IDE corruption datapoint

IDE corruption datapoint

[David S. Miller]

FWIW, just like my alim5x3 based Sparcs, I have to boot with ide=nodma
on my VAIO laptop which uses PIIX chipsets, otherwise I get data
corruptions.

On the VAIO it shows up as corrupted kernel builds.  The build
is successful but the images won't boot properly, rebuilding
the same kernel tree with ide=nodma results in a working image.
This shows up with both 2.4.21 and 2.4.22-current.

I don't know what the situation is wrt. 2.4.20 (ie. before the IDE
huge merge from -ac), but I can check that out.

With 2.6.0-test1 both cases (alim5x3 based Sparcs and the VAIO)
seem perfectly fine with DMA enabled.

This suggests, at least to me, that it's something generic in
the IDE layer of 2.4.x vanilla vs. 2.6.0-test1 that's causing
the problems, rather than a chipset specific issue.

Just FYI...

Re: IDE corruption datapoint

[Alan Cox]

> FWIW, just like my alim5x3 based Sparcs, I have to boot with ide=nodma
> on my VAIO laptop which uses PIIX chipsets, otherwise I get data
> corruptions.

Ok thats I think a first ever and only report for the VAIO. Are you using
ACPI and what compiler are you using ? Do you have APIC support enabled ?
Does it occur with a vanilla Red Hat 2.4.20- kernel ?

> On the VAIO it shows up as corrupted kernel builds.  The build
> is successful but the images won't boot properly, rebuilding
> the same kernel tree with ide=nodma results in a working image.
> This shows up with both 2.4.21 and 2.4.22-current.

If you cp lots of data around then compare what is the error pattern,
thats basically essential info to debug this as is the drive data,
mode selected and so on

> This suggests, at least to me, that it's something generic in
> the IDE layer of 2.4.x vanilla vs. 2.6.0-test1 that's causing
> the problems, rather than a chipset specific issue.

That would be unlikely since someone found the ali problem with
sparc/alpha but just never bothered to send a patch to fix it.

Alan

Re: IDE corruption datapoint

[David S. Miller]

On Wed, 16 Jul 2003 08:47:48 -0400 (EDT)
Alan Cox <alan@redhat.com> wrote:

> > FWIW, just like my alim5x3 based Sparcs, I have to boot with ide=nodma
> > on my VAIO laptop which uses PIIX chipsets, otherwise I get data
> > corruptions.
> 
> Ok thats I think a first ever and only report for the VAIO. Are you using
> ACPI and what compiler are you using ? Do you have APIC support enabled ?
> Does it occur with a vanilla Red Hat 2.4.20- kernel ?

No ACPI, stock compiler from RH8.0 with all updates installed:

gcc (GCC) 3.2 20020903 (Red Hat Linux 8.0 3.2-7)

CONFIG_X86_GOOD_APIC is enabled, but there is no IO-APIC in this
VAIO and /proc/interrupts shows all XT-PIC.

Does not occur with vanilla RH 2.4.20- kernel.

WRT. the VAIO, I originally suspected some of the DMI table changes
that went into 2.4.21, but I reverted that (and nearly all of the
other i386 specific 2.4.21 changes) to no avail.

> If you cp lots of data around then compare what is the error pattern,
> thats basically essential info to debug this as is the drive data,
> mode selected and so on

I'll do this at next opportunity.  It'll take a while since it
requires a whole kernel build in order to trigger the corruption
in the final linked image :(

I'll try using the disk bashing tools from feral.com to see if
that triggers anything easily.

> > This suggests, at least to me, that it's something generic in
> > the IDE layer of 2.4.x vanilla vs. 2.6.0-test1 that's causing
> > the problems, rather than a chipset specific issue.
> 
> That would be unlikely since someone found the ali problem with
> sparc/alpha but just never bothered to send a patch to fix it.

Interesting.  I've never seen any evidence from Ben or anyone else who
saw the problem on Sparc say much other than "2.6.0-test1 doesn't have
the problem, 2.4.x does"

Can you show the patch in question?  I guess it's in 2.6.0-test1
already?

