odd timer bug, similar to VIA 686a symptoms

odd timer bug, similar to VIA 686a symptoms

[Neale Banks]

Greetings,

Using Vojtech's arch/i386/kernel/time.c patch ported to 2.2[1], I recently
got the "via686a" bug check triggered after the machine suspended itself
due low battery, including the following lines grepped and snipped out of
kern.log:

May 28 11:19:49 gull kernel: timer.c: VIA bug check triggered. Value read 65500 [0xffdc], re-read 65486 [0xffce]
May 28 11:19:50 gull kernel: timer.c: VIA bug check triggered. Value read 65500 [0xffdc], re-read 65486 [0xffce]
May 28 11:19:51 gull kernel: timer.c: VIA bug check triggered. Value read 65499 [0xffdb], re-read 65484 [0xffcc]
May 28 11:19:52 gull kernel: timer.c: VIA bug check triggered. Value read 65498 [0xffda], re-read 65484 [0xffcc]
May 28 11:19:53 gull kernel: timer.c: VIA bug check triggered. Value read 65500 [0xffdc], re-read 65486 [0xffce]
May 28 11:19:54 gull kernel: timer.c: VIA bug check triggered. Value read 65500 [0xffdc], re-read 65485 [0xffcd]
May 28 11:19:55 gull kernel: timer.c: VIA bug check triggered. Value read 65500 [0xffdc], re-read 65486 [0xffce]
May 28 11:19:56 gull kernel: timer.c: VIA bug check triggered. Value read 65500 [0xffdc], re-read 65486 [0xffce]
May 28 11:19:57 gull kernel: timer.c: VIA bug check triggered. Value read 65499 [0xffdb], re-read 65484 [0xffcc]
May 28 11:19:58 gull kernel: timer.c: VIA bug check triggered. Value read 65497 [0xffd9], re-read 65483 [0xffcb]

Anyone got any good theories what's going on here, given that this is a
~1995 vintage laptop with a Pentium-120 (which I'm assured doesn't have a
VIA 686a ;-)?

Thanks,
Neale.

[1] OK, I stuffed part of the porting of Vojtech's patch (revised version
to come Real Soon Now) but that shouldn't affect the validity of the
greppage from kern.log

Re: odd timer bug, similar to VIA 686a symptoms

[Vojtech Pavlik]

On Wed, May 29, 2002 at 01:46:27PM +0100, Alan Cox wrote:
> On Wed, 2002-05-29 at 10:25, Neale Banks wrote:
> >> May 28 11:19:54 gull kernel: timer.c: VIA bug check triggered. Value read 65500 [0xffdc], re-read 65485 [0xffcd]
> > May 28 11:19:55 gull kernel: timer.c: VIA bug check triggered. Value read 65500 [0xffdc], re-read 65486 [0xffce]
> > May 28 11:19:56 gull kernel: timer.c: VIA bug check triggered. Value read 65500 [0xffdc], re-read 65486 [0xffce]
> > May 28 11:19:57 gull kernel: timer.c: VIA bug check triggered. Value read 65499 [0xffdb], re-read 65484 [0xffcc]
> > May 28 11:19:58 gull kernel: timer.c: VIA bug check triggered. Value read 65497 [0xffd9], re-read 65483 [0xffcb]
> > 
> > Anyone got any good theories what's going on here, given that this is a
> > ~1995 vintage laptop with a Pentium-120 (which I'm assured doesn't have a
> > VIA 686a ;-)?
> 
> Neptune chipsets at least had latching bugs on timer reads. What chipset
> is the laptop ?

This is unlikely to be the latching bug - note the values are near to
65535 - that means the timer is reprogrammed to count from 0xffff down
instead from LATCH. That is because of the suspend I presume. What's
weird is that the VIA fix doesn't program it to the correct value, or
perhaps is that missing from the patch?

-- 
Vojtech Pavlik
SuSE Labs

Re: odd timer bug, similar to VIA 686a symptoms

[Alan Cox]

On Wed, 2002-05-29 at 10:25, Neale Banks wrote:
>> May 28 11:19:54 gull kernel: timer.c: VIA bug check triggered. Value read 65500 [0xffdc], re-read 65485 [0xffcd]
> May 28 11:19:55 gull kernel: timer.c: VIA bug check triggered. Value read 65500 [0xffdc], re-read 65486 [0xffce]
> May 28 11:19:56 gull kernel: timer.c: VIA bug check triggered. Value read 65500 [0xffdc], re-read 65486 [0xffce]
> May 28 11:19:57 gull kernel: timer.c: VIA bug check triggered. Value read 65499 [0xffdb], re-read 65484 [0xffcc]
> May 28 11:19:58 gull kernel: timer.c: VIA bug check triggered. Value read 65497 [0xffd9], re-read 65483 [0xffcb]
> 
> Anyone got any good theories what's going on here, given that this is a
> ~1995 vintage laptop with a Pentium-120 (which I'm assured doesn't have a
> VIA 686a ;-)?

Neptune chipsets at least had latching bugs on timer reads. What chipset
is the laptop ?

