Re: ACPI on a sony c1xd?

[Bernd Ernesti <netbsd%lists.veego.de@localhost>]

On Sun, Mar 09, 2008 at 01:03:39PM -0400, Jared D. McNeill wrote:
> Bernd Ernesti wrote:
> 
> >how can I suspend a sony c1xd with an ACPI kernel?
> >There is no problem to do a supend to disk with a non ACPI kernel while 
> >pressing
> >the Fn-F12 key combination. But it does nothing with a ACPI kernel.
> 
> Please provide the output of 'sysctl hw.acpi.supported_states' to see if 
> the machine supports S3 at all. The NetBSD kernel does not currently 
> support S4.

S0 S1 S3 S4 S5

Hmm, so S4 is suspend to disk and S3 is suspend to RAM, which in my case
is not a solution, because sometime I suspend the notebook for weeks or
months and remove the battery.

> >And the system is very slow, which is easy to understand while looking
> >at the dmesg difference between a non-ACPI and an ACPI kernel:
> >
> >-cpu0: Intel Mobile Pentium II (686-class), 397.02 MHz, id 0x66d
> >+cpu0: Intel Mobile Pentium II (686-class), 112.96 MHz, id 0x66d
> 
> Does the system support Speedstep?

It has some kind of logic to reduce the frequence, but if that is Speedstep ...

> >And I get this warning while inserting the network card:
> >
> >tlp0: WARNING: powerhook_establish is deprecated
> 
> tlp needs to be converted to PMF before the system will attempt to suspend.

I can test patches, but converting it to PMF is not what I can do at the
moment.

> >Not to mention the other warnings:
> >
> >WARNING: can't reserve area for I/O APIC.
> 
> This system is probably too old to have an I/O APIC, I wouldn't worry 
> about it.

Ok.

> >acpi_pci_link: WARNING: powerhook_establish is deprecated
> >acpi_pci_link: WARNING: powerhook_establish is deprecated
> >acpi_pci_link: WARNING: powerhook_establish is deprecated
> >acpi_pci_link: WARNING: powerhook_establish is deprecated
> 
> Again, these need to be converted to PMF. ioapic handles capturing and 
> restoring IRQ routing on modern laptops, but we need to use 
> acpi_pci_link on older ones. Our acpi_pci_link code doesn't attach as a 
> device driver to the ACPI device nodes as it should, and PMF needs a 
> device_t to be associated with a power handler.
> 
> You might want to capture/restore this state in acpi's PMF handler 
> directly as a workaround.

I would do if I had an idea what I should do here.

It seems that a non acpi kernel let me do a suspend to disk, but the states
are horrible messed up, so I get pci errors from the kernel (or bios?).

I will open two prs, one for tlp and one for the acpi_pci_link warnings.

Bernd