Bug 4392

Summary: powernow-k8 driver doesn't work with Socket 754 AMD Sempron 3100+
Product: Power Management Reporter: Artem S. Tashkinov (aros)
Component: cpufreqAssignee: cpufreq (cpufreq)
Status: CLOSED PATCH_ALREADY_AVAILABLE    
Severity: blocking CC: email, perryw
Priority: P2    
Hardware: i386   
OS: Linux   
Kernel Version: 2.6.11 Subsystem:
Regression: --- Bisected commit-id:
Attachments: This patch solved my problems

Description Artem S. Tashkinov 2005-03-23 05:49:13 UTC
Distribution: cpufreq powernow-k8 modules doesn't work with Sempron 3100 CPU
correctly.

Hardware Environment:
CPU: AMD Sempron 3100+ Socket 745
MB: Gigabyte K8NS Pro
RAM: 512Mb PC3200 Hynix
PowerSupply: 350Watts Unknown

Software Environment: Fedora Core 1 partly upgaded to FC4 test1
Problem Description:

When loading the powernow-k8 module on any recent kernel (I've tried
2.6.8 through 2.6.10) on my Sempron desktop, I see:
powernow-k8: Found 1 AMD Athlon 64 / Opteron processors (version 1.00.09e)
powernow-k8:    0 : fid 0x8 (1800 MHz), vid 0x6 (1400 mV)
powernow-k8:    1 : fid 0x0 (900 MHz), vid 0x18 (950 mV)
cpu_init done, current fid 0x8, vid 0x4
powernow-k8: vid trans failed, vid 0x3, curr 0x4
powernow-k8: transition frequency failed
powernow-k8: vid trans failed, vid 0x3, curr 0x4
powernow-k8: transition frequency failed

... and many more of those 'vid trans failed' errors (one every time
there's an attempted freq change).

Actually as soon as my PC is idle the CPU speed is set to 900MHz and
cpufreq_ondemand can't change it to smth else. I tried using cpufreq_userspace
driver with the same effect: after userspace cpufreq daemon changed CPU speed to
900MHz it couldn't restore it to 1800MHz on high load.

Note: 
Original AMD cpu driver supplied for Windows XP works OK.
Comment 1 Artem S. Tashkinov 2005-05-01 21:13:44 UTC
Created attachment 5007 [details]
This patch solved my problems

This patch resolves the issues while following closely AMD specifications.
Comment 2 Perry Wolf 2005-05-14 22:05:56 UTC
I was having a similar problem on my Winchester 3000+ on a Gigabyte K8N Ultra-9
(nForce4 Ultra chipset).  The patch fixed my problem on gentoo-sources
2.6.11-r7.  The same problem existed on vanilla 2.6.12-rc4.
Comment 3 Steffen Weber 2005-05-21 11:25:55 UTC
I got the same error messages as the original poster on an Athlon 64 3500+
(Venice) @ Gigabyte K8N-Ultra SLI (So939 nForce 4) using
gentoo-sources-2.6.11-r9 and this patch fixed them for me, too. :-)
Comment 4 Andrew Morton 2005-05-25 17:14:01 UTC
Can I please have an update on this one?  Was the patch merged?  Is 2.6.12-rc5
functioning correctly?

Thanks.
Comment 5 Perry Wolf 2005-05-25 17:52:37 UTC
I still see the same problem in 2.6.12-rc5.  The patch still corrects the
problem for me as well.
Comment 6 Artem S. Tashkinov 2005-05-26 05:09:36 UTC
This patch is not included into 2.6.12-rc5.
Comment 7 Perry Wolf 2005-06-18 23:49:27 UTC
I just did a quick test of 2.6.12.  Seems that the problem is fixed on my system.
Comment 8 Steffen Weber 2005-06-19 02:07:12 UTC
For me it is fixed in 2.6.12, too. :-)  
Comment 9 Artem S. Tashkinov 2005-06-20 21:15:26 UTC
As for 2.6.12 the bug is no longer valid. Another patch has obviously fixed this
problem.