[Bug 231107] New: CPU frequency policy
https://bugzilla.novell.com/show_bug.cgi?id=231107 Summary: CPU frequency policy Product: openSUSE 10.2 Version: Final Platform: x86-64 OS/Version: Other Status: NEW Severity: Normal Priority: P5 - None Component: Kernel AssignedTo: kernel-maintainers@forge.provo.novell.com ReportedBy: f.alabas@ttnet.net.tr QAContact: qa@suse.de I have a compaq presario M2000 laptop whose CPU works on powernow-k8 kernel driver. The CPU has 3 frequency steps 800, 1600 and 1800 MHz. When I boot the machine with AC plugged it works on these 3 steps according to kernel's ondemand governor. Later if AC unplugged it works on 2 steps; 800 and 1600 MHz, again when I plugged in the AC it works on 3 steps. When I boot the machine with AC unplugged, it works on 2 step; 800 and 1600 MHz. Until here everthing is ok. But when I plugged in the AC, the cpu continues working on the 2 step and does not step on to 1800 MHz which is the highest hardware limit while AC plugged. Summary, only if I boot with AC unplugged and then plug in the AC, the kernel does not adapt the cpu frequency policy to AC plugged mode. I need to change it manually using "cpufreg-set -u 1.8GHz". On the other hand, other parameters such as screen brightness reacts well to AC on/off conditions. The /proc/cpuinfo file;------------------------------------ processor : 0 vendor_id : AuthenticAMD cpu family : 15 model : 36 model name : AMD Turion(tm) 64 Mobile Technology ML-32 stepping : 2 cpu MHz : 1800.000 cache size : 512 KB fpu : yes fpu_exception : yes cpuid level : 1 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 syscall nx mmxext fxsr_opt lm 3dnowext 3dnow up pni lahf_lm bogomips : 1593.56 TLB size : 1024 4K pages clflush size : 64 cache_alignment : 64 address sizes : 40 bits physical, 48 bits virtual power management: ts fid vid ttp tm stc The output of "cpufreq-info -c 0" ( AC plugged )---------------------------- driver: powernow-k8 CPUs which need to switch frequency at the same time: 0 hardware limits: 800 MHz - 1.80 GHz available frequency steps: 1.80 GHz, 1.60 GHz, 800 MHz available cpufreq governors: conservative, ondemand, userspace, powersave, performance current policy: frequency should be within 800 MHz and 1.80 GHz. The governor "ondemand" may decide which speed to use within this range. The output of "cpufreq-info -c 0" ( AC unplugged )------------------------- driver: powernow-k8 CPUs which need to switch frequency at the same time: 0 hardware limits: 800 MHz - 1.80 GHz available frequency steps: 1.80 GHz, 1.60 GHz, 800 MHz available cpufreq governors: conservative, ondemand, userspace, powersave, performance current policy: frequency should be within 800 MHz and 1.60 GHz. The governor "ondemand" may decide which speed to use within this range. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #1 from f.alabas@ttnet.net.tr 2006-12-29 15:56 MST ------- Created an attachment (id=111213) --> (https://bugzilla.novell.com/attachment.cgi?id=111213&action=view) boot.msg files for ac and battery I am not sure, but the boot log files may be useful. I have compared them, there is no difference. Anyway I am attaching them. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 gregkh@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- AssignedTo|kernel- |trenn@novell.com |maintainers@forge.provo.nove| |ll.com | -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 trenn@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |ASSIGNED ------- Comment #2 from trenn@novell.com 2007-01-13 06:52 MST ------- As this is a HP laptop, this sounds related to bug #175702 and bug #179702. Can you please try whether the kernel I put here: ftp.suse.com/pub/people/trenn/hp_fixes_final/ does help. Be aware that you have to boot the kernel twice as the system gets fixed when shutting it down! Does everything work fine after booting and rebooting this kernel? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 trenn@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |f.alabas@ttnet.net.tr -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 f.alabas@ttnet.net.tr changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED Info Provider|f.alabas@ttnet.net.tr | ------- Comment #3 from f.alabas@ttnet.net.tr 2007-01-15 11:09 MST ------- Your kernel in ( ~/hp_fixes_final) did not help me. The scenario of my issue is the same with that kernel, too. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 trenn@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |hmacht@novell.com Status|ASSIGNED |NEEDINFO Info Provider| |f.alabas@ttnet.net.tr ------- Comment #4 from trenn@novell.com 2007-01-15 12:17 MST ------- Holger can this be the "unsorted frequencies" bug? Hmm, you are using the ondemand governor (cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor)? Can you check whether you run the latest BIOS for this machine and if not update, pls. If this does not work: Do you have BIOS configuration options to modify PowerNow! or cpufreq related settings? Can you also attach acpidump please (of latest BIOS). -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #5 from f.alabas@ttnet.net.tr 2007-01-15 13:59 MST ------- Yes, I have also checked the (......../cpufreq/scaling_governor file) and the policy is ondemand. I have the latest BIOS installed on my machine. I see just one configuration option with two parameters for "optimization of battery usage on battery operation mode." The first parameter is "battery-life" which tries to use the cpu at lower frequencies on battery operation mode. The second parameter is "performance" which simply ignores the battery operation mode. I have tested both parameters but, no help for my issue. Here is the output of acpidump. ( of latest BIOS ) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 f.alabas@ttnet.net.tr changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED Info Provider|f.alabas@ttnet.net.tr | ------- Comment #6 from f.alabas@ttnet.net.tr 2007-01-15 14:01 MST ------- Created an attachment (id=113030) --> (https://bugzilla.novell.com/attachment.cgi?id=113030&action=view) The output of acpidump command -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 trenn@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |f.alabas@ttnet.net.tr ------- Comment #7 from trenn@novell.com 2007-01-18 08:38 MST ------- Does: processor.max_cstates=1 help? If you do a cat /proc/acpi/processor/*/power There should be line: max_cstate: C1 then, instead of: max_cstate: C8 -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 f.alabas@ttnet.net.tr changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED Info Provider|f.alabas@ttnet.net.tr | ------- Comment #8 from f.alabas@ttnet.net.tr 2007-01-18 17:20 MST ------- I tried "processor.max_cstate=1" parameter at GRUB and have noted the "max_cstate" value as C1 instead of C8 in the related "power" process file but NO help ! (I also tried to reload the powernow-k8 module with that parameter, but I think this is not supported cause modprobe gives error.) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #9 from trenn@novell.com 2007-01-19 05:26 MST ------- It seems the system should (maybe only in some cases, the first if loop) run on reduced frequency on battery (corresponding DSDT BIOS code, in the _PSR method of the AC device object. The _PSR method tells the kernel whether AC is online or offline): If (LAnd (LEqual (\_SB.PCI0.LPC0.EC0.BATP, 0x34), LEqual (\_SB.POTL, 0x01))) { Store (0x00, \_PR.CPU0._PPC) Notify (\_PR.CPU0, 0x80) } Else { If (LEqual (\_SB.ACST, 0x01)) { Store (0x00, \_PR.CPU0._PPC) Notify (\_PR.CPU0, 0x80) } Else { If (LEqual (\_SB.PWDC, 0x06)) { Store (0x02, \_PR.CPU0._PPC) Notify (\_PR.CPU0, 0x80) } If (LEqual (\_SB.PWDC, 0x05)) { Store (0x01, \_PR.CPU0._PPC) Notify (\_PR.CPU0, 0x80) } } } This: If (LEqual (\_SB.ACST, 0x01)) is the check for whether AC is online or offline and _PPC is set to zero in this case (means all frequencies allowed). If above is zero _PPC is either set to 0x1 (all but the highest) or 0x2 (all but the two highest freqs are allowed). After setting _PPC an ACPI event (Notify (\_PR.CPU0, 0x80)) should get generated to tell the kernel that _PPC needs to be reevaluated. Summary: DSDT BIOS code seems to be correct. If the machine is running on lower freqs on battery, this may be intended, but on AC all frequncies must always be allowed. Now it's hard to debug this. Can you have a look at (when booted with AC unplugged and then plug in AC) whether the ACPI processor events are generated correctly (this should be run before pluggin in AC): tail -f /var/log/acpid If they are reported correctly (an AC and a processor event), does it help to unplug and replug AC? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #10 from f.alabas@ttnet.net.tr 2007-01-19 09:43 MST ------- I looked at the "acpid" log file for the AC and processor events; they all seem working well by reacting and generating necessary events according to triggers from shifting battery to AC method and vice versa such as; after boot on battery, in "acpid" file; CPU0 events are received, notified and completed; 0x80 0x01 ACAD events are ok,too; 0x00 0x00 Then when AC plugged, events are processed and values are changed; CPU0; 0x80 0x00 ACAD; 0x00 0x01 Consequent unplugging-plugging trials bring NO change on processor frequency policy. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #11 from trenn@novell.com 2007-01-20 05:24 MST ------- Tks. Hmm, the order of happening events should be the other way around. If you (un)plug AC you must first see the ACAD, then a cpu event. Can you paste all lines added to acpid log file when you plug and unplug AC adapter when cpufreq is affected, please. If this looks fine, I don't see another possibility than trying a kernel with cpufreq and ACPI debug comipiled in. You can try whether the kernel-debug (rpm) boots for you. If not, pls tell me how familiar you are with compiling a SUSE kernel. Maybe we can meet via IRC or ICQ and I can give you a helping hand... -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #12 from f.alabas@ttnet.net.tr 2007-01-20 06:22 MST ------- Created an attachment (id=114033) --> (https://bugzilla.novell.com/attachment.cgi?id=114033&action=view) acpid events for plug-unplug 15:06:12 when system is on battery 15:06:23 when AC plugged-in 15:07:07 when AC unplugged again -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #13 from f.alabas@ttnet.net.tr 2007-01-20 06:33 MST ------- I think the acpid is working correctly according to your indication on your comment #11. Now, if we need kernel compilation, I do not know consistent information so I will need your help. On the other hand, I am not sure how to proceed with kernel-debug rpm ? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #14 from f.alabas@ttnet.net.tr 2007-01-20 10:21 MST ------- I installed and tested a boot from kernel-debug rpm. At kdb console, I enter command go, then boot failed at; [ACPI debug] Integer:0x00000033 A bit later, i see another message which is generated periodically as; [ACPI debug] String:[0x1A] "_Q09: Delay Timer Handler" then boot seems stop, because I can not interfere the "boot/the periodically generated message" anymore. I tried to reboot the machine with acpi=off, just to see if the machine boots, however the machine hung at boot with a last message related to ALSA atiixp. I am sorry because of this stupid comment, but I do not know how to provide you further information as I have no experience about using kernel-debug rpm and compiling a kernel, especially Suse kernel :-(( -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 trenn@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |FIXED ------- Comment #15 from trenn@novell.com 2007-01-23 05:02 MST ------- You should see a similar or the same problem as described here: http://bugzilla.kernel.org/show_bug.cgi?id=7859 I added the patch from comment #6 there, which should also fix your problem. Please reopen if you should still have problems with the next update kernel. You can test it by installing the kernel of the day (better wait a day or two until everything got built and synced): ftp.suse.com/pub/projects/kernel/kotd/10.2-x86_64/SL102_BRANCH/kernel-default.x86_64.rpm Check whether the change is included by doing: rpm -qp --changelog kernel-default.x86_64.rpm |less You possibly have to reinstall kmp (rpm -qa |grep kmp) packages to get e.g. additional kernel packages like nvidia or wlan to work again. Please reopen if you still have problems with the update kernel. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 f.alabas@ttnet.net.tr changed: What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |REOPENED Resolution|FIXED | ------- Comment #16 from f.alabas@ttnet.net.tr 2007-02-06 11:42 MST ------- I've installed kotd-20070205 which includes the changes you made. Unfortunately there is no resolution for my issue ! With the new kernel, I have also followed up your previous suggestions and investigations to find something useful, but things seem like all the same. Can my issue be related to PolicyKit or Hal Daemon ? At the first boot of kotd, I have seen some strange messages at the boot screen, but not later. strange lines was exactly; starting PolicyKit Daemon ??? ??????????? ??? ??????????? ??? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #17 from f.alabas@ttnet.net.tr 2007-02-06 12:41 MST ------- I have made an online chat with an HP technician to gather some information that might be useful. According to my poor perception, he said nothing different than in your comment #9. Shortly; when on AC, 3 frequencies are allowed. When on BAT, if "battery mode-related BIOS option" is set to "battery-life", then only 800 MHz is allowed; seems "3rd if sub-loop" at comment #9, or if "battery mode-related BIOS option is set to "performance, then 800 and 1600 MHz are allowed; seems "4th if sub-loop" at comment #9. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 trenn@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|REOPENED |ASSIGNED ------- Comment #18 from trenn@novell.com 2007-02-07 03:31 MST ------- With the kernel of the day you installed, do you have these lines in the changelog?: ("rpm -qp --changelog /path/kernel-xy.rpm| less" or rpm -q --changelog kernel-xy |less (installed version)) Wed Jan 24 09:20:21 CET 2007 - trenn@suse.de - patches.fixes/cpufreq_dont_call_ppc_on_init.patch: ACPI: fix cpufreq regression (231107, http://bugzilla.kernel.org/show_bug.cgi?id=7859). If the kernel is too new, you may have missed this one, which should fix the issue. If you verified that it's in, it still could be a BIOS issue, not informing the kernel about the changed cpufreq limits and I can have a look at your acpidump information and try to find out. I debugged this (with a mainline kernel) on an other turion which also limits cpufreq on battery and everything worked fine there with this patch. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 trenn@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |f.alabas@ttnet.net.tr -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 f.alabas@ttnet.net.tr changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED Info Provider|f.alabas@ttnet.net.tr | ------- Comment #19 from f.alabas@ttnet.net.tr 2007-02-07 12:58 MST ------- Certainly, there are the lines... I had installed the new kernel with "rpm -ihv" and booted from it at least two times before making test. BTW, I found some documentation at the below link for compiling a kernel with ACPI debug. Do you think this doc is consistent ? Shall I follow it up ? Maybe with some additional hints by you ? http://powersave.sourceforge.net/powersave/ACPI_005fDebugging.html#ACPI_005f... -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #20 from trenn@novell.com 2007-02-08 07:50 MST ------- First be sure you really still have the problem: cat /dev/zero >/dev/null & will give you 100% load per CPU (start it twice for a dual core machine) with fg and CTRL-c you can kill these processes again and cpufreq will go down. watch -n1 cat /sys/devices/system/cpu/cpu*/cpufreq/{scaling_max_freq,scaling_cur_freq} shows you the max and current frequencies. Play a bit with that with the new kernel and report back the exact behaviour after (un)pluggin AC. (Also have a look at /proc/acpi/thermal_zone/*/temperature and trip_points files), if the temperature exceeds the passive cooling limit, frequncies might also got limited then. 15 Overriding the DSDT /usr/share/doc/packages/powersave/ should be newer. This might be of interest for debugging whether we face a BIOS problem if max freq still stays low after plugging in AC adapter. Ping me again if you are sure something is still wrong and I go through BIOS stuff and try to find the culprit or a useful debugging cenario. