Computer Resets on Resume From S3

Computer Resets on Resume From S3

[John Knottenbelt]

Hello

I have a KT4 Ultra (MS-6590) computer, with latest BIOS. 
I am running kernel 2.6.0-test9-bk21.
When I enter S3, by "echo 3 > /proc/acpi/sleep", the machine suspends all the 
tasks and then appears to power-down (all the hard disks are powered down, 
and the power-supply fan is off).

I know that the machine is not truly off, because I can trigger a resume by 
pressing a key on the keyboard. Also /proc/acpi/alarm works fine.

The trouble is that resume from S3 immediately resets the computer, so it as 
if I had just turned the computer on, rather than resuming the previous 
session.

Can anybody suggest some debugging that I can try to narrow this problem down? 
Perhaps by inserting a forced oops() somewhere so that I can see where it 
gets to before rebooting?

What happens on S3 resume? Is this fault occuring in the DSDT? Is it occuring 
in the kernel? What would the relevant kernel procedures to examine be?

Any help much appreciated

John



-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/

Re: Computer Resets on Resume From S3

[Pavel Machek]

Hi!

> 
> I have a KT4 Ultra (MS-6590) computer, with latest BIOS. 
> I am running kernel 2.6.0-test9-bk21.
> When I enter S3, by "echo 3 > /proc/acpi/sleep", the machine suspends all the 
> tasks and then appears to power-down (all the hard disks are powered down, 
> and the power-supply fan is off).
> 
> I know that the machine is not truly off, because I can trigger a resume by 
> pressing a key on the keyboard. Also /proc/acpi/alarm works fine.
> 
> The trouble is that resume from S3 immediately resets the computer, so it as 
> if I had just turned the computer on, rather than resuming the previous 
> session.
> 
> Can anybody suggest some debugging that I can try to narrow this problem down? 
> Perhaps by inserting a forced oops() somewhere so that I can see where it 
> gets to before rebooting?

Inserting infinite loops can be used for debugging. Or debug leds on
pci card...
								Pavel

-- 
When do you have a heart between your knees?
[Johanka's followup: and *two* hearts?]


-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/

Re: Computer Resets on Resume From S3

[John Knottenbelt]

On Friday 21 November 2003 9:16 pm, Pavel Machek wrote:
>
> Inserting infinite loops can be used for debugging. Or debug leds on
> pci card...
> 								Pavel

Right, I inserted an infinite loop in drivers/acpi/sleep/main.c, just after  
printk(KERN_DEBUG "Back to C!\n"); in acpi_pm_enter()

        /* Johnny's infinite loop */
        for (;;) ;

However, the computer must have rebooted before then. Next thing I tried was 
an infinite loop in arch/i386/kernel/acpi/wakeup.S as the first thing for the 
wakeup_start routine.

ENTRY(wakeup_start)
wakeup_code:
        wakeup_code_start = .
        .code16

1:
        jmp      1b

This too, doesn't catch it. This makes me think that perhaps there is 
something going wrong with the ACPI wake-up address initialisation?

My dmesg (with acpi debugging enabled) is at http://www.doc.ic.ac.uk/~jak97/
dsdt/dmesg.

One warning of possible significance is:
ACPI: No IRQ known for interrupt pin A of device 0000:00:11.1 - using IRQ 255
The 11.1 device is my IDE controller. Could this be an issue? I do not have 
any problems with disks ordinarily.

Please let me know if anything stands out.

Thanks

John



-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/

Re: Computer Resets on Resume From S3

[Pavel Machek]

Hi!

> > Inserting infinite loops can be used for debugging. Or debug leds on
> > pci card...
> 
> Right, I inserted an infinite loop in drivers/acpi/sleep/main.c, just after  
> printk(KERN_DEBUG "Back to C!\n"); in acpi_pm_enter()
> 
>         /* Johnny's infinite loop */
>         for (;;) ;
> 
> However, the computer must have rebooted before then. Next thing I tried was 
> an infinite loop in arch/i386/kernel/acpi/wakeup.S as the first thing for the 
> wakeup_start routine.
> 
> ENTRY(wakeup_start)
> wakeup_code:
>         wakeup_code_start = .
>         .code16
> 
> 1:
>         jmp      1b
> 
> This too, doesn't catch it. This makes me think that perhaps there
> > is 

I've seen something similar before -- it usually meant we left
hardware in "weird" state and it crashed in BIOS.

Can you add "noapic" first, then turn off all unneccesary drivers? No
SMP, please.
								Pavel

-- 
Horseback riding is like software...
...vgf orggre jura vgf serr.


