Bug#311185: linux-2.6: Does the Debian Kernel Team still have concerns about CONFIG_PREEMPT ? - Debian

This is a discussion on Bug#311185: linux-2.6: Does the Debian Kernel Team still have concerns about CONFIG_PREEMPT ? - Debian ; Package: linux-2.6 Followup-For: Bug #311185 If I understand the following output correctly, the current (testing) Linux kernel does *not* have any preemption activated: $ grep -i EMPT /boot/config-2.6.22-3-amd64 CONFIG_PREEMPT_NONE=y # CONFIG_PREEMPT_VOLUNTARY is not set # CONFIG_PREEMPT is not