Re: IDE corruption datapoint

[Alan Cox]

> CONFIG_X86_GOOD_APIC is enabled, but there is no IO-APIC in this
> VAIO and /proc/interrupts shows all XT-PIC.

It matters if you have APIC/IO-APIC support enabled. In marcelo's tree
disable_irq is broken in APIC mode for non SMP.

> WRT. the VAIO, I originally suspected some of the DMI table changes
> that went into 2.4.21, but I reverted that (and nearly all of the
> other i386 specific 2.4.21 changes) to no avail.

They really wouldn't relate to the IDE code at all. IDE or block layer
stuff is by far the most likely.

> Interesting.  I've never seen any evidence from Ben or anyone else who
> saw the problem on Sparc say much other than "2.6.0-test1 doesn't have
> the problem, 2.4.x does"
> 
> Can you show the patch in question?  I guess it's in 2.6.0-test1
> already?

I got mail from two people confirming some "patch" fixed it, but nobody
ever sent me the patch in question. I'd guess its making sure the PIO side
is always set up at boot

Re: IDE corruption datapoint

[David S. Miller]

On Wed, 16 Jul 2003 09:24:07 -0400 (EDT)
Alan Cox <alan@redhat.com> wrote:

> > CONFIG_X86_GOOD_APIC is enabled, but there is no IO-APIC in this
> > VAIO and /proc/interrupts shows all XT-PIC.
> 
> It matters if you have APIC/IO-APIC support enabled. In marcelo's tree
> disable_irq is broken in APIC mode for non SMP.

I'll test with this option enabled then.

> > Can you show the patch in question?  I guess it's in 2.6.0-test1
> > already?
> 
> I got mail from two people confirming some "patch" fixed it, but nobody
> ever sent me the patch in question. I'd guess its making sure the PIO side
> is always set up at boot

That was probably from me, I forwarded you an email from Ivan saying
"this aparently fixes Alpha, I'm asking Ben Collins to test back in
2.4.x" but the latter never actually occurred :-)

Re: IDE corruption datapoint

[Alan Cox]

> That was probably from me, I forwarded you an email from Ivan saying
> "this aparently fixes Alpha, I'm asking Ben Collins to test back in
> 2.4.x" but the latter never actually occurred :-)

I guess by the DaveM rule of bug handling it clearly doesnt matter so I
should just ignore it 8)

Re: IDE corruption datapoint

[David S. Miller]

On Wed, 16 Jul 2003 09:38:12 -0400 (EDT)
Alan Cox <alan@redhat.com> wrote:

> I guess by the DaveM rule of bug handling it clearly doesnt matter so I
> should just ignore it 8)

If you're busy with something else, this is _EXACTLY_ what you should
do, delete and defer.

The lists hold all the information, and someone else more inclined can
either stick it into some bug database or try to follow up on it.

Re: IDE corruption datapoint

[Ben Collins]

[-- Attachment #1: Type: text/plain, Size: 488 bytes --]

On Wed, Jul 16, 2003 at 04:20:20AM -0700, David S. Miller wrote:
> 
> FWIW, just like my alim5x3 based Sparcs, I have to boot with ide=nodma
> on my VAIO laptop which uses PIIX chipsets, otherwise I get data
> corruptions.

Let's not forget this patch, which has allowed me to run 2.4.21 and
2.4.22-pre's with DMA enabled on my sb100 (alim5x3) without corruption.


-- 
Debian     - http://www.debian.org/
Linux 1394 - http://www.linux1394.org/
Subversion - http://subversion.tigris.org/

[-- Attachment #2: 3-ali-ide-dma-fix.diff --]
[-- Type: text/plain, Size: 1876 bytes --]

On Tue, May 27, 2003 at 02:31:52PM +0200, Willy Tarreau wrote:
> Sorry, I pasted the .config that I used just after, and which allowed me to
> boot. Later I set CONFIG_BLK_DEV_ALI15X3 again and CONFIG_BLK_DEV_IDEDMA_PCI,
> but I left CONFIG_IDEDMA_PCI_AUTO disabled. I now can boot and enable DMA
> later. That's weird, but it works.