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #21 from f.alabas@ttnet.net.tr 2007-02-08 18:00 MST ------- :-( still have the same problem. With full load on CPU, (un)plug AC does not have any effect on processor speed. In both case,the output of "watch -n1 cat /sys/devices/system/cpu/cpu0/cpufreq/{scaling_max_freq,scaling_cur_freq}" 1600000 1600000 The processor works under passive cooling limit as output of cat /proc/acpi/thermal_zone/THRM/temperature when idle temperature= ~45 when full load temp= ~60 and output of cat /proc/acpi/thermal_zone/THRM/trip_points critical (S5): 105 C passive: 93 C: tc1=2 tc2=3 tsp=100 devices=0xffff810037ed7a10 I have also checked the files of /proc/acpi/ac_adapter/ACAD/state /proc/acpi/battery/BAT1/state and they show the correct status in response to AC plug/unplug conditions. As a reminder from bug opening, the problem exists only if I boot the machine when AC unplugged. Then plugging in the AC does not tell the processor to reach highest available frequency which is the above story. However, when I boot the machine with AC plugged, system works ok. AC-plugged under full load the output of "watch -n1 cat /sys/devices/system/cpu/cpu0/cpufreq/{scaling_max_freq,scaling_cur_freq}" 1800000 1800000 and when AC-unplugged the output is 1600000 1600000 then plugging again 1800000 1800000 The viewpoint is; while boot on AC, BIOS and Kernel coordinates, so no problem happens; but while boot on Battery, BIOS and Kernel lost the coordination, so either BIOS does not send whatever the necessary signal to Kernel or Kernel does not grant whatever the signal BIOS send. So something works strange if the machine boots on battery although all acpi-related events and values in the acpi-related files react well. One more thing; although there is no information about "powering type at boot", the most similar issue to my case had been reported for a HP nx6125 laptop. Interestingly, that machine that I know well is the twin brother of mine. i.e. same cpu (except L2 cache size), same chipset etc... I noted some other, maybe unusual thing in the file /sys/devices/system/cpu/cpu0/cpufreq/ondemand/ignore_nice_load When boot on battery its value is 1, then AC plugged its value is 0. When boot on AC plugged its value is 0, then AC unplugged its value is still 0. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 trenn@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |seife@novell.com Status|ASSIGNED |NEEDINFO Info Provider| |f.alabas@ttnet.net.tr ------- Comment #22 from trenn@novell.com 2007-02-09 03:39 MST -------
I noted some other, maybe unusual thing in the file /sys/devices/system/cpu/cpu0/cpufreq/ondemand/ignore_nice_load When boot on battery its value is 1, then AC plugged its value is 0. When boot on AC plugged its value is 0, then AC unplugged its value is still 0.
This is a good point! Maybe we have a race or locking problem here. Hmm, but I cannot see anything obvious in the code.. The ignore_nice_load is set when you run on battery, in this case lower priority programs started with nice factor do not count to CPU load to increase CPU frequency. Can you try to reproduce the bug by setting up cpufreq by hand: Set: CPUFREQ_ENABLED="no" in /etc/sysconfig/powersave/cpufreq If you reboot (without AC to try whether the problem got workarounded), check that powernow-k8 module and cpufreq_ondemand modules were not loaded. Then load them: modprobe powernow-k8 modprobe cpufreq_ondemand and activate the ondemand governor: echo ondemand >/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor Try the watch/monitor commands from above and play with AC (un)plug. Does it work now? If it works now, can you revert the sysconfig variable to: CPUFREQ_ENABLED="" again and change in /etc/sysconfig/powersave/scheme_powersave CONSIDER_NICE="yes" (I am not sure whether it's still this variable that is used for setting ignore_nice in 10.2, Seife or Holger may want to correct me here). Reboot and play again with AC (un)plug. If it now again works we probably have a nasty locking problem when writing to ignore_nice when using ondemand governor. If setting up cpufreq manually does not work, I exect a BIOS problem, as I got confirms that switching up after unplugging AC works generally (but this was not a SUSE system and he used powernowd. We need to check whether setting the ignore_nice by powersaved is the problem). -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #23 from seife@novell.com 2007-02-09 04:30 MST ------- (In reply to comment #22)
Can you try to reproduce the bug by setting up cpufreq by hand: Set: CPUFREQ_ENABLED="no" in /etc/sysconfig/powersave/cpufreq
This does nothing in 10.2, the modules are loaded by the hal init script.
