Sa45QXt9pNn3LNg?nav=index.html|src/|src/arch|src/arch/i386|src/arch/i386/pci|related/arch/i386/pci/fixup.c
> http://linus.bkbits.net:8080/linux-2.5/cset@3cfcec0fOJreGFyCWkPeT7EWiydYFw?nav=index.html|src/|src/arch|src/arch/i386|src/arch/i386/pci|related/arch/i386/pci/fixup.c
> 
> Care to explain why 'trash' fixup is needed in 2.6 but not in 2.4?

Because 2.4 was never used on the affected machines, where this fixup
was needed - those machines sere putting nonsense into the BARs. I don't
recall exactly what model they were, though I remember they were one of
the first machines with ICH MMIO support.

-- 
Vojtech Pavlik
SuSE Labs, SuSE CR

RE: Multiple (ICH3) IDE-controllers in a system

[Shai Fultheim]

Ok.
I would suggest one of the followings:
1. If we can't identify those machine, I would recommend to drop that patch,
since probably the BIOS is taking care of it nowadays.
2. If we believe we can't do (1) above, lets have it rest only the first
controller it is being called for.  This will make any of the other
controllers usable if their ports are set right.

Any comments?
 

--Shai


-----Original Message-----
From: linux-ide-owner@vger.kernel.org
[mailto:linux-ide-owner@vger.kernel.org] On Behalf Of Vojtech Pavlik
Sent: Wednesday, May 05, 2004 23:46
To: Bartlomiej Zolnierkiewicz
Cc: shai@ftcon.com; linux-ide@vger.kernel.org; linux-kernel@vger.kernel.org
Subject: Re: Multiple (ICH3) IDE-controllers in a system

On Wed, May 05, 2004 at 05:16:43PM +0200, Bartlomiej Zolnierkiewicz wrote:
> 
> Hi Vojtech,
> 
> Do I correctly assume that these fixups for Intel chipsets are from you?

Yes.

>
http://linus.bkbits.net:8080/linux-2.5/cset@3cfbacdfzHvfqp0Sa45QXt9pNn3LNg?n
av=index.html|src/|src/arch|src/arch/i386|src/arch/i386/pci|related/arch/i38
6/pci/fixup.c
>
http://linus.bkbits.net:8080/linux-2.5/cset@3cfcec0fOJreGFyCWkPeT7EWiydYFw?n
av=index.html|src/|src/arch|src/arch/i386|src/arch/i386/pci|related/arch/i38
6/pci/fixup.c
> 
> Care to explain why 'trash' fixup is needed in 2.6 but not in 2.4?

Because 2.4 was never used on the affected machines, where this fixup
was needed - those machines sere putting nonsense into the BARs. I don't
recall exactly what model they were, though I remember they were one of
the first machines with ICH MMIO support.

-- 
Vojtech Pavlik
SuSE Labs, SuSE CR
-
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: Multiple (ICH3) IDE-controllers in a system

[Shai Fultheim]

[-- Attachment #1: Type: text/plain, Size: 3049 bytes --]

Vojtech and Bartlomiej,

Attached patch that fix the problem according to (2) below.  Let me know if
you can apply it.

Thanks. 


--- linux-2.6.5-mm6.orig/arch/i386/pci/fixup.c  2004-04-29
05:23:35.000000000 -0700
+++ linux-2.6.5-mm6/arch/i386/pci/fixup.c       2004-05-07
01:45:31.000000000 -0700
@@ -92,6 +92,15 @@
        int i;
 
        /*
+        * Runs the fixup only for the first IDE controller
+        * (Shai Fultheim - shai@ftcon.com)
+        */
+       static int called = 0;
+       if (called)
+               return;
+       called = 1;
+
+       /*
         * There exist PCI IDE controllers which have utter garbage
         * in first four base registers. Ignore that.
         */



--Shai


-----Original Message-----
From: linux-ide-owner@vger.kernel.org
[mailto:linux-ide-owner@vger.kernel.org] On Behalf Of Shai Fultheim
Sent: Thursday, May 06, 2004 12:19
To: 'Vojtech Pavlik'; 'Bartlomiej Zolnierkiewicz'
Cc: linux-ide@vger.kernel.org; linux-kernel@vger.kernel.org
Subject: RE: Multiple (ICH3) IDE-controllers in a system

Ok.
I would suggest one of the followings:
1. If we can't identify those machine, I would recommend to drop that patch,
since probably the BIOS is taking care of it nowadays.
2. If we believe we can't do (1) above, lets have it rest only the first
controller it is being called for.  This will make any of the other
controllers usable if their ports are set right.

Any comments?
 

--Shai


-----Original Message-----
From: linux-ide-owner@vger.kernel.org
[mailto:linux-ide-owner@vger.kernel.org] On Behalf Of Vojtech Pavlik
Sent: Wednesday, May 05, 2004 23:46
To: Bartlomiej Zolnierkiewicz
Cc: shai@ftcon.com; linux-ide@vger.kernel.org; linux-kernel@vger.kernel.org
Subject: Re: Multiple (ICH3) IDE-controllers in a system

On Wed, May 05, 2004 at 05:16:43PM +0200, Bartlomiej Zolnierkiewicz wrote:
> 
> Hi Vojtech,
> 
> Do I correctly assume that these fixups for Intel chipsets are from you?

Yes.

>
http://linus.bkbits.net:8080/linux-2.5/cset@3cfbacdfzHvfqp0Sa45QXt9pNn3LNg?n
av=index.html|src/|src/arch|src/arch/i386|src/arch/i386/pci|related/arch/i38
6/pci/fixup.c
>
http://linus.bkbits.net:8080/linux-2.5/cset@3cfcec0fOJreGFyCWkPeT7EWiydYFw?n
av=index.html|src/|src/arch|src/arch/i386|src/arch/i386/pci|related/arch/i38
6/pci/fixup.c
> 
> Care to explain why 'trash' fixup is needed in 2.6 but not in 2.4?

Because 2.4 was never used on the affected machines, where this fixup
was needed - those machines sere putting nonsense into the BARs. I don't
recall exactly what model they were, though I remember they were one of
the first machines with ICH MMIO support.

-- 
Vojtech Pavlik
SuSE Labs, SuSE CR
-
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

-
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[-- Attachment #2: pci-ide-fixup.patch --]
[-- Type: application/octet-stream, Size: 499 bytes --]

--- linux-2.6.5-mm6.orig/arch/i386/pci/fixup.c	2004-04-29 05:23:35.000000000 -0700
+++ linux-2.6.5-mm6/arch/i386/pci/fixup.c	2004-05-07 01:45:31.000000000 -0700
@@ -92,6 +92,15 @@
 	int i;
 
 	/*
+	 * Runs the fixup only for the first IDE controller
+	 * (Shai Fultheim - shai@ftcon.com)
+	 */
+	static int called = 0;
+	if (called)
+		return;
+	called = 1;
+
+	/*
 	 * There exist PCI IDE controllers which have utter garbage
 	 * in first four base registers. Ignore that.
 	 */