Bug 17712

Summary: lm_sensors returns wrong CPU temperature values with k8temp
Product: Drivers Reporter: mkkot (marcin2006)
Component: Hardware MonitoringAssignee: Jean Delvare (jdelvare)
Status: CLOSED CODE_FIX    
Severity: low CC: jdelvare, maciej.rutecki, rjw
Priority: P1    
Hardware: All   
OS: Linux   
Kernel Version: 2.6.35.4 Tree: Mainline
Regression: Yes
Bug Depends on:    
Bug Blocks: 16055    
Attachments: dmesg

Description mkkot 2010-09-03 11:07:10 UTC
Created attachment 28942 [details]
dmesg

I kindly ask for indulgence as this is my first bugreport here.

I have a Gigabyte M61P-S3 motherboard with an AMD CPU installed:

model name      : AMD Athlon(tm) 64 X2 Dual Core Processor 5600+

I use cpu scaling feature if this matters (powernow-k8 cpufreq_ondemand). After recent kernel update (using Archlinux 64 bit packages):

[2010-09-01 11:01] upgraded kernel26 (2.6.34.1-1 -> 2.6.35.4-1)

my plasmoid which measures CPU core temperatures started to display very lowered values. After some digging I found it is probably using lm_sensors, which reports bad values as well:

[root@linux mk]# sensors
k8temp-pci-00c3
Adapter: PCI adapter
Core0 Temp:   +5.0°C                                    
Core0 Temp:   +6.0°C                                    
Core1 Temp:   +7.0°C                                    
Core1 Temp:   +2.0°C

[root@linux mk]# 

I should have there at least 27°C.

I use a standard often updated Archlinux system with no stange configuration or devices attached. Probably a dmesg output can tell you more.
Comment 1 Jean Delvare 2010-09-05 13:44:08 UTC
I believe this is caused by commit d535bad90dad4eb42ec6528043fcfb53627d4f89 (hwmon: (k8temp) Fix temperature reporting for ASB1 processor revisions), which affected more CPU models than was originally intended. Should be fixed by commit a05e93f3b3fc2f53c1d0de3b17019e207c482349 (hwmon: (k8temp) Differentiate between AM2 and ASB1) which limits the scope of the former commit to ASB1 models (I guess you have an AM2 model.)

The latter commit was already accepted by Greg KH for 2.6.32-stable and 2.6.35-stable, so you will get the fix when kernel 2.6.35.5 is released and your distribution upgrades to it.
Comment 2 Rafael J. Wysocki 2010-09-05 22:05:13 UTC
Fixed in the mainline, so closing.
Comment 3 mkkot 2010-09-25 11:43:17 UTC
Thanks, it works in 2.6.35.5.