Re: odd timer bug, similar to VIA 686a symptoms

[Vojtech Pavlik]

On Wed, May 29, 2002 at 11:40:40PM +1000, Neale Banks wrote:
> On Wed, 29 May 2002, Vojtech Pavlik wrote:
> 
> [...]
> > On Wed, May 29, 2002 at 01:46:27PM +0100, Alan Cox wrote:
> [...]
> > > Neptune chipsets at least had latching bugs on timer reads. What chipset
> > > is the laptop ?
> > 
> > This is unlikely to be the latching bug - note the values are near to
> > 65535 - that means the timer is reprogrammed to count from 0xffff down
> > instead from LATCH. That is because of the suspend I presume. What's
> > weird is that the VIA fix doesn't program it to the correct value, or
> > perhaps is that missing from the patch?
> 
> Yes, my version of your patch includes an option to disable the via686a
> fix (and this was in effect at the time - I'm still cringing in fear from
> nasty FS corruption that ensued after a "probable hardware bug: restoring
> chip configuration" message last October :-( - yes it may be unlikely
> that it's related, but I don't yet have any other suspects.  FWIW, the
> machine also locked up then).

It shouldn't be able to do anything like that ...

> Any suggestions smarter than backing up everything "important" and running
> the battery down with the VIA fix enabled?

Nope. ;)

-- 
Vojtech Pavlik
SuSE Labs

Re: odd timer bug, similar to VIA 686a symptoms

[Neale Banks]

On 29 May 2002, Alan Cox wrote:

> On Wed, 2002-05-29 at 10:25, Neale Banks wrote:
> >> May 28 11:19:54 gull kernel: timer.c: VIA bug check triggered. Value read 65500 [0xffdc], re-read 65485 [0xffcd]
> > May 28 11:19:55 gull kernel: timer.c: VIA bug check triggered. Value read 65500 [0xffdc], re-read 65486 [0xffce]
> > May 28 11:19:56 gull kernel: timer.c: VIA bug check triggered. Value read 65500 [0xffdc], re-read 65486 [0xffce]
> > May 28 11:19:57 gull kernel: timer.c: VIA bug check triggered. Value read 65499 [0xffdb], re-read 65484 [0xffcc]
> > May 28 11:19:58 gull kernel: timer.c: VIA bug check triggered. Value read 65497 [0xffd9], re-read 65483 [0xffcb]
> > 
> > Anyone got any good theories what's going on here, given that this is a
> > ~1995 vintage laptop with a Pentium-120 (which I'm assured doesn't have a
> > VIA 686a ;-)?
> 
> Neptune chipsets at least had latching bugs on timer reads. What chipset
> is the laptop ?

Does it help in that it's got a CMD640?  If not, what am I looking for?

Thanks,
Neale.

Re: odd timer bug, similar to VIA 686a symptoms

[Neale Banks]

On Wed, 29 May 2002, Vojtech Pavlik wrote:

[...]
> On Wed, May 29, 2002 at 01:46:27PM +0100, Alan Cox wrote:
[...]
> > Neptune chipsets at least had latching bugs on timer reads. What chipset
> > is the laptop ?
> 
> This is unlikely to be the latching bug - note the values are near to
> 65535 - that means the timer is reprogrammed to count from 0xffff down
> instead from LATCH. That is because of the suspend I presume. What's
> weird is that the VIA fix doesn't program it to the correct value, or
> perhaps is that missing from the patch?

Yes, my version of your patch includes an option to disable the via686a
fix (and this was in effect at the time - I'm still cringing in fear from
nasty FS corruption that ensued after a "probable hardware bug: restoring
chip configuration" message last October :-( - yes it may be unlikely
that it's related, but I don't yet have any other suspects.  FWIW, the
machine also locked up then).

Any suggestions smarter than backing up everything "important" and running
the battery down with the VIA fix enabled?

Thanks,
Neale.

Re: odd timer bug, similar to VIA 686a symptoms

[Neale Banks]

On 29 May 2002, Alan Cox wrote:

> On Wed, 2002-05-29 at 14:18, Neale Banks wrote:
> > Does it help in that it's got a CMD640?  If not, what am I looking for?
> 
> Thats the IDE controller. Look at lspci and the bridges and see what
> they are. Those are generally the core chipset of the PC

Straight from the metaphorical horses's virtual mouth:

gull:~# lspci 
00:00.0 Host bridge: Acer Laboratories Inc. [ALi] M1451 (rev ad)
00:02.0 Non-VGA unclassified device: Acer Laboratories Inc. [ALi] M1449 (rev b2)
00:06.0 VGA compatible controller: Chips and Technologies F65545
00:07.0 IDE interface: CMD Technology Inc PCI0640 (rev 02)

Is the "[ALi] M1451" and "[ALi] M1449" are what we are looking for here?

Thanks,
Neale.

Re: odd timer bug, similar to VIA 686a symptoms

[Alan Cox]

On Wed, 2002-05-29 at 14:18, Neale Banks wrote:
> Does it help in that it's got a CMD640?  If not, what am I looking for?

Thats the IDE controller. Look at lspci and the bridges and see what
they are. Those are generally the core chipset of the PC

Re: odd timer bug, similar to VIA 686a symptoms

[Alan Cox]

On Wed, 2002-05-29 at 14:47, Neale Banks wrote:
> Straight from the metaphorical horses's virtual mouth:
> 
> gull:~# lspci 
> 00:00.0 Host bridge: Acer Laboratories Inc. [ALi] M1451 (rev ad)
> 00:02.0 Non-VGA unclassified device: Acer Laboratories Inc. [ALi] M1449 (rev b2)
> 00:06.0 VGA compatible controller: Chips and Technologies F65545
> 00:07.0 IDE interface: CMD Technology Inc PCI0640 (rev 02)
> 
> Is the "[ALi] M1451" and "[ALi] M1449" are what we are looking for here?

Yes - thanks.