[Bug 558968] New: cpufreq no longer works for Pentium 4 / Celeron D processors
http://bugzilla.novell.com/show_bug.cgi?id=558968 http://bugzilla.novell.com/show_bug.cgi?id=558968#c0 Summary: cpufreq no longer works for Pentium 4 / Celeron D processors Classification: openSUSE Product: openSUSE 11.2 Version: Final Platform: i686 OS/Version: openSUSE 11.2 Status: NEW Severity: Major Priority: P5 - None Component: Kernel AssignedTo: kernel-maintainers@forge.provo.novell.com ReportedBy: bruce.lilly@gmail.com QAContact: qa@suse.de Found By: --- Blocker: --- User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.1.4) Gecko/20091016 SUSE/3.5.4-1.1.2 Firefox/3.5.4 Upgrading from opensuse 11.1 (2.6.27.37-pae) to 11.2 (2.6.31.5-0.1-pae) results in no cpufreq control. Laptop (HP ZE5385US) runs hot, with fans running full blast; battery time in minutes rather than hours. Laptop is unusable as such. Looks like although p4-clockmod.c source is present, it's not built per shipping config file. Editing config and rebuilding results in loadable p4-clockmod.ko, but neither ondemand nor conservative governors work. Problem is the change in the source file of the magic number (probably should be a header file macro or enum value!) 10000000 to 10000001 (source code comment mentions ondemand but not conservative). Editing the source code back to 10000000 and rebuilding results in working cpufreq. Conservative governor seems less reponsive than previous kernel (frequency increases with load, but rarely drops back down under light load). Ondemand (with above change) works fine; if anything better than previous kernel. Same issue with desktop machine (E-machines W3503) with Celeron D processor, same fix. Reproducible: Always Steps to Reproduce: 1. verify cpufreq working fine with 2.6.27.37-pae opensuse 11.1 kernel and p4-clockmod module loaded 2. upgrade to opensuse 11.2, 2.6.31.5-0.1-pae kernel 3. no p4-clockmod module, none of the provided modules work on this hardware (speedstep-lib loads, of course, but does nothing w/o p4-clockmod on this hardware) 4. edit config to enable build of p4-clockmod module 5. change 10000001 to 10000000 in p4-clockmod.c source 6. with kernel source package installed, rebuild 7. load speedstep-lib and p4-clockmod and verify operation Actual Results: No cpufreq with opensuse 11.2 as shipped on P4 hardware, including laptops Expected Results: With p4-clockmod change reverted and module built and loaded, operation of cpufreq on P4 is reasonable /proc/cpuinfo from laptop: processor : 0 vendor_id : GenuineIntel cpu family : 15 model : 2 model name : Intel(R) Pentium(R) 4 CPU 2.66GHz stepping : 7 cpu MHz : 333.325 cache size : 512 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 2 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe up pebs bts cid xtpr bogomips : 5316.37 clflush size : 64 power management: /proc/cpuinfo from desktop machine: processor : 0 vendor_id : GenuineIntel cpu family : 15 model : 6 model name : Intel(R) Celeron(R) D CPU 3.33GHz stepping : 4 cpu MHz : 833.312 cache size : 512 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 6 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe lm constant_tsc up pebs bts pni dtes64 monitor ds_cpl cid cx16 xtpr pdcm lahf_lm bogomips : 6666.02 clflush size : 64 power management: -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=558968
http://bugzilla.novell.com/show_bug.cgi?id=558968#c
Jiri Kosina
http://bugzilla.novell.com/show_bug.cgi?id=558968
http://bugzilla.novell.com/show_bug.cgi?id=558968#c1
--- Comment #1 from Alexey Starikovskiy
http://bugzilla.novell.com/show_bug.cgi?id=558968
http://bugzilla.novell.com/show_bug.cgi?id=558968#c2
--- Comment #2 from Bruce Lilly
http://bugzilla.novell.com/show_bug.cgi?id=558968
http://bugzilla.novell.com/show_bug.cgi?id=558968#c3
--- Comment #3 from Bruce Lilly
http://bugzilla.novell.com/show_bug.cgi?id=558968
http://bugzilla.novell.com/show_bug.cgi?id=558968#c4
Thomas Renninger
http://bugzilla.novell.com/show_bug.cgi?id=558968
http://bugzilla.novell.com/show_bug.cgi?id=558968#c5
Bruce Lilly
I don't want the p4-clockmod driver be built if anyhow possible.
