Cpufreq for 2.4: what is the best source?

Hi all,

What is the best source for CPUFREQ for a 2.4.22-rc2 kernels? Is it the 2.4.22-rc2-ac1 kernel version or the 2.4 tarballs on ftp://ftp.linux.org.uk/pub/linux/cpufreq/ or the LINUX_2_4 tag from CVS?

(BTW I want to use it with a P4-M CPU)

Does someone know when codemonkey comes on-line again? Or is there another place where I can find up to date documentation?

Regards,

Bas.

Either 2.4.22-rc2-ac1 or the latest 2.4 backport on www.codemonkey.org.uk - but as that's offline at the moment, I'd try -ac. Or 2.6.0-test3-bk-current :)

IIRC there's zero extra documentation on "codemonkey.org.uk" The best documentation is within the linux kernel sources [2.4.-patched, 2.4-ac, or 2.6.] in Documentation/cpu-freq/

Dominik

On Friday 15 August 2003 17:41, Dominik Brodowski wrote:

Hi,

unfortunely I can't use -ac 2.4. Kernels ... (because of my hardware and not 2.6. because of a Cisco VPN). Is (or was) www.codemonkey.org.uk really the only place, to get the cpufreq backports against vanilla 2.4. kernels?

What about cvs? And the ftp location mentioned by Bas Mevissen is outdated?

BTW I need cpufreq for Pentium M.

Regards
Georg Sauthoff

The ftp is updated each night. Whether it contains anything useful depends on whether people check the changes into the cvs tree(s).

If people want a 2.4-ac snapshot and the maintainers are willing to create a cvs branch for them, its pretty simple to add it to nightly scripts.

Russell King ([email protected])                The developer of ARM Linux
             http://www.arm.linux.org.uk/personal/aboutme.html

Yes.

The ftp location takes the latest cvs sources. And the CVS sources aren't kept up to date for the 2.4. backports.

Dominik

Can someone please make a backport of CPUFREQ to 2.4.22-rc2? The -ac has
a new scheduler (backport from 2.6?) and that breaks swsusp for the
moment. So I would like to go for 2.4.22-rc2 (or later) with the -ac patches.

It seems that more people are waiting for a 2.4.22+swsusp+cpufreq, so it is not a waste of time.

Regards,

Bas.

Already done. As Dave's site isn't yet back online, it can temporarily be accessed at
http://www.brodo.de/cpufreq_tmp/cpufreq-2.4.22-rc2-1__.gz

Dominik

Hmmm. A few problems with this patch:

1) double quotes "userspace" governor configuration text confuse make menuconfig. So it shows no help text. Escaping them didn't help, so I made a new text. Note that this is a problem in the -ac too.

--- linux-2.4.22-rc2-sgm1/arch/i386/config.in.orig 2003-08-18
19:37:12.000000000 +0200
+++ linux-2.4.22-rc2-sgm1/arch/i386/config.in 2003-08-18 20:32:59.000000000
+0200
@@ -201,7 +201,7 @@
bool ' CPU frequency table helpers' CONFIG_CPU_FREQ_TABLE define_bool CONFIG_CPU_FREQ_PROC_INTF y
comment 'CPUFreq governors'
- bool ' "userspace" for userspace frequency scaling'
CONFIG_CPU_FREQ_GOV_USERSPACE
+ bool ' Support for governing from userspace' CONFIG_CPU_FREQ_GOV_USERSPACE
define_bool CONFIG_CPU_FREQ_24_API y
comment 'CPUFreq processor drivers'
dep_tristate ' AMD Mobile K6-2/K6-3 PowerNow!' CONFIG_X86_POWERNOW_K6
$CONFIG_CPU_FREQ_TABLE

2) Compile error from speedstep-lib.c caused by exported symbols. Patch below:

--- linux-2.4.22-rc2-sgm1/arch/i386/kernel/Makefile.orig 2003-08-18
20:58:28.000000000 +0200
+++ linux-2.4.22-rc2-sgm1/arch/i386/kernel/Makefile 2003-08-18
21:56:15.000000000 +0200
@@ -14,7 +14,7 @@

O_TARGET := kernel.o

-export-objs := mca.o mtrr.o msr.o cpuid.o microcode.o i386_ksyms.o time.o
setup.o
+export-objs := mca.o mtrr.o msr.o cpuid.o microcode.o i386_ksyms.o time.o
setup.o speedstep-lib.o

obj-y := process.o semaphore.o signal.o entry.o traps.o irq.o vm86.o \
ptrace.o i8259.o ioport.o ldt.o setup.o time.o sys_i386.o \

Files attached for convieniance :-)

Regards,

Bas.

@@ -201,7 +201,7 @@
    bool ' CPU frequency table helpers' CONFIG_CPU_FREQ_TABLE
    define_bool CONFIG_CPU_FREQ_PROC_INTF y
    comment 'CPUFreq governors'
-   bool ' "userspace" for userspace frequency scaling' CONFIG_CPU_FREQ_GOV_USERSPACE
+   bool ' Support for governing from userspace' CONFIG_CPU_FREQ_GOV_USERSPACE
    define_bool CONFIG_CPU_FREQ_24_API y
    comment 'CPUFreq processor drivers'
    dep_tristate ' AMD Mobile K6-2/K6-3 PowerNow!' CONFIG_X86_POWERNOW_K6 $CONFIG_CPU_FREQ_TABLE

@@ -14,7 +14,7 @@
 
 O_TARGET := kernel.o
 
-export-objs     := mca.o mtrr.o msr.o cpuid.o microcode.o i386_ksyms.o time.o setup.o
+export-objs     := mca.o mtrr.o msr.o cpuid.o microcode.o i386_ksyms.o time.o setup.o speedstep-lib.o
 
 obj-y	:= process.o semaphore.o signal.o entry.o traps.o irq.o vm86.o \
 		ptrace.o i8259.o ioport.o ldt.o setup.o time.o sys_i386.o \

How up to date is this one? It looks like it doesn't work with a ICH4 P4-M combination. With 2.6.0-test3 it works out of the box.

With speedstep_ich.o and (..._lib.o) loaded, I get:
cd /proc/sys/cpu ; cat *
2000000 # speed_max, which is correct
0 # speed_min
0 # spped_max

BTW where can I find documentation about how to set a different policy with the patch above? The documentation in the tree deliberately doesn't detail it...

(I mean setting the policy with e.g. echo userspace >/proc/cpufreq)

Bas.

echo 0%100%userspace > /proc/cpufreq

Ducrot Bruno

--  Which is worse:  ignorance or apathy?
--  Don't know.  Don't care.