cpufreq patches: cleanups, bugfixes, debug infrastructure

The following quite large number of patches is currently in my queue. They are mostly trivial bugfixes and/or cleanups; and because of the large number I won't send them to the list [a few have been sent to the list before, were marked as RFC, and didn't generate any feedback]. If you are interested you can take a look at the patches at the specified URL.

The patches apply on current cpufreq-bk, and even though they were created on top of -rc2, they do apply well on top of -rc3, too.

Dominik

http://www.brodo.de/patches/2004-10-06/1_cpufreq-2.6.9-rc2-userspace-remember-users-wish

Sometimes, a user or userspace daemon wants a specific frequency is set, but that frequency is (temporarily) unavailable or becomes unavailable due to e.g. thermal considerations. As soon as this situation changes, the user wish should be re-evaluated. The attached patch saves the user wish in "cpu_set_freq", and re-uses that whenever appropriate.

Thanks to Thomas Renninger for testing.

http://www.brodo.de/patches/2004-10-06/2_cpufreq-2.6.9-rc2-speedstep-centrino-dont-break-on-FFFFFFFF-ed-frequencies

Some BIOSes export invalid _PSS entries where all bits are set to one. While this is totally contrary to the ACPI specification, BIOS vendors try to tell userspace that these entries should not be considered or used. In order to not fail on such entries, the speedstep-centrino already contains a method to ignore these things. However, due to a wrong ordering of checks the driver aborts nonetheless. So move some checks around.

<3_ is unused>

http://www.brodo.de/patches/2004-10-06/4_cpufreq-2.6.9-rc2-powernow-k7-static

struct acpi_processor_performance *acpi_processor_perf should be static.

http://www.brodo.de/patches/2004-10-06/5_cpufreq-2.6.9-rc2-ondemand-static

dbs_tuners_ins should be static.

http://www.brodo.de/patches/2004-10-06/6_cpufreq-2.6.9-rc2-core-static

Attribute definitions in cpufreq core should be static.

http://www.brodo.de/patches/2004-10-06/7_cpufreq-2.6.9-rc2-userspace-static

cpufreq_sysctl_table should be static even though it'll be gone soon.

http://www.brodo.de/patches/2004-10-06/8_cpufreq-2.6.9-rc2-p4-clockmod-dothan-thinko

Dothan is stepping 13 == 0x0D instead of 0x13 == 19

Thanks to Brian Miles for debugging this issue.

http://www.brodo.de/patches/2004-10-06/9_cpufreq-2.6.9-rc2-rename-cpufreq-acpi

NOTE: For this patch to work you need to move the file arch/i386/kernel/cpu/cpufreq/acpi.c in bitkeeper to arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.c _before_ applying this patch. I haven't done it by means of diff -ruN so that the file history can be kept. END NOTE.

The "acpi" cpufreq driver uses the far too generic module name "acpi". Fix it by renaming it to "cpufreq-acpi". A MODULE_ALIAS("acpi"); will allow for backwards-compatible naming until 2005-06-30.

http://www.brodo.de/patches/2004-10-06/10_cpufreq-2.6.9-rc2-Kconfig-unify-CPU_FREQ

Unify the CPU_FREQ config option.

<11_ is unused>

http://www.brodo.de/patches/2004-10-06/12_cpufreq-2.6.9-rc2-clarify-cpufreq-set-will-go

Clarify that cpufreq_set() and cpufreq_setmax() are parts of the 2.4. API to the userspace governor which will be removed soon after 2005-01-01.

http://www.brodo.de/patches/2004-10-06/13_cpufreq-2.6.9-rc2-debug_infrastructure

Add a unified cpufreq debug infrastructure.

cpufreq_debug_printk will print messages if

a) debugging for the specified part is activated (add 1 for core,
2 for drivers,
4 for governors
and pass value as
cpufreq.debug=<value>
on the kernel command line).

b) and either
b1) printk_ratelimit() allows it to be printed, b2) the user disables ratelimit'ing by passing
cpufreq.debug_ratelimit=0
on the kernel command line, _or_
b3) during driver initialization [unless a different
driver has been initialized successfully] and unloading, and whenever a new policy is set.

The last point may cause for numerous messages if an userspace-based dynamic governor is used which mis-uses the policy interface to set specific frequencies. Oh, the ACPI processor.c interface to the file "performance" uses the same trick, but that interface is marked deprecated as well, so I don't care. And debugging isn't activated normally, you know...

http://www.brodo.de/patches/2004-10-06/14_cpufreq-2.6.9-rc2-debug-speedstep-lib

Use the unified cpufreq debug infrastructure in the speedstep library. Modified existing dprintks and added a few which might be / have been of interest.

http://www.brodo.de/patches/2004-10-06/15_cpufreq-2.6.9-rc2-debug-speedstep-smi

Use the unified cpufreq debug infrastructure in the speedstep-smi driver. Modified existing dprintks and added a few which might be / have been of interest.

