[lkml]   [2006]   [Jan]   [25]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
    SubjectRe: pthread_mutex_unlock (was Re: sched_yield() makes OpenLDAP slow)
    Lee Revell wrote:
    > On Wed, 2006-01-25 at 10:26 -0800, Howard Chu wrote:
    >> The SUSv3 text seems pretty clear. It says "WHEN
    >> pthread_mutex_unlock()
    >> is called, ... the scheduling policy SHALL decide ..." It doesn't say
    >> MAY, and it doesn't say "some undefined time after the call."
    > This does NOT require pthread_mutex_unlock() to cause the scheduler to
    > immediately pick a new runnable process. It only says it's up the the
    > scheduling POLICY what to do. The policy could be "let the unlocking
    > thread finish its timeslice then reschedule".

    This is obviously some very old ground.

    Kaz's post clearly interprets the POSIX spec differently from you. The
    policy can decide *which of the waiting threads* gets the mutex, but the
    releasing thread is totally out of the picture. For good or bad, the
    current pthread_mutex_unlock() is not POSIX-compliant. Now then, if
    we're forced to live with that, for efficiency's sake, that's OK,
    assuming that valid workarounds exist, such as inserting a sched_yield()
    after the unlock.

    But then we have to deal with you folks' bizarre notion that
    sched_yield() can legitimately be a no-op, which also defies the POSIX
    spec. Again, in SUSv3 "The /sched_yield/() function shall force the
    running thread to relinquish the processor until it again becomes the
    head of its thread list. It takes no arguments." There is no language
    here saying "sched_yield *may* do nothing at all." There are of course
    cases where it will have no effect, such as when called in a
    single-threaded program, but those are the exceptions that define the
    rule. Otherwise, the expectation is that some other runnable thread will
    acquire the CPU. Again, note that sched_yield() is a core function of
    the Threads specification, while scheduling policies are an optional
    feature. The function's core behavior (give up the CPU and make some
    other runnable thread run) is invariant; the current thread gives up the
    CPU regardless of which scheduling policy is in effect or even if
    scheduling policies are implemented at all. The only behavior that's
    open to implementors is which *of the other runnable threads* is chosen
    to take the place of the current thread.

    -- Howard Chu
    Chief Architect, Symas Corp.
    Director, Highland Sun
    OpenLDAP Core Team

    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to
    More majordomo info at
    Please read the FAQ at

     \ /
      Last update: 2006-01-25 23:18    [W:0.029 / U:8.388 seconds]
    ©2003-2017 Jasper Spaans. hosted at Digital OceanAdvertise on this site