Perhaps not that weird. From my experience, ALi DMA is sensitive to
some of "PIO timings". That is, if SRM hasn't initialized the chipset
properly (on Nautilus it has, BTW), DMA won't work. When you boot with
DMA disabled, driver has to set right PIO mode, so you can safely
enable DMA later.

Can you (and Jason) try this patch with CONFIG_IDEDMA_PCI_AUTO=y?

Ivan.

--- linux/drivers/ide/pci/alim15x3.c.orig	Tue Apr 22 19:17:22 2003
+++ linux/drivers/ide/pci/alim15x3.c	Tue May 27 17:42:17 2003
@@ -525,10 +525,14 @@ static int ali15x3_config_drive_for_dma(
 
 	drive->init_speed = 0;
 
+	/* Set reasonable PIO timings first - some of them are needed
+	   for DMA as well. */
+	hwif->tuneproc(drive, 255);
+
 	if ((id->capability & 1) != 0 && drive->autodma) {
 		/* Consult the list of known "bad" drives */
 		if (hwif->ide_dma_bad_drive(drive))
-			goto ata_pio;
+			goto no_dma_set;
 		if ((id->field_valid & 4) && (m5229_revision >= 0xC2)) {
 			if (id->dma_ultra & hwif->ultra_mask) {
 				/* Force if Capable UltraDMA */
@@ -550,11 +554,9 @@ try_dma_modes:
 			if (!config_chipset_for_dma(drive))
 				goto no_dma_set;
 		} else {
-			goto ata_pio;
+			goto no_dma_set;
 		}
 	} else {
-ata_pio:
-		hwif->tuneproc(drive, 255);
 no_dma_set:
 		return hwif->ide_dma_off_quietly(drive);
 	}
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: IDE corruption datapoint

[Alan Cox]

> Let's not forget this patch, which has allowed me to run 2.4.21 and
> 2.4.22-pre's with DMA enabled on my sb100 (alim5x3) without corruption.

That one certainly looks right

Re: IDE corruption datapoint

[David S. Miller]

On Wed, 16 Jul 2003 10:34:43 -0400 (EDT)
Alan Cox <alan@redhat.com> wrote:

> > Let's not forget this patch, which has allowed me to run 2.4.21 and
> > 2.4.22-pre's with DMA enabled on my sb100 (alim5x3) without corruption.
> 
> That one certainly looks right

Someone needs to forward this to Marcelo, if enough context from
this thread is provided to him, he'll apply it. :-)

Re: IDE corruption datapoint

[Ben Collins]

On Wed, Jul 16, 2003 at 11:36:46PM -0700, David S. Miller wrote:
> On Wed, 16 Jul 2003 07:34:11 -0700
> "David S. Miller" <davem@redhat.com> wrote:
> 
> > Someone needs to forward this to Marcelo, if enough context from
> > this thread is provided to him, he'll apply it. :-)
> 
> Is someone taking care of this? :-)

I can send it to Marcelo, but I'm sure he'll wait to see some backing
from atleast Alan before he takes it.

-- 
Debian     - http://www.debian.org/
Linux 1394 - http://www.linux1394.org/
Subversion - http://subversion.tigris.org/

Re: IDE corruption datapoint

[David S. Miller]

On Thu, 17 Jul 2003 12:38:50 -0400
Ben Collins <bcollins@debian.org> wrote:

> On Wed, Jul 16, 2003 at 11:36:46PM -0700, David S. Miller wrote:
> > On Wed, 16 Jul 2003 07:34:11 -0700
> > "David S. Miller" <davem@redhat.com> wrote:
> > 
> > > Someone needs to forward this to Marcelo, if enough context from
> > > this thread is provided to him, he'll apply it. :-)
> > 
> > Is someone taking care of this? :-)
> 
> I can send it to Marcelo, but I'm sure he'll wait to see some backing
> from atleast Alan before he takes it.

This is why I said, above, "if enough context from this thread".
In particular, include Alan's reply where he said "looks about
right" wrt. this patch :-)