aic79xx -> can't suspend

aic79xx -> can't suspend

[Lee Revell]

I have a machine with an Adaptec 2940U2W adapter running 2.6.11.  When I
try to go into standby like so:

    echo standby > /sys/power/state

this is what happens:

Stopping tasks:
===============================================================
stopping tasks failed (1 tasks remaining)
Restarting tasks...<6> Strange, ahc_dv_0 not stopped
done 

The machine fails to suspend, and ahc_dv_0 is now wedged, consuming all
available CPU until reboot.

PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  SWAP WCHAN     COMMAND
735 root      39  19     0    0    0 S 99.3  0.0  53:09.97    0 text.lock [ahc_dv_0]

Lee

Re: aic79xx -> can't suspend

[Jim Crilly]

On 06/23/05 01:51:43PM -0400, Lee Revell wrote:
> I have a machine with an Adaptec 2940U2W adapter running 2.6.11.  When I
> try to go into standby like so:
> 
>     echo standby > /sys/power/state
> 
> this is what happens:

AFAIK no SCSI drivers have had power management functions implemented, a
quick grep for PM_ in drivers/scsi seems to confirm that only the PCMCIA
SCSI drivers even look for PM events.

> 
> Lee
>

Jim.

Re: aic79xx -> can't suspend

[Lee Revell]

On Thu, 2005-06-23 at 15:32 -0400, Jim Crilly wrote:
> On 06/23/05 01:51:43PM -0400, Lee Revell wrote:
> > I have a machine with an Adaptec 2940U2W adapter running 2.6.11.  When I
> > try to go into standby like so:
> > 
> >     echo standby > /sys/power/state
> > 
> > this is what happens:
> 
> AFAIK no SCSI drivers have had power management functions implemented, a
> quick grep for PM_ in drivers/scsi seems to confirm that only the PCMCIA
> SCSI drivers even look for PM events.

Thanks, I suspected as much.  I'll take a look at implementing it then,
can't be too hard.  It would be nice if I could suspend this machine,
because I have a small apartment and it gets hot in here.

Lee

Re: aic79xx -> can't suspend

[Lee Revell]

On Thu, 2005-06-23 at 15:32 -0400, Jim Crilly wrote: 
> On 06/23/05 01:51:43PM -0400, Lee Revell wrote:
> > I have a machine with an Adaptec 2940U2W adapter running 2.6.11.  When I
> > try to go into standby like so:
> > 
> >     echo standby > /sys/power/state
> > 
> > this is what happens:
> 
> AFAIK no SCSI drivers have had power management functions implemented, a
> quick grep for PM_ in drivers/scsi seems to confirm that only the PCMCIA
> SCSI drivers even look for PM events.

Hmm, actually, this is still pretty bad.  Shouldn't the suspend fail
with "power management not implemented for device aic9xxx?" rather than
crashing and burning?

Also, who is the aic9xxx maintainer?  I could not find any reference to
it at all in MAINTAINERS.

Finally, is there an authoritative, up to date driver developers guide
for the SCSI subsystem, like the one for ALSA?

TIA,

Lee

Re: aic79xx -> can't suspend

[Pavel Machek]

Hi!

> > > I have a machine with an Adaptec 2940U2W adapter running 2.6.11.  When I
> > > try to go into standby like so:
> > > 
> > >     echo standby > /sys/power/state
> > > 
> > > this is what happens:
> > 
> > AFAIK no SCSI drivers have had power management functions implemented, a
> > quick grep for PM_ in drivers/scsi seems to confirm that only the PCMCIA
> > SCSI drivers even look for PM events.
> 
> Hmm, actually, this is still pretty bad.  Shouldn't the suspend fail
> with "power management not implemented for device aic9xxx?" rather than
> crashing and burning?

That check is simply not there. We don't know if device is "suspend
not yet implemented" or "suspend is not needed" class... Sorry.

								Pavel
-- 
Boycott Kodak -- for their patent abuse against Java.

Re: aic79xx -> can't suspend

[Lee Revell]

On Thu, 2005-06-23 at 15:32 -0400, Jim Crilly wrote:
> On 06/23/05 01:51:43PM -0400, Lee Revell wrote:
> > I have a machine with an Adaptec 2940U2W adapter running 2.6.11.
> > When I try to go into standby like so:
> > 
> >     echo standby > /sys/power/state
> > 
> > this is what happens:
> 
> AFAIK no SCSI drivers have had power management functions implemented,
> a quick grep for PM_ in drivers/scsi seems to confirm that only the
> PCMCIA SCSI drivers even look for PM events. 

Actually it is implemented in the aic7xxx driver, see ahc_suspend and
ahc_resume.

I tried it with 2.6.12, and I no longer have the problem with the ahc_dv
thread as it no longer exists (AFAICT the functionality is handled by
the SCSI midlayer now?).

Now it just immediately resumes:

[4297399.286000] PM: Preparing system for standby sleep
[4297399.609000] Stopping tasks: ================================|
[4297399.610000] Restarting tasks... done

How can I debug this further?

Lee

Re: aic79xx -> can't suspend

[Pavel Machek]

> On Thu, 2005-06-23 at 15:32 -0400, Jim Crilly wrote:
> > On 06/23/05 01:51:43PM -0400, Lee Revell wrote:
> > > I have a machine with an Adaptec 2940U2W adapter running 2.6.11.
> > > When I try to go into standby like so:
> > > 
> > >     echo standby > /sys/power/state
> > > 
> > > this is what happens:
> > 
> > AFAIK no SCSI drivers have had power management functions implemented,
> > a quick grep for PM_ in drivers/scsi seems to confirm that only the
> > PCMCIA SCSI drivers even look for PM events. 
> 
> Actually it is implemented in the aic7xxx driver, see ahc_suspend and
> ahc_resume.
> 
> I tried it with 2.6.12, and I no longer have the problem with the ahc_dv
> thread as it no longer exists (AFAICT the functionality is handled by
> the SCSI midlayer now?).
> 
> Now it just immediately resumes:
> 
> [4297399.286000] PM: Preparing system for standby sleep
> [4297399.609000] Stopping tasks: ================================|
> [4297399.610000] Restarting tasks... done
> 
> How can I debug this further?

Enable debugging in drivers/base/power/*...

								Pavel

-- 
Boycott Kodak -- for their patent abuse against Java.