Is e_powersaver safe to auto-load?

Is e_powersaver safe to auto-load?

[Ben Hutchings]

[-- Attachment #1: Type: text/plain, Size: 1111 bytes --]

The e_powersaver driver comes with a dire warning in Kconfig help:

    VIA C7 Enhanced PowerSaver (DANGEROUS)

    This adds the CPUFreq driver for VIA C7 processors.  However, this driver
    does not have any safeguards to prevent operating the CPU out of spec
    and is thus considered dangerous.  Please use the regular ACPI cpufreq
    driver, enabled by CONFIG_X86_ACPI_CPUFREQ.

Following the bug report <http://bugs.debian.org/566208>, Debian's
cpufrequtils init script was changed to stop loading it.  The driver was
later disabled in Debian kernel configurations.  Rafał appears to have
made it somewhat safer since then, but the help text has not been
changed.

Since commit fa8031aefec0cf7ea6c2387c93610d99d9659aa2 ('cpufreq: Add
support for x86 cpuinfo auto loading v4') the driver will be auto-loaded
on some systems.  If it is still dangerous then it should surely not be
auto-loaded.  But if it is now believed to be safe then the help text
should be updated to say so.

Ben.

-- 
Ben Hutchings
The two most common things in the universe are hydrogen and stupidity.

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 828 bytes --]

Re: Is e_powersaver safe to auto-load?

[Rafał Bilski]

> The e_powersaver driver comes with a dire warning in Kconfig help:
>
>      VIA C7 Enhanced PowerSaver (DANGEROUS)
>
>      This adds the CPUFreq driver for VIA C7 processors.  However, this driver
>      does not have any safeguards to prevent operating the CPU out of spec
>      and is thus considered dangerous.  Please use the regular ACPI cpufreq
>      driver, enabled by CONFIG_X86_ACPI_CPUFREQ.
>
> Following the bug report<http://bugs.debian.org/566208>, Debian's
> cpufrequtils init script was changed to stop loading it.  The driver was
> later disabled in Debian kernel configurations.  Rafał appears to have
> made it somewhat safer since then, but the help text has not been
> changed.
>
> Since commit fa8031aefec0cf7ea6c2387c93610d99d9659aa2 ('cpufreq: Add
> support for x86 cpuinfo auto loading v4') the driver will be auto-loaded
> on some systems.  If it is still dangerous then it should surely not be
> auto-loaded.  But if it is now believed to be safe then the help text
> should be updated to say so.
>
> Ben.
>
To best of my knowledge driver is safe. It is/was safe to use it on all C7 and
Nano processors. Both C7 and Nano don't allow to set frequency or voltage
from outside the processor specs. Nano doesn't allow to set voltage outside
the specs of the given frequency. It will reduce frequency if voltage is too low.
But it was unsafe to use it on systems using faster then neccesary processor
running on lower speed. Which means systems using 1GHz CPu running all
the time on 400MHz due to lack of neccesary heatsink. It should be safe
now because driver is respecting BIOS frequency limit.
But to be honest it isn't needed anymore. Only people with C7-A processors
may need it because their BIOS doesn't have ACPI P-States tables or these
tables are invalid. Some people with C7-D without "paralax " may need it
because it allows now to set lower voltage for the maximum frequency
(most of VIA processors were 1,6GHz reduced to lower frequency and higher
voltage after quality testing). Otherwise it is to no use as "paralax" processors
will use values set by driver as a guide because P-states tables are hardcoded
into them. And nobody cares if driver is using hardware or SMI as long it is
doing it's job.
So in my opinion this driver should not be autoloaded. Only "acpi p-states"
driver can be autoloaded and if it fails it should be user responsibility to
modprobe "e_powersaver".

Sorry for bad english.
Rafał

Re: Is e_powersaver safe to auto-load?

[Rafał Bilski]

> The e_powersaver driver comes with a dire warning in Kconfig help:
>
>      VIA C7 Enhanced PowerSaver (DANGEROUS)
>
>      This adds the CPUFreq driver for VIA C7 processors.  However, this driver
>      does not have any safeguards to prevent operating the CPU out of spec
>      and is thus considered dangerous.  Please use the regular ACPI cpufreq
>      driver, enabled by CONFIG_X86_ACPI_CPUFREQ.
>
> Following the bug report<http://bugs.debian.org/566208>, Debian's
> cpufrequtils init script was changed to stop loading it.  The driver was
> later disabled in Debian kernel configurations.  Rafał appears to have
> made it somewhat safer since then, but the help text has not been
> changed.
>
> Since commit fa8031aefec0cf7ea6c2387c93610d99d9659aa2 ('cpufreq: Add
> support for x86 cpuinfo auto loading v4') the driver will be auto-loaded
> on some systems.  If it is still dangerous then it should surely not be
> auto-loaded.  But if it is now believed to be safe then the help text
> should be updated to say so.
>
> Ben.
>
Hello again,

Mainboard in question is VIA Epia SN10000EG rev. A1. I have access to it
now. I've tested it briefly for a couple of hours and it seems to work fine with
3.2 kernel. I will let you know if more extensive testing will give different result.

Rafał

eps: Detected VIA Model A C7
eps: Current voltage = 1004mV
eps: Current multiplier = 10
eps: Highest voltage = 1004mV
eps: Highest multiplier = 10
eps: Lowest voltage = 956mV
eps: Lowest multiplier = 8
eps: ACPI limit 1.0GHz