-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/

Re: Computer Resets on Resume From S3

[John Knottenbelt]

On Saturday 22 November 2003 1:19 pm, Pavel Machek wrote:
> I've seen something similar before -- it usually meant we left
> hardware in "weird" state and it crashed in BIOS.
>
> Can you add "noapic" first, then turn off all unneccesary drivers? No
> SMP, please.
> 								Pavel

Hi,

I boot with noapic and also init=/bin/bash so no modules loaded. Same 
behaviour. 

My kernel config is at: http://www.doc.ic.ac.uk/~jak97/dsdt/config.gz
And the dmesg is: http://www.doc.ic.ac.uk/~jak97/dsdt/dmesg
Other info at http://www.doc.ic.ac.uk/~jak97/dsdt

I am running with standard DSDT with relaxed AML checking, however I have also 
tried a fixed DSDT previously and that didn't seem to make any difference at 
all.

Are there any other useful places I might put an infinite loop?

What about something to check before the machine goes into S3?

I am compiling for Athlon architecture, do you think I should try i386?

Could it be a gcc problem? gcc is version 3.3.2.

Thanks

John



-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/

Re: Computer Resets on Resume From S3

[Pavel Machek]

Hi!

> > I've seen something similar before -- it usually meant we left
> > hardware in "weird" state and it crashed in BIOS.
> >
> > Can you add "noapic" first, then turn off all unneccesary drivers? No
> > SMP, please.
> > 								Pavel
> 
> Hi,
> 
> I boot with noapic and also init=/bin/bash so no modules loaded. Same 
> behaviour. 
> 
> My kernel config is at: http://www.doc.ic.ac.uk/~jak97/dsdt/config.gz
> And the dmesg is: http://www.doc.ic.ac.uk/~jak97/dsdt/dmesg
> Other info at http://www.doc.ic.ac.uk/~jak97/dsdt
> 
> I am running with standard DSDT with relaxed AML checking, however I have also 
> tried a fixed DSDT previously and that didn't seem to make any difference at 
> all.
> 
> Are there any other useful places I might put an infinite loop?

I do not have better ideas.

> What about something to check before the machine goes into S3?
> 
> I am compiling for Athlon architecture, do you think I should try
> i386?

That should not make a difference.

> Could it be a gcc problem? gcc is version 3.3.2.

I do not think gcc could cause this.
								Pavel
-- 
When do you have a heart between your knees?
[Johanka's followup: and *two* hearts?]


-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/

Re: Computer Resets on Resume From S3

[John Knottenbelt]

Actually, I'm not sure that the computer is going into S3 properly.
Earlier I had assumed that it did go into S3 because the wake-up events (such 
as keyboard press) caused the computer to wake up. However, these wake-up 
events also cause the computer to wake up from S5.

My hunch is that the wake-up address is not being set correctly. Does anybody 
know how to check this?

Looking at the ACPI Spec this has something to do with the FADT/FACS?

Thanks

John



-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/

Re: Computer Resets on Resume From S3

[Sérgio Monteiro Basto]

On Sat, 2003-11-22 at 17:23, Pavel Machek wrote:
> Hi!
> 
> > > I've seen something similar before -- it usually meant we left
> > > hardware in "weird" state and it crashed in BIOS.
> > >
> > > Can you add "noapic" first, then turn off all unneccesary drivers? No
> > > SMP, please.
> > > 								Pavel
> > 
> > Hi,
> > 
> > I boot with noapic and also init=/bin/bash so no modules loaded. Same 
> > behaviour. 
> > 
> > My kernel config is at: http://www.doc.ic.ac.uk/~jak97/dsdt/config.gz
> > And the dmesg is: http://www.doc.ic.ac.uk/~jak97/dsdt/dmesg
> > Other info at http://www.doc.ic.ac.uk/~jak97/dsdt
> > 
> > I am running with standard DSDT with relaxed AML checking, however I have also 
> > tried a fixed DSDT previously and that didn't seem to make any difference at 
> > all.
> > 
> > Are there any other useful places I might put an infinite loop?
> 
> I do not have better ideas.
> 
> > What about something to check before the machine goes into S3?
> > 
> > I am compiling for Athlon architecture, do you think I should try
> > i386?
> 
> That should not make a difference.
> 
> > Could it be a gcc problem? gcc is version 3.3.2.
> 
> I do not think gcc could cause this.

Asking why fedora core 1, why compile kernel with gcc 3.2.3 and use
gcc3.3.2 for compile everything else.
reply:
if you try (compiling kernel with gcc3.3.x), then it just magically
works for you, and you go on happy, possibly with silent data
corruption, or possibly not.  Roll a 20 sided dice to find out.

-- 
SérgioMB




-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/