If you reboot (without AC to try whether the problem got workarounded), check that powernow-k8 module and cpufreq_ondemand modules were not loaded. Then load them: modprobe powernow-k8 modprobe cpufreq_ondemand and activate the ondemand governor: echo ondemand >/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
Try the watch/monitor commands from above and play with AC (un)plug. Does it work now? If it works now, can you revert the sysconfig variable to: CPUFREQ_ENABLED="" again and change in /etc/sysconfig/powersave/scheme_powersave CONSIDER_NICE="yes" (I am not sure whether it's still this variable that is used for setting ignore_nice in 10.2, Seife or Holger may want to correct me here). Reboot and play again with AC (un)plug. If it now again works we probably have a nasty locking problem when writing to ignore_nice when using ondemand governor.
If setting up cpufreq manually does not work, I exect a BIOS problem, as I got confirms that switching up after unplugging AC works generally (but this was not a SUSE system and he used powernowd. We need to check whether setting the ignore_nice by powersaved is the problem).
Powersaved does nothing with cpufreq at all. It is all handled by HAL and the desktop applets. But there was a bug in HAL wrt. ignore_nice which has been fixed with an online update IIRC. So installing the HAL online update and testing again would be my first suggestion. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 f.alabas@ttnet.net.tr changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED Info Provider|f.alabas@ttnet.net.tr | ------- Comment #24 from f.alabas@ttnet.net.tr 2007-02-09 10:29 MST ------- Thomas, before having seen the comment #23 from Stefan, I have already followed the method you wrote me. (In reply to comment #22)
Can you try to reproduce the bug by setting up cpufreq by hand: Set: CPUFREQ_ENABLED="no" in /etc/sysconfig/powersave/cpufreq
If you reboot (without AC to try whether the problem got workarounded), check that powernow-k8 module and cpufreq_ondemand modules were not loaded. Then load them: modprobe powernow-k8 modprobe cpufreq_ondemand and activate the ondemand governor: echo ondemand >/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
Try the watch/monitor commands from above and play with AC (un)plug. Does it work now?
No, it does not work.....
If it works now, can you revert the sysconfig variable to: CPUFREQ_ENABLED="" again and change in /etc/sysconfig/powersave/scheme_powersave CONSIDER_NICE="yes" (I am not sure whether it's still this variable that is used for setting ignore_nice in 10.2, Seife or Holger may want to correct me here). Reboot and play again with AC (un)plug. If it now again works we probably have a nasty locking problem when writing to ignore_nice when using ondemand governor.
It does not work too... I have also tested my issue by disabling the powersaved from within the runlevel services and this does not help. After having disabled the powersaved service, I booted the machine on battery and still see that the highest frequency is not allowed after AC-plugged. I tried and played removing/inserting the powernow-k8, cpufreq-* modules and assigning different schemes into scaling_governor, however nothing help me to bring frequency to the highest available !! As to comment #23, I have the latest HAL package. (hal-0.5.8_git20061106-31.1) If Stefan means; - b.n.c #223517: fix logic for consider nice in CPU Freq addon (hal-fix-consider-nice-logic_223517.diff) from hmacht@suse.de which is seen in the changelog, then I have the latest fixed version of HAL. However, do you think my findings about "ignore_nice_load values" in comment #21 is as what it should be ? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 trenn@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |FIXED ------- Comment #25 from trenn@novell.com 2007-02-28 05:13 MST ------- This could already befixed in latest kod (by the HP laptop psmouse cleanup fixes). Can you pls retest with latest kernel (update for 10.2 should come out soon). And reopen if this still does not work. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 f.alabas@ttnet.net.tr changed: What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |REOPENED Resolution|FIXED | ------- Comment #26 from f.alabas@ttnet.net.tr 2007-03-01 09:46 MST ------- I have installed and tested the kotd (with psmouse-fiddle-with-reset.patch ). There is no good news, it did not work for me. I have reread all the bugs around the internet about "scaling_max_frequency", and if I do not miss some, most of them are related to Intel processors, Intel Core Duo's.... Only two report for AMD's, but still of double cores; one is AMD Turion X2 of a Fujitsu Siemens and another one for AMD Turion X2 based nx6325 but seems a cause mainly from fan. They all seem resolved with your action. I would like to speculate that my issue is an "old & covert" problem with some line of **AMD single core Turion** based HP machines and maybe that's why the patches that are originated from the similar BUT intel processor-based reports are not working for me. I made a paranoid-check for all the files in /proc/acpi, the things seem working perfect with acpi. Assuming the kernel is working ok, then what about the powernow-k8 module ? I look forward your suggestion for finding at least single-bit of a hint. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
