ACPI sleep states on Tyan Thunder K8W S2885

ACPI sleep states on Tyan Thunder K8W S2885

[Kyle Rose]

I can't seem to get my Tyan board (AMD 81x1 chipset) to go to sleep such
that wake-on-LAN will wake it back up.  On my other machines, when I
shutdown -h, it (presumably) puts the machine into S5 state
automatically, and WOL works like a charm; on this machine, shutdown -h
  puts the machine into an actual "off" state in which WOL won't wake it
back up.

Moreover, if I try to echo 5 > /proc/acpi/sleep with full debugging, I
get absolutely nothing in dmesg.

Here are the ACPI-related lines from my boot log (minus the lines
regarding ACPI routing of specific IRQ's):

PCI: Using ACPI for IRQ routing
hpet_acpi_add: no address or irqs in _CRS
ACPI wakeup devices:
PCI1 USB0 USB1 PS2K GOLA GLAN GOLB SMBC AC97 MODM PWRB
ACPI: (supports S0 S1 S4 S5)
ReiserFS: hda5: found reiserfs format "3.6" with standard journal
ReiserFS: hda5: using ordered data mode
ACPI: 'PS2K' and 'PCI1' have the same GPE, can't disable/enable one
seperately
ACPI: 'GLAN' and 'PCI1' have the same GPE, can't disable/enable one
seperately

/proc/acpi/wakeup:

Device  Sleep state     Status
PCI1       4             enabled
USB0       4            disabled
USB1       4            disabled
PS2K       1             enabled
GOLA       4            disabled
GLAN       4             enabled
GOLB       4            disabled
SMBC       4            disabled
AC97       4            disabled
MODM       4            disabled
PWRB       1            *enabled


and /proc/acpi/sleep:

S0 S1 S4 S5

Furthermore, if I shut down from Windows, it *does* go into what I
presume is the S5 state, so this is a software problem, not hardware.

Any suggestions on debugging?

Cheers,
Kyle

Re: ACPI sleep states on Tyan Thunder K8W S2885

[Adam Belay]

On Fri, Apr 29, 2005 at 12:36:23PM -0400, Kyle Rose wrote:
> I can't seem to get my Tyan board (AMD 81x1 chipset) to go to sleep such
> that wake-on-LAN will wake it back up.  On my other machines, when I
> shutdown -h, it (presumably) puts the machine into S5 state
> automatically, and WOL works like a charm; on this machine, shutdown -h
>   puts the machine into an actual "off" state in which WOL won't wake it
> back up.

The "off" state is always considered S5.  Adding wake devices does not make
the state S5.

> 
> Moreover, if I try to echo 5 > /proc/acpi/sleep with full debugging, I
> get absolutely nothing in dmesg.

-->snip

> Furthermore, if I shut down from Windows, it *does* go into what I
> presume is the S5 state, so this is a software problem, not hardware.
> 
> Any suggestions on debugging?
> 
> Cheers,
> Kyle

S5 wake devices can be a very fuzzy area.  In general, the ACPI spec
recommends that wake capabilities are enabled before halting the system.
Therefore, your network card driver may need to specifically handle this.

For starters, we should probably look at "lspci -vv".  I'm assuming your
network card is PCI.  This will allow us to see if it's capable of waking
from D3-cold, the state it will most likely be in during S5.

I would also check your BIOS configuration and see if it's possible to
specifically enable wake-on-lan.  Let me know if this helps.

Finally, what kernel version are you using?

Thanks,
Adam

Re: ACPI sleep states on Tyan Thunder K8W S2885

[Kyle Rose]

> S5 wake devices can be a very fuzzy area.  In general, the ACPI spec
> recommends that wake capabilities are enabled before halting the system.
> Therefore, your network card driver may need to specifically handle this.

I'll check into that.

> For starters, we should probably look at "lspci -vv".  I'm assuming your
> network card is PCI.  This will allow us to see if it's capable of waking
> from D3-cold, the state it will most likely be in during S5.

I'll have to do this tonight.  I can't start the machine remotely. :)

> I would also check your BIOS configuration and see if it's possible to
> specifically enable wake-on-lan.  Let me know if this helps.

It only has an option for "wake on PME," which I've activated.  Again,
as I've said, shutting down from Win XP allows WOL, so the difference is
between how Linux and Windows handle the GigE device or handle shutdown.

> Finally, what kernel version are you using?

Sorry, I should have given more details.  It's Linux-2.6.11.7 with no
other patches.  I think the earliest I tested this was with 2.6.9, and
it didn't work there either.

Thanks,
Kyle