lkml.org 
[lkml]   [2004]   [Oct]   [15]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    Patch in this message
    /
    Date
    From
    Subject[patch, 2.6.9-rc4-mm1, 2.6.9-rc4] fix the prof=schedule feature

    * Andrew Morton <akpm@osdl.org> wrote:

    > > in the next mail i will send a patch against 2.6.9-rc4 too (which
    > > luckily doesnt have the crash bug, but has the feature mismerge).
    >
    > OK. So it sounds like no additional patch will be needed for -mm?

    the sched.c chunks below are still needed. prof=schedule must be driven
    from schedule(), not from setscheduler().

    The profile.c one is a cleanup. [ and you surely must be a great fan of
    profiler cleanups by now B-) ]

    the patch applies to 2.6.9-rc4 cleanly as well, and the sched.c bits are
    necessary there too. I have tested this patch with and without
    prof=schedule and it now produces the expected results:

    314 work_resched 14.2727
    389 log_wait_commit 1.2006
    391 kjournald 0.6872
    499 need_resched 9.9800
    1549 worker_thread 2.5310
    1649 wait_for_completion 7.4615
    2551 do_exit 2.6435
    2999 schedule_timeout 16.7542
    3073 io_schedule 64.0208
    7863 preempt_schedule 126.8226

    the above is the scheduling profile of a bootup. ~40% of the reschedules
    were caused by PREEMPT, ~20% by IO and the remaining ~40% by process
    communication.

    without this patch the output is a boring empty profile ...

    Ingo

    Signed-off-by: Ingo Molnar <mingo@elte.hu>

    --- linux/kernel/sched.c.orig
    +++ linux/kernel/sched.c
    @@ -2646,6 +2646,7 @@ asmlinkage void __sched schedule(void)
    dump_stack();
    }
    }
    + profile_hit(SCHED_PROFILING, __builtin_return_address(0));

    need_resched:
    preempt_disable();
    @@ -3222,7 +3223,6 @@ static int setscheduler(pid_t pid, int p
    policy != SCHED_NORMAL)
    goto out_unlock;
    }
    - profile_hit(SCHED_PROFILING, __builtin_return_address(0));

    /*
    * Valid priorities for SCHED_FIFO and SCHED_RR are
    --- linux/kernel/profile.c.orig
    +++ linux/kernel/profile.c
    @@ -22,14 +22,14 @@ static int __init profile_setup(char * s
    int par;

    if (!strncmp(str, "schedule", 8)) {
    - prof_on = 2;
    + prof_on = SCHED_PROFILING;
    printk(KERN_INFO "kernel schedule profiling enabled\n");
    if (str[7] == ',')
    str += 8;
    }
    if (get_option(&str,&par)) {
    prof_shift = par;
    - prof_on = 1;
    + prof_on = CPU_PROFILING;
    printk(KERN_INFO "kernel profiling enabled (shift: %ld)\n",
    prof_shift);
    }
    -
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.kernel.org
    More majordomo info at http://vger.kernel.org/majordomo-info.html
    Please read the FAQ at http://www.tux.org/lkml/

    \
     
     \ /
      Last update: 2005-03-22 14:07    [W:0.025 / U:60.188 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site