I have installed and tested the kotd From 10.2 branch? Please also try the one from x86_64/HEAD/..., maybe things already got fixed
https://bugzilla.novell.com/show_bug.cgi?id=231107 trenn@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|REOPENED |NEEDINFO Info Provider| |f.alabas@ttnet.net.tr ------- Comment #27 from trenn@novell.com 2007-03-01 10:55 MST ------- mainline. Before doing that: You should also search for a BIOS update.., hmm I doubt it is the BIOS... Have you booted the fixed kernel twice? You need to do that: Rebooting a broken kernel (psmouse issue) will result in broken kernel on next boot (even it has the psmouse issue included). Rebooting a fixed kernel (psmouse issue) will result in fixed kernel on next boot. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 f.alabas@ttnet.net.tr changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED Info Provider|f.alabas@ttnet.net.tr | ------- Comment #28 from f.alabas@ttnet.net.tr 2007-03-02 07:48 MST ------- Today, I have retested with the kernels from: ftp://ftp.suse.com/pub/projects/kernel/kotd/10.2-x86_64/SL102_BRANCH/ http://software.opensuse.org/download/Kernel:/SL102_BRANCH/openSUSE_10.2/x86... http://software.opensuse.org/download/Kernel:/HEAD/openSUSE_10.2/x86_64/ I have booted the machine,at least twice,upon installation of the new kernel. For a BIOS issue, as common, no support from HP, however I have convinced them to make a test using Wndows to see whether the the issue is available, which is actually the only way they can release a BIOS update. However I am not sure if they really think about making this test. During my conversation with them, they referred an operating system related issue which seems similar to my report and already resolved by the mentioned operating system itself. I need to put the link here for your review. http://support.microsoft.com/kb/330512 Lastly, if there is no other option, I must kill my machine and recreate it with Wndows installation, reproduce the same issue, thus to provide a concrete proof for HP. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #29 from trenn@novell.com 2007-03-02 08:21 MST ------- Hmm, do you have ICQ? Maybe this goes a bit faster... As I saw this working on other machines and even BIOS DSDT looks fine, I currently have not much of an idea... Do you know how to compile a SUSE kernel?: rpm kernel-source.rpm cd /usr/src/linux cp arch/x86_64/defconfig.default .config # Enable ACPI_DEBUG=y and CONFIG_CPU_FREQ_DEBUG=y in .config make -j3 make install modules_install -> adjust /boot/grub/menu.lst We should communicate via ICQ or something then to and try out some more things... -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #30 from trenn@novell.com 2007-03-16 09:00 MST ------- Created an attachment (id=125033) --> (https://bugzilla.novell.com/attachment.cgi?id=125033&action=view) Patch that should fix this. Sent upstream, but no answer yet, I probably need to send again -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #31 from f.alabas@ttnet.net.tr 2007-03-20 13:37 MST ------- seems the issue doesn't get impression by the upstreamers !? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 ------- Comment #32 from f.alabas@ttnet.net.tr 2007-04-13 04:04 MST ------- Sigh...4 weeks after Thomas sent the patch......kernel-2.6.21 stable is almost released, but the patch is still not accepted/included in it. How long should it take more ? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107 trenn@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |FIXED ------- Comment #33 from trenn@novell.com 2007-04-23 07:33 MST ------- I wanted to wait for some response, I have now added a workaround that should help. This should also affect SLE10 -> also added. I posted again (on Thursday) and didn't get an answer again..., maybe Dave is holding this off for .22, I will contact him directly as soon as .21 is out and nothing happened... -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=231107#c34
--- Comment #34 from Fatih Alabas
https://bugzilla.novell.com/show_bug.cgi?id=231107#c35
--- Comment #35 from Thomas Renninger
participants (1)
-
bugzilla_noreply@novell.com