http://www.brodo.de/patches/2004-10-06/16_cpufreq-2.6.9-rc2-debug-speedstep-ich

Use the unified cpufreq debug infrastructure in the speedstep-ich driver. Modified existing dprintks and added a few which might be / have been of interest.

http://www.brodo.de/patches/2004-10-06/17_cpufreq-2.6.9-rc2-debug-speedstep-centrino

Use the unified cpufreq debug infrastructure in the speedstep-centrino driver. Modified existing dprintks and added a few which might be / have been of interest, and cleared some whitespace/tab confusion. The big comment regarding which frequency to pass as "old" value wasn't up to date any longer in two regards: the driver's behaviour had changed, and the core is know able to handle such things... anyways, the comment could be removed.

http://www.brodo.de/patches/2004-10-06/18_cpufreq-2.6.9-rc2-debug-p4-clockmod

Use the unified cpufreq debug infrastructure in the p4-clockmod driver. Modified existing dprintks and added a few which might be / have been of interest.

http://www.brodo.de/patches/2004-10-06/19_cpufreq-2.6.9-rc2-debug-longrun

Use the unified cpufreq debug infrastructure in the longrun driver, and add some dprintks which might be / have been of interest.

http://www.brodo.de/patches/2004-10-06/20_cpufreq-2.6.9-rc2-debug-powernow-k8

Use the unified cpufreq debug infrastructure in the powernow-k8 driver. Modified existing dprintks.

http://www.brodo.de/patches/2004-10-06/21_cpufreq-2.6.9-rc2-debug-powernow-k7

Use the unified cpufreq debug infrastructure in the powernow-k7 driver. Modified existing dprintks, and removed the driver-only debug module parameter.

http://www.brodo.de/patches/2004-10-06/22_cpufreq-2.6.9-rc2-debug-longhaul

Use the unified cpufreq debug infrastructure in the longhaul driver. Modified existing dprintks, and removed the driver-only debug module parameter.

http://www.brodo.de/patches/2004-10-06/23_cpufreq-2.6.9-rc2-debug-gx-suspmod

Use the unified cpufreq debug infrastructure in the gx-suspmod driver. Modified existing dprintks.

http://www.brodo.de/patches/2004-10-06/24_cpufreq-2.6.9-rc2-debug-acpi-cpufreq

Use the unified cpufreq debug infrastructure in the acpi-cpufreq driver, and convert the driver's ACPI debug messages to cpufreq debug messages.

http://www.brodo.de/patches/2004-10-06/25_cpufreq-2.6.9-rc2-debug-performance-and-powersave

Add (one each) dprintk for the performance and powersave cpufreq governors.

http://www.brodo.de/patches/2004-10-06/26_cpufreq-2.6.9-rc2-debug-userspace

Add a few dprintks for the userspace cpufreq governor.

http://www.brodo.de/patches/2004-10-06/27_cpufreq-2.6.9-rc2-debug-freq-table

Add a few useful dprintks in the frequency table helpers.

http://www.brodo.de/patches/2004-10-06/28_cpufreq-2.6.9-rc2-debug-core

Add a few useful dprintks to the cpufreq core

http://www.brodo.de/patches/2004-10-06/29_cpufreq-2.6.9-rc2-export-affected_cpus

Show the CPUs which can only change frequencies at the same time in a sysfs file named "affected_cpus". The usefulness of such a file was first noted by Zwane Mwaikambo.

http://www.brodo.de/patches/2004-10-06/30_cpufreq-2.6.9-rc2-powernow-k7-mobile-vid-table-is-only-needed-in-case-of-debug

module_vid_table is only used if DEBUG isn't set, so put #ifdefs around it.

http://www.brodo.de/patches/2004-10-06/31_cpufreq-2.6.9-rc2-longhaul-print-speed-only-needed-in-case-of-debug

print_speed and speedbuffer are only used if DEBUG isn't set, so put #ifdefs around them.

http://www.brodo.de/patches/2004-10-06/32_cpufreq-2.6.9-rc2-core-sysfs--use__ATTR

Use __ATTR in cpufreq.c attribute definitions. Fixes the lack of MODULE_OWNER setting [can't cause problems yet because cpufreq can only be built into the kernel], and allows for code recuction.

http://www.brodo.de/patches/2004-10-06/33_cpufreq-2.6.9-rc2-userspace-sysfs-refcnt

Set .owner in cpufreq_userspace.c sysfs attribute definition.

http://www.brodo.de/patches/2004-10-06/34_cpufreq-2.6.9-rc2-ondemand-sysfs-use__ATTR

Use __ATTR in cpufreq_ondemand.c attribute definitions. Fixes the lack of MODULE_OWNER setting, and allows for code recuction.

http://www.brodo.de/patches/2004-10-06/35_cpufreq-2.6.9-rc2-freq_table-sysfs-refcnt

Set .owner in freq_table's sysfs definition.

http://www.brodo.de/patches/2004-10-06/36_cpufreq-2.6.9-rc2-ondemand-return-EINVAL-on-errors

Return -EINVAL instead of a non-error-value if the user echo'es something bad into attributes generated by the ondemand cpufreq governor.