171.06 still locking up dual core with 8800GT

[bitmage]

Same symptoms as always, hard lock, no error messages.
maxcpus=1 prevents problem at cost of losing half of processing power.
Just like 169.04, 169.07, 169.09, 169.12.

Things tried: Different network card. Setting irqbalance to run in one-shot mode. Disabling irqbalance entirely. Search my previous threads for other attempts.

XP runs fine on same machine.

nvidia log attached.

[burgeon]

Originally Posted by bitmage

Same symptoms as always, hard lock, no error messages.
maxcpus=1 prevents problem at cost of losing half of processing power.
Just like 169.04, 169.07, 169.09, 169.12.

Things tried: Different network card. Setting irqbalance to run in one-shot mode. Disabling irqbalance entirely. Search my previous threads for other attempts.

XP runs fine on same machine.

nvidia log attached.

I have exactly the same problem. The only working option is maxncpus=1 for me. I doubt if this can be really an option that justifies running linux on multiple cores.

[chunkey]

Originally Posted by bitmage

Same symptoms as always, hard lock, no error messages.
maxcpus=1 prevents problem at cost of losing half of processing power.
Just like 169.04, 169.07, 169.09, 169.12.

nvidia log attached.

The log doesn't contain any valuable infomation about the real causes of the crashes...
But it seems that you have something on the serial port?

/proc/interrupts
           CPU0       CPU1       
  0:     650412          0   IO-APIC-edge      timer
  1:          2          0   IO-APIC-edge      i8042
  4:        230       6002   IO-APIC-edge      serial
[...]

Is this a serial console, or something like that? Because you can catch very useful Oops or the panic log with it!


Howto enable the serial console (but you'll need another PC with a terminal!)

[bitmage]

Originally Posted by chunkey

Is this a serial console, or something like that? Because you can catch very useful Oops or the panic log with it!

It's the monitoring connection to my UPS, but that was an excellent idea. I disabled the UPS monitoring software, and set up the serial console with another machine to log the messages.

I removed maxcpus=1, rebooted, and went to bed. In the morning my machine was locked up tight. Alas, nothing unusual was logged to the serial console. You just see normal log messages (smartd, etc)...and then silence until the messages from my reboot start appearing.

Whatever this crash is, it halts the CPU so fast it doesn't even have time to scream.

[chunkey]

Originally Posted by bitmage

I removed maxcpus=1, rebooted, and went to bed. In the morning my machine was locked up tight. Alas, nothing unusual was logged to the serial console. You just see normal log messages (smartd, etc)...and then silence until the messages from my reboot start appearing.

Whatever this crash is, it halts the CPU so fast it doesn't even have time to scream.

what was your baud-rate? The 9600bps is too slow nowadays... you should try 115200 !
BTW, when your kernel freezes can you please check if your keyboard LEDs are flashing?

[bitmage]

Had it at 14400. No flashing of the keyboard LEDs, just NumLock stuck on. I don't think the kernel is having a chance to panic.

[chunkey]

Originally Posted by bitmage

Had it at 14400. No flashing of the keyboard LEDs, just NumLock stuck on. I don't think the kernel is having a chance to panic.

oh? no LEDs... Well, then the next step would be a watchdog-card or firewire (ohci1394 can read/write to the system memory over DMA) but I have no idea how to do that...

[mlauss]

Originally Posted by chunkey

firewire (ohci1394 can read/write to the system memory over DMA) but I have no idea how to do that...

Firescope:

http://freshmeat.net/projects/firesc...ease_id=273725