PROBLEM: BUG: Constant freezes and kernel panics on a quad core (with dumps)
Bruno Barberi Gnecco
situations, including under no load (not even the X running). Dumps are provided below.
Sometimes it's a plain freeze, sometimes a kernel panic with the blinking leds and a dump.
It seems the problem is related to memory, most of the messages are related to null memory
references (see more below). System ran once for several hours before freezing.
* System specification: Intel Core 2 Quad Q9550 @ 2.83GHz, 4GB DDR3, Motherboard Asus
P5QC. No overclocking.
* Tested under the following kernels/distros, all had the bug. Since I used their vanilla
kernels, I'm not posting a copy of the .configs:
- 2.6.27.7-smp (slackware 12.2 hugesmp kernel; distro heavily updated), 32 bits
- 2.6.29.6-smp (slackware 13.0 hugesmp kernel; fresh install), both 32 and 64 bits
- 2.6.29.6-nosmp (slackware 13.0 huge kernel; fresh install), both 32 and 64 bits
- 2.6.31-14-generic SMP (ubuntu 9.10 netbook remix live cd), 32 bits
* Here are some screen dumps (sorry, no serial cable for a text dump).
http://ultraxs.com/share-208E_4B118ED2.html
http://ultraxs.com/share-2FDD_4B118ED2.html
http://ultraxs.com/share-D0D9_4B118ED2.html
http://ultraxs.com/share-498F_4B118ED2.html
http://ultraxs.com/share-0AD8_4B118ED2.html
* Turning off ACPI and APIC, either through the BIOS or through kernel parameters did not
help, even when only one processor is recognized.
* I ran memtest and no errors were detected. I completed the tests three times.
* Nothing is ever output to the logs.
* A sure way to get a freeze is to recompile the kernel, but the freeze does not happen at
the same point of the compilation and I can't even say if the compilation triggers the bug
or if it's just a coincidence and the compilation just takes long enough for it to happen.
* Any other information, testing or debugging you need, just ask. I just can't use the PC,
it usually freezes within minutes. Please CC any replies to me [brunobg at gmail]. Thank
you very much for any help.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majo...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Michael Breuer
similar grief in the past:
Bad IDE cable
Bad SATA cable
Bad power supply
Bad motherboard
Bad memory (memtest doesn't necessarily access things the same way as
the kernel)
Bad cards (pci, agp, whatever)
Any of the above with loose connections
And did I mention bad power supply?
Bruno Barberi Gnecco
> similar grief in the past:
It is possible, but I can rule most of them out. I think it's unlikely.
> Bad IDE cable
> Bad SATA cable
Ruled out. Only SATA drives. I used two different HDs (separatedly) with different
cables, and the live CD. It's not the cables or the drives.
> Bad power supply
> Bad motherboard
Both are brand new, but of course that doesn't rule them out.
Regarding the PS, I have checked voltages with a multimeter and they are more than fine,
and the wattage is enough for the system, so it'd have to be a very weird transient glitch
that affects only memory access. See also below.
Any ideas to rule the MB out, other than "get a new one"?
> Bad memory (memtest doesn't necessarily access things the same way as
> the kernel)
Ruled out. I replaced with a 2GB DDR2, still got the bug: "BUG: Bad page map in process".
> Bad cards (pci, agp, whatever)
Ruled out. The only card is the video card. I replaced it with a very old PCI board and
still got error. This also pretty much rules out that the PS is underpowered, since I
powered only the MB and the HD.
Could it be one of the onboard things? I disabled everything but the LAN, and still got it.
> Any of the above with loose connections
I already reconnected everything twice. Could still be a loose connection of one of the
wires in the connector, but it's very very unlikely to give such a specific error on
memory access.
> And did I mention bad power supply?
Yes you did, and I'll try to get another one to be sure, but it could still be a software
bug too.
I tried to install an old Win2K I had here. It doesn't handle the big HD well and ends up
not booting after installation (it can't partition it, also), but it didn't freeze (even
though it formatted the disk and copied the files, which was not slow). So +1 to being a
kernel bug.
The most common error message I get is: "BUG: unable to handle kernel NULL pointer
dereference."
Any other suggestions? Any other information which could help?
Thanks for the answer,
--
Bruno Barberi Gnecco <brunobg_at_users.sourceforge.net>
Robert Hancock
>> I'd think this is a hardware problem. Some things that have caused me
>> similar grief in the past:
>
> It is possible, but I can rule most of them out. I think it's unlikely.
>
>> Bad IDE cable
>> Bad SATA cable
>
> Ruled out. Only SATA drives. I used two different HDs (separatedly) with
> different cables, and the live CD. It's not the cables or the drives.
>
>> Bad power supply
>> Bad motherboard
>
> Both are brand new, but of course that doesn't rule them out.
Checked CPU temperature under load?
>
> Regarding the PS, I have checked voltages with a multimeter and they are
> more than fine, and the wattage is enough for the system, so it'd have
> to be a very weird transient glitch that affects only memory access. See
> also below.
Most of the time transients will be the issue when a power supply causes
problems and that can't be seen with a normal voltmeter. It's not
typical for the rails to be low all the time unless the power supply is
heavily overloaded.
>
> Any ideas to rule the MB out, other than "get a new one"?
>
>> Bad memory (memtest doesn't necessarily access things the same way as
>> the kernel)
>
> Ruled out. I replaced with a 2GB DDR2, still got the bug: "BUG: Bad page
> map in process".
>
>> Bad cards (pci, agp, whatever)
>
> Ruled out. The only card is the video card. I replaced it with a very
> old PCI board and still got error. This also pretty much rules out that
> the PS is underpowered, since I powered only the MB and the HD.
>
> Could it be one of the onboard things? I disabled everything but the
> LAN, and still got it.
>
>> Any of the above with loose connections
>
> I already reconnected everything twice. Could still be a loose
> connection of one of the wires in the connector, but it's very very
> unlikely to give such a specific error on memory access.
>
>> And did I mention bad power supply?
>
> Yes you did, and I'll try to get another one to be sure, but it could
> still be a software bug too.
>
> I tried to install an old Win2K I had here. It doesn't handle the big HD
> well and ends up not booting after installation (it can't partition it,
> also), but it didn't freeze (even though it formatted the disk and
> copied the files, which was not slow). So +1 to being a kernel bug.
>
> The most common error message I get is: "BUG: unable to handle kernel
> NULL pointer dereference."
If you're getting random crashes in different places then it usually is
some kind of hardware problem, unless there's some kind of random memory
corruption going on, but that seems a bit unlikely.
Bruno Barberi Gnecco
>>> Regarding the PS, I have checked voltages with a multimeter and they are
>>> more than fine, and the wattage is enough for the system, so it'd have
>>> to be a very weird transient glitch that affects only memory access. See
>>> also below.
>> Most of the time transients will be the issue when a power supply causes
>> problems and that can't be seen with a normal voltmeter. It's not
>> typical for the rails to be low all the time unless the power supply is
>> heavily overloaded.
>
>
> You can't check any PSU with any multimeter I've ever seen unless it's a
> catastrophic failure, or as you said, so overloaded that it can't
> regulate (in which case it would have shut down if it were decent
> quality...). Non-catastophic PSU failures are often filter problems
> that a multimeter isn't fast enough to see. Many switchers are
> deplorably noisy, and rely on the caps at the end of the transmission
> line, so one poor quality or dried out cap on MB can screw the pooch
> too.
>
>>> Any ideas to rule the MB out, other than "get a new one"?
>>>
>>>> Bad memory (memtest doesn't necessarily access things the same way as
>>>> the kernel)
>>> Ruled out. I replaced with a 2GB DDR2, still got the bug: "BUG: Bad page
>>> map in process".
>>>
>>>> Bad cards (pci, agp, whatever)
>>> Ruled out. The only card is the video card. I replaced it with a very
>>> old PCI board and still got error. This also pretty much rules out that
>>> the PS is underpowered, since I powered only the MB and the HD.
>>>
>>> Could it be one of the onboard things? I disabled everything but the
>>> LAN, and still got it.
>>>
>>>> Any of the above with loose connections
>
> connectors with vibration. Pull ram, and reseat. Resist the urge to
> clean any connector with anything other than no-residue contact cleaner.
>
> Another thing to watch out for is crappy heat sink compound. That dries
> out, doesn't conduct heat well enough. Under load, such a problem may
> build VERY fast with modern CPU current draw. If all else fails, pull
> your CPU heatsink, clean and re-apply fresh compound.
>
>>> I already reconnected everything twice. Could still be a loose
>>> connection of one of the wires in the connector, but it's very very
>>> unlikely to give such a specific error on memory access.
>>>
>>>> And did I mention bad power supply?
>>> Yes you did, and I'll try to get another one to be sure, but it could
>>> still be a software bug too.
>
> with everything from PC hardware to very high dollar HW. It's the point
> of maximum electrical stress. It's also a spot where many people try to
> save money... big mistake that.
>
> (removes HW guy hat;)
Follow-up, with thanks to everybody who helped: I tried a different PSU and still got the
problem, and I also got a BSOD with Windows. So it seems to be a problem with the
motherboard or the processor.
Thanks a lot again,