Hyperthreading problems with 2.6.7

Submitted by Anonymous
on August 21, 2004 - 9:19am

Hey there,

I just compiled a 2.6.7 kernel on Debian Sarge, and actually everything works fine, besides one very concerning issue:
I'm running an Intel P4 3.0GHz with Hyperthreading, so I activated HT support in the kernel config. However, (at least under X), the second CPU is *always* under full load, resulting in a pretty laggy system.

The dmesg output regarding the two virtual CPUs is as follows:

anthrax:/usr/src/linux# dmesg | grep CPU
Initializing CPU#0
CPU: After generic identify, caps: bfebfbff 00000000 00000000 00000000
CPU: After vendor identify, caps: bfebfbff 00000000 00000000 00000000
CPU: Trace cache: 12K uops, L1 D cache: 8K
CPU: L2 cache: 512K
CPU: Physical Processor ID: 0
CPU: After all inits, caps: bfebfbff 00000000 00000000 00000080
Intel machine check reporting enabled on CPU#0.
CPU0: Intel P4/Xeon Extended MCE MSRs (12) available
CPU0: Thermal monitoring enabled
CPU0: Intel(R) Pentium(R) 4 CPU 3.00GHz stepping 09
per-CPU timeslice cutoff: 1462.98 usecs.
enabled ExtINT on CPU#0
Initializing CPU#1
masked ExtINT on CPU#1
CPU: After generic identify, caps: bfebfbff 00000000 00000000 00000000
CPU: After vendor identify, caps: bfebfbff 00000000 00000000 00000000
CPU: Trace cache: 12K uops, L1 D cache: 8K
CPU: L2 cache: 512K
CPU: Physical Processor ID: 0
CPU: After all inits, caps: bfebfbff 00000000 00000000 00000080
Intel machine check reporting enabled on CPU#1.
CPU1: Intel P4/Xeon Extended MCE MSRs (12) available
CPU1: Thermal monitoring enabled
CPU1: Intel(R) Pentium(R) 4 CPU 3.00GHz stepping 09
..... CPU clock speed is 2998.0170 MHz.
checking TSC synchronization across 2 CPUs: passed.
Brought up 2 CPUs
CPU0: online
CPU1: online
ACPI: Processor [CPU1] (supports C1)
ACPI: Processor [CPU2] (supports C1)

I'm not an expert, but in my eyes this looks OK. I'm pretty desparate on what could be causing this issue.

Can you help me here?
Thanks in advance.

hmm

Anonymous
on
August 21, 2004 - 10:04am

Hmm, it's OK, both "virtual" cpus are there and they're working...
But back to business. I'd say (only if an older, newer 2.6 kernel work and you've nothing changed) that the scheduler can be broken, or it can be a problem with the display-drivers(ati/nvidia closed source drivers)...

I'm currently using the 'ati'

Anonymous
on
August 21, 2004 - 11:00am

I'm currently using the 'ati' driver module shipped with sarge, though I will update to the proprietary ATI driver to get hardware accalerated 3D (the DRI is not working with the debian driver).

I'm not 100% sure, but I doubt that it's a bug or that something is really broken. I searched the web but couldn't find anybody else having this problem. I guess I just screwed something up in the kernel config.

Any other ideas what could lead to this issue?

hmmm Part II

Anonymous
on
August 21, 2004 - 11:11am

Well I don't know what's causing that lacks, maybe if you find out
which progress eat the CPU, we've at least a hint. Or did you ever check an older/newer 2.6 kernel with the same config, or what happend if you kill X and ran "top" in a console, or disable HT?

(Well your problems seems to be a interrupt problem, but I'm not sure...)

I've a 2 * 667 Pentium III Box, running a vanilla 2.6.7 Kernel.

First of all, it's the first

Anonymous
on
August 21, 2004 - 11:42am

First of all, it's the first 2.6 kernel I've ever used. I don't know if other versions have the same problem. I just ran the KDE system monitor, which showed both CPUs at 100% idle load, while the sysmeter from blackbox only shows the second CPU under full load (frankly I'm not sure what exactly it's referring to).
More surprisingly, none of the processes I'm currently running uses more than a few percent processor power. Top says:
Cpu(s): 0.2% us, 0.2% sy, 0.0% ni, 99.7% id,
I guess 'id' means idle.

Little tip for top.

Anonymous
on
August 21, 2004 - 12:00pm

Press "1" in top, show's each CPU workload, that's very helpful...

me too

Anonymous
on
October 21, 2004 - 8:03am

I have a very similar problem, except that my cpu is not always under full load, yet the system is VERY slow. Booting takes a LONG time. I am running 2.6.8-1-686-smp on a 3GHz PIV with HT enabled. The boot takes around ten minutes. I installed 2.6.8-1-386 and amazingly enough the system screams. At the moment I have no clue as to what is causing this. I am running 2.6.8-1-686-smp and 2.6.8-1-686 on several other boxes without the slowness. The only difference is the HT. I am going to try and disable it and see if my problem is realted to that. If anyone has any ideas I would greatly appreciate it.

i865PE - ASUS motherboard problem?

Anonymous
on
December 7, 2004 - 5:31am

We had the same problem with an ASUS P4P800 motherboard.
It seems there is some problem with the MEMORY handling, because with two memory modules it is SLOOOOW. But when we pulled out one ot the two memory modules leaving only one in bank A1 the system worked fine.

getting closer to the origin

Anonymous
on
December 9, 2004 - 1:45pm

It's not the problem with the memory modules. It slows down above 512MB of main memory, regardless of how many modules forms it.
It seems that the on-board video controller and the kerner conflicts. (Maybe the kernel uses the in-mapped slow video memory for stack???) Anyway, to work around the problem simply say to the kernel not to use the last 64MB of memory or so. For example give the kernel parameter "mem=960M" at boot time if you have 1G of main memory.
(W)

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.