Why?
P4 is a power waster and p4-clockmod won't help you much with that,
Nonsense; p4-clockmod has worked with both Pentium 4 and Celeron D for years, is the only Linux cpufreq module that does work with these processors on the specified hardware, and continues to work if the recently added bug is removed and the module is built and loaded (as described earlier).
you should better buy a new fan (and probably also a new battery which, if it has the same age than the CPU should dying soon).
Unless you're offering to underwrite the expense of replacement and to provide environmentally responsible disposal of existing hardware, your comments are neither helpful nor appropriate. Moreover they are irrelevant, as the issue is not hardware problems, but missing kernel functionality which was present and working in earlier releases.
I agree that there are some corner cases where p4-clockmod might make some sense if you want to cut down the TDP (Thermal Design Power) with the powersave governor or setting a max_freq limit (buying a better CPU will pay off rather quickly in performance and power waste, though).
Corner, schmorner; on the laptop especially, extending battery run time and reducing fan noise are important; I again remind you that this functionality provided by p4-clockmod was working fine for years.
Still most people don't want to have it running by default.
Loadable modules need not be loaded. Cpufreq-utils allows setting the governor to performance, if the user so desires. [i.e. your comment is a red herring]
Bruce: If you like to, I can give you some hints how to build a p4-clockmod-kmp package in our build service in a way that others can also use it.
According to the documentation for the build service, it is intended for special purpose vendor hardware. This is not special purpose hardware, rather widely used general purpose hardware and the existing kernel source that supports it. I am interested only in having working systems; I have confirmed that FreeBSD 8.0-RELEASE as distributed fully supports P4 speedstep [and does not have the glibc/kernel ntp bugs (bug 557716) which were also introduced in, and are still unresolved (now more than seven weeks after release) in opensuse 11.2. I have no interest in jumping through hoops to generate and maintain packages for functionality which ought to have remained in the core distribution, put up with rude and irrelevant remarks about replacing hardware, etc. It's far simpler (and cost-free) to convert 4 systems here from opensuse 11.2 to FreeBSD 8.0. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=558968
http://bugzilla.novell.com/show_bug.cgi?id=558968#c6
--- Comment #6 from Thomas Renninger
http://bugzilla.novell.com/show_bug.cgi?id=558968
http://bugzilla.novell.com/show_bug.cgi?id=558968#c
Thomas Renninger
http://bugzilla.novell.com/show_bug.cgi?id=558968
http://bugzilla.novell.com/show_bug.cgi?id=558968#c7
Bruce Lilly
Please read: http://thread.gmane.org/gmane.linux.kernel.cpufreq/6147
Contains a two-sentence unsupported assertion.
Confirms similar problem on different hardware with P4.
There are two arguments for keeping this: - Broken HW (e.g. not working fan) - Let people believe they save power, while they don't
These are not "arguments", they are strawmen (and not convincing ones); nobody but you is talking about "Broken HW (e.g. not working fan)", and the power savings with working p4-clockmod (e.g. openSuSE 11.1 or with resored working code built as a module as described earlier) are both 1) real and measureable (e.g. via DC power input current on a laptop) and 2) clear and obvious (by reduced CPU temperature and power consumption when p4-clockmod is operating) Your claim is that there is no power saving; you'll need to supply measurements (rather than ivory-tower pontification) on real hardware to prove that; I (and apparently others) have real data from real systems that show real power savings. [...]
I agree that it is a problem that Window shows e.g. 300 MHz which is not true and people think they miss an important power saving feature, but this is wrong.
You keep repeating your unsubstantiated claim that there is no power savings; THAT is clearly wrong, and you need to look at actual operation of real hardware. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=558968
http://bugzilla.novell.com/show_bug.cgi?id=558968#c8
Thomas Renninger
I (and apparently others) have real data from real systems that show real power savings. Please provide the data.
-- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=558968
http://bugzilla.novell.com/show_bug.cgi?id=558968#c10
Thomas Renninger
I (and apparently others) have real data from real systems Bruce: I still wait for this data!
-- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
participants (1)
-
bugzilla_noreply@novell.com