You are not logged in.
Pages: 1
Hi
After upgrading from the stock 2.6.11->2.6.12 the powernow-k7 (or powernow-k8) module no longer recognizes my CPU (mobile athlon). It produces a "no such device". Previously with 2.6.11 everything worked ok. I've seen people complaining about this issue, but has anyone got a solution?
Michal Stanislawski
Offline
try modprobe --force it.
iphitus
Offline
It didn't help. Still the "no such device" message appears.
I'm not sure if it's of any importance but with kernel 2.6.11 "dmesg | grep power" resulted in:
ACPI: CPU0 (power states: C1[C1] C2[C2])
powernow: PowerNOW! Technology present. Can scale: frequency and voltage.
powernow: SGTC: 13333
powernow: Minimum speed 1396 MHz. Maximum speed 1994 MHz.
while with kernel 2.6.12:
ACPI: CPU0 (power states: C1[C1] C2[C2])
powernow: PowerNOW! Technology present. Can scale: frequency and voltage.
additionally in 2.6.12 with cpufreq debug on i get the following when trying to modprobe powernow-k7:
cpufreq-core: trying to register driver powernow-k7
cpufreq-core: adding CPU 0
cpufreq-core: initialization failed
cpufreq-core: no CPU initialized for driver powernow-k7
cpufreq-core: unregistering CPU 0
Thanks for your help
Michal Stanislawski
Offline
Ok... it took me a whole day but it works!
Problem:
Arch 2.6 stock kernel has SMP enabled by default -> by tracing it to the source it seems that the recalibrate_cpu_khz() function found in arch/i386/kernel/timers/timer_tsc.c only works for non-SMP systems with 2.6.12 kernels. The only problem is that in 2.6.11 kernels this was not an issue.
Solution:
Build a custom kernel and disable the SMP option
Best Regards
Michal Stanislawski
Offline
Cheers, been trying to sort that out for ages... just need some time to do a recompile now.
Offline
Happy I could help. It's strange it's not stated in the wiki since everyone with mobile athlons and stock kernels surly have this problem.
Michal Stanislawski
Offline
thank's for information
Offline
Ok - we can disable SMP in kernel26cko but I'll need someone to explain what difference that might make to the intersted parties
Offline
Symmetric multi-processing support SMP
This enables support for systems with more than one CPU. If you have
a system with only one CPU, like most personal computers, say N. If
you have a system with more than one CPU, say Y.
If you say N here, the kernel will run on single and multiprocessor
machines, but will use only one CPU of a multiprocessor machine. If
you say Y here, the kernel will run on many, but not all,
singleprocessor machines. On a singleprocessor machine, the kernel
will run faster if you say N here [...]
--------------------
So I guess disabling SMP is a good idea and shouldn't break anything (apart from the fact that SMP systems will use only one CPU ) In my opinion most people still use singleprocessor machines (or am I wrong?). And if they do use a SMP system... building a custom kernel on such a powerfull machine takes seconds (yeah, I know... until they have this SMP kernel they only have one processor so it's not as powerful as one might thing but that's just details)
Michal Stanislawski
Offline
I can't find out of the stock kernel olny recently moved to SMP support - can anyone confirm how long the stock kernel has been using SMP - why is this suddently and issue?
Offline
I know it's been there for at least 5 or 6 months, but not sure how long before that. I like SMP being enabled and many, if not most of the major distros now have it enabled by default.
Just my two cents...
oz
Offline
The problem is not with SMP itself but with changes to the powernow module.
Michal Stanislawski
Offline
OK - it's not going to be disabled in kernel26cko - i was directed here from another thread by the way, presumably to consider the case for disabling it. case closed
Offline
so? laptop users doomed?
Offline
SMP allows the kernel to use more than one processor on a system, this also includes those with hyperthreading and dual core. If there is no SMP support compiled in, it will boot and run, but only using one cpu.
Most other distros package seperate SMP and non SMP kernels.
iphitus
Offline
Rubinas, you only have to compile a custom kernel and disable SMP. Besides only mobile athlon users are affected by this AFAIK.
Michal Stanislawski
Offline
geez, i have recompiled kernel, ant it works well.
in 26.12 brach, i noticed what freq_table are'nt going crazy, then
turn on or off power. it looks promising.
but still i'am dissapointed :S
what i have spend 30 min on kernel compiling :S
BTW i think what laptop users are more then dual core or multiple processors users
and as mention previuos post, should be one more kernel special for this ocassions... multi and laptop kernels..
zzzzz...
Offline
geez, i have recompiled kernel, ant it works well.
in 26.12 brach, i noticed what freq_table are'nt going crazy, then
turn on or off power. it looks promising.
but still i'am dissapointed :S
what i have spend 30 min on kernel compiling :SBTW i think what laptop users are more then dual core or multiple processors users
and as mention previuos post, should be one more kernel special for this ocassions... multi and laptop kernels..
zzzzz...
I would actually think there are more dual core/smp users than there are users who need the powernow-k7 module.
Either way, has someone filed a bug for this? With the suggestion of having 2 kernels?
Offline
Nah I reckon there are more people with Athlon's who at least would like the option of frequency scaling than with dual procs.
Plus, if you're messing about with >1 cpu, you probably don't mind recompiling. This is all moot anyway if you use gcc 4 and nvidia, as you have to recompile anyway
Offline
Nah I reckon there are more people with Athlon's who at least would like the option of frequency scaling than with dual procs.
Plus, if you're messing about with >1 cpu, you probably don't mind recompiling. This is all moot anyway if you use gcc 4 and nvidia, as you have to recompile anyway
I donno, I think theres more pentium 4s out there with hyperthreading then there are Athlon Mobiles. The higher end Pentium 4s are now dual core too. So no, just becuase someone has a dual core processor, doesnt mean they are someone who doesnt mind recompiling.
Besides, if you want it to change, file a bug, otherwise, I doubt it'll be changing in the near future.
Offline
I have to agree with iphitus. The market is flooded with P4s, so disabling SMP in the kernel would be a big step backwards at this point in time. However, the option of a separate kernel with SMP disabled does sound like a good alternative if the maintainer wants to do that.
oz
Offline
Pages: 1