Bug 879847

Summary: Only 2 cores of AMD Phenom 945 quad-core activated.
Product: [Fedora] Fedora Reporter: Ralph Loader <suckfish>
Component: kernelAssignee: Kernel Maintainer List <kernel-maint>
Status: CLOSED CANTFIX QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: medium Docs Contact:
Priority: unspecified    
Version: 18CC: gansalmon, itamar, jacob.shin, jonathan, kernel-maint, madhu.chinakonda
Target Milestone: ---   
Target Release: ---   
Hardware: x86_64   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2012-11-28 23:43:37 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
Output of acpidump
none
dmesg output
none
lspci.txt
none
/proc/cpuinfo none

Description Ralph Loader 2012-11-24 20:13:03 UTC 
User-Agent:       Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/17.0 Firefox/17.0
Build Identifier: 

With 3.6.7-5.fc18.x86_64, the kernel is only starting 2 of 4 cores:

$ egrep 'processor|model name' /proc/cpuinfo 
processor	: 0
model name	: AMD Phenom(tm) II X4 945 Processor
processor	: 1
model name	: AMD Phenom(tm) II X4 945 Processor

In the past, the kernel was correctly starting all 4 cores.

I'll attach acpidump output.

This kernel boot log message might or might not be relevant:

Nov 25 07:53:48 i kernel: setup_percpu: NR_CPUS:128 nr_cpumask_bits:128 nr_cpu_ids:4 nr_node_ids:1


Reproducible: Always

Steps to Reproduce:
1.  Boot fedora 18.
2.  cat /proc/cpuinfo
Actual Results:  
2 cores are running

Expected Results:  
4 cores should be running
Comment 1 Ralph Loader 2012-11-24 20:15:06 UTC 
Created attachment 651220 [details]
Output of acpidump

Output of acpidump
Comment 2 Dave Jones 2012-11-27 19:46:30 UTC 
can you attach the output of dmesg please.
Comment 3 Ralph Loader 2012-11-27 23:52:34 UTC 
Created attachment 653203 [details]
dmesg output

First 1k lines of dmesg output (I'm assuming post-boot messages are irrelevant & have trimmed the file.  Let me know if my assumption is wrong).
Comment 4 Jacob Shin 2012-11-28 05:07:59 UTC 
Hi,

Ralph,

Could you also attach:

1. The full output of:
   $ cat /proc/cpuinfo

2. And as root, the output of:
   $ lspci -s 0:18.* -xxxx

please ?

Thanks!
Comment 5 Jacob Shin 2012-11-28 05:12:23 UTC 
Oh, and I also forgot to ask ..

Could you go through your BIOS options and make sure that there aren't any obvious things like "downcore" or "disable core" that is set? Just in case.

Thanks!
Comment 6 Ralph Loader 2012-11-28 06:27:37 UTC 
Created attachment 653296 [details]
lspci.txt

Output of lspci -s 0:18.* -xxxx
Comment 7 Ralph Loader 2012-11-28 06:28:45 UTC 
Created attachment 653309 [details]
/proc/cpuinfo
Comment 8 Ralph Loader 2012-11-28 06:58:37 UTC 
Rebooted yet again to look through BIOS options.  Did not find any relevant.

Rebooted in Linux.

All 4 cores back.

WTF!  Your guess is probably better than mine.

Unless the information attached above contains a smoking gun, feel free to close as 'cannot reproduce'... if it reoccurs, I'll reopen...
Comment 9 Jacob Shin 2012-11-28 17:56:02 UTC 
Weird..

Yes, in one of the CPU PCI configuration registers:

> 00:18.3 Host bridge: Advanced Micro Devices [AMD] Family 10h Processor Miscellaneous Control
...
> 190: 0c 00 00 00 00 00 00 00 00 04 78 00 00 00 00 00
       ^^

Bit 2 and 3 are set, which disables core 2 and 3. This is (was?) set by the BIOS.

If you see all 4 cores now, run the lspci command again, and see if those bits are clear now .. you should see 00.

Not really sure why it was set and now it's clear, and why your BIOS decided to change it's mind after reboot.

Thanks!

-Jacob
Comment 10 Ralph Loader 2012-11-28 23:43:37 UTC 
Ok, the evidence is that this is nothing to do with fedora and everything to do with some buggy lump of closed source bios.

Closing.