RE: [ACPI] Vaio FX403 IRQ problem and general report

[Diefenbaugh, P. S <pau...@in...>]

Tom:

> broken Vaios out there with the exactly same problem and I now wonder if
> there's anything planed to integrate some type of fix into future acpi
> releases since the current solution, althogh working fine for me,
obviously
> isn't that clean at all.

This isn't Vaio-specific: we've got an issue with resolving the bus number
on subordinate PCI-PCI bridges.  Should have a fix out next week.

> There is nothing about a fan in /proc/acpi although I've compiled the
kernel
> with CONFIG_ACPI_FAN=y. Does that mean that fan control simply isn't
> possible on that machin or something eles?

Your DSDT (AML) has to specify a fan, and many systems don't.

> Could someone sum up what effects should be excpected from throttling and
> what I does in general?

Throttling is a linear (e.g. 50% performance at 50% power) performance
control that is typically used by thermal passive cooling.

> Echoing anything above 1 to CPU0/limit doesn't have any effect when beeing
> at
> L0 (by having effect I just mean it isn't reported back when cating
> CPU0/limit). Moreover an echo x > limit (with x > 1) while limit is at L1
> let
> it  jump back to L0. What exactly is limiting about? In limit there is for

See my previous email describing the limit interface.

> example the line "L0:                  100% [P0:T0]"
> For P0 means maximum performance, max. power consumption, right?

P0 means maximum performance (e.g. SpeedStep) state.

> T0, or more general Tx means throttling, right? Since for L1 to L4 Px
alway
> is P0 and Tx goes from T0 to T4 what's the difference to "normal"
> throttling?

None.  The purpose of this limit interface is to gracefully support P-states
once we develop SpeedStep support.

> cat CPU0/performace returns <not supported>
> Is that related to SpeedStep (i've read something about that this is,
unlike
> amd's powernow, not supported)?

Support for the ACPI 2.0 processor performance control interface will be
added shortly.

> Echoing something to CPU0/power doesn't seem to have any effect (not sure
if
> it sould). It's alway at C2.

Writes to power do nothing currently.

> In /proc/acpi there also is a directory named embedded_controller that
> contains just EC0/info. Might that have something to do with for example
> those extra buttons above the keyboard that are dead under Linux (well you
> can get them to life by using sonypi kernel module that actually does that
> over acpi)?

I'm not sure how the EC is used on your system (it is often used to detect
button events, though).

> cat event always returns "cat: event: Device or resource busy"
> Is that normal?

That's a bug.  Investigating...

Thanks for the data!

-- Paul