lkml.org 
[lkml]   [2010]   [Jun]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    Date
    From
    SubjectRe: [linux-pm] [PATCH v4] pm_qos: make update_request non blocking
    On Thu, 10 Jun 2010 08:39:04 -0500
    James Bottomley <James.Bottomley@suse.de> wrote:

    > On Thu, 2010-06-10 at 09:45 +0200, Florian Mickler wrote:
    > > On Wed, 09 Jun 2010 13:05:49 -0400
    > > James Bottomley <James.Bottomley@suse.de> wrote:
    > > > On Wed, 2010-06-09 at 18:32 +0200, Florian Mickler wrote:
    > > > > On Wed, 09 Jun 2010 12:07:25 -0400
    > > > > James Bottomley <James.Bottomley@suse.de> wrote:
    > > > > > OK, so the expression of the race is that the latest notification gets
    > > > > > lost. If something is tracking values, you'd really like to lose the
    > > > > > previous one (which is now irrelevant) not the latest one. The point is
    > > > > > there's still a race.
    > > > > >
    > > > > > James
    > > > > >
    > > > >
    > > > > Yeah, but for blocking notification it is not that bad.
    > > >
    > > > The network latency notifier uses the value to recalculate something.
    > > > Losing the last value will mean it's using stale data.
    > >
    > > Actually after pondering a bit, it is not stale data that gets
    > > delivered: (Correct me if I'm wrong)
    > >
    > > The update_notify() function determines the extreme value and then
    > > calls the blocking_notifier_chain.
    > >
    > > But just before the update_notify() function get's called, the
    > > work-structure is reset and re-queue-able. So it is possible to queue it
    > > already even before the extreme_value in update_notify get's
    > > determined.
    > >
    > > So the notified value is always the latest or there is another
    > > notification underway.
    >
    > Well, no ... it's a race, and like all good races the winner is non
    > deterministic.

    Can you point out where I'm wrong?

    U1. update_request gets called
    U2. new extreme value gets calculated under spinlock
    U3. notify gets queued if its WORK_PENDING_BIT is not set.

    run_workqueue() does the following:
    R1. clears the WORK_PENDING_BIT
    R2. calls update_notify()
    R3. reads the current extreme value
    R4. notification gets called with that value


    If another update_request comes to schedule_work before
    run_workqueue() has cleared the WORK_PENDING_BIT, the work will not be
    requeued, but R3 isn't yet executed. So the notifiers will get the last
    value.

    If another update_request comes to schedule_work() after
    run_workqueue() has cleared the WORK_PENDING_BIT, the work will be
    requeued and another update_notify will be executed later.

    >
    > If the update comes in before the work queue is run, then everyone
    > eventually sees the new value. If it comes in as the chain is running,
    > some notifiers see the old value and some the new. If it comes in
    > during back end processing, no-one sees the new value.
    >
    > James
    >

    No, I think that is not correct. There will always be another
    update_notify()-call after the last update_request-recalculation due to
    the fact that run_workqueue() calls work_clear_pending() before calling
    the callback function of that work.

    (We could even requeue the work from within that callback function if we
    wanted to keep the number of notifications the same. But then we would
    need to queue the extreme_values in the framework. I think this
    isn't needed.)

    Cheers,
    Flo


    \
     
     \ /
      Last update: 2010-06-10 16:43    [W:0.025 / U:29.416 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site