[9fans] smp issues with dual core cpu? - Plan9

This is a discussion on [9fans] smp issues with dual core cpu? - Plan9 ; hello i have a dual core 2 (intel dual core 2 6600) cpu, when booting without *nomp= the kernel prints both cpu cores but get hanged randomly, sometimes it runs for a while and i can run commands, sometimes it ...

+ Reply to Thread
Results 1 to 6 of 6

Thread: [9fans] smp issues with dual core cpu?

  1. [9fans] smp issues with dual core cpu?

    hello

    i have a dual core 2 (intel dual core 2 6600) cpu, when booting
    without *nomp= the kernel prints both cpu cores but get hanged
    randomly, sometimes it runs for a while and i can run commands,
    sometimes it doesn't start the shell. Ctrl+tt keystrokes work.

    i have disabled usb, network card (rtl), serial and parallel ports,
    audio, and all the devices the bios let me disable, but the problem is
    still here.

    timesync is not running on the machine, and cron seems to behave correctly.

    when using one core it work as expected.

    any hints about what could be happening? i have tested plan9 time ago
    on dual core 2 machines and worked fine with both cores enabled.

    the booting messages seems normal (the same as when booting with a
    single core enabled but the cpu1: line) i didn't run diff on them. . .

    thanks

    gabi

  2. Re: [9fans] smp issues with dual core cpu?

    I have this same problem on a dual cpu machine. Things run for a while
    and then hang. In rio, if a window hangs, opening a new window, if rio
    itself hasn't already hung, will give me a working shell that will
    eventually hang but works for a while. Putting *nomp= back in gives me
    a working system.

    On 9/5/07, Gabriel Diaz wrote:
    > hello
    >
    > i have a dual core 2 (intel dual core 2 6600) cpu, when booting
    > without *nomp= the kernel prints both cpu cores but get hanged
    > randomly, sometimes it runs for a while and i can run commands,
    > sometimes it doesn't start the shell. Ctrl+tt keystrokes work.
    >
    > i have disabled usb, network card (rtl), serial and parallel ports,
    > audio, and all the devices the bios let me disable, but the problem is
    > still here.
    >
    > timesync is not running on the machine, and cron seems to behave correctly.
    >
    > when using one core it work as expected.
    >
    > any hints about what could be happening? i have tested plan9 time ago
    > on dual core 2 machines and worked fine with both cores enabled.
    >
    > the booting messages seems normal (the same as when booting with a
    > single core enabled but the cpu1: line) i didn't run diff on them. . .
    >
    > thanks
    >
    > gabi
    >


  3. Re: [9fans] smp issues with dual core cpu?

    Vesa + mp = trouble.

    Do you have a native driver for your graphic card?.

    --
    - curiosity sKilled the cat

  4. Re: [9fans] smp issues with dual core cpu?

    hello

    you're right, i can use nvidia with hwaccel off, now seems that all works.

    thanks

    gabi


    On 9/6/07, Gorka Guardiola wrote:
    > Vesa + mp = trouble.
    >
    > Do you have a native driver for your graphic card?.
    >
    > --
    > - curiosity sKilled the cat
    >


  5. Re: [9fans] smp issues with dual core cpu?

    On 9/6/07, Gorka Guardiola wrote:
    > Vesa + mp = trouble.
    >
    > Do you have a native driver for your graphic card?.


    now that's interesting!

    So there is more than one process doing a call to the VESA bios? is
    that the issue? Or is it that rio is moving from one cpu to another
    and calling vesa?

    If the latter, could you just wire rio down to one cpu?

    thanks

    ron

  6. Re: [9fans] smp issues with dual core cpu?

    On 9/6/07, ron minnich wrote:
    > now that's interesting!
    >
    > So there is more than one process doing a call to the VESA bios? is
    > that the issue? Or is it that rio is moving from one cpu to another
    > and calling vesa?
    >
    > If the latter, could you just wire rio down to one cpu?
    >
    > thanks
    >


    I do not fully understand the reason, but it has to do with the way
    interrupts are routed in mp (jmk explained this to me). When you do a vesa call
    the processor goes to 16 bit mode to execute the Rom functions. Apparently
    you can get an interrupt routed to you when you are in this mode and then
    bad things happen and there is no way to stop this. As I said, I donīt fully
    undertand why or how, ask jmk.

    --
    - curiosity sKilled the cat

+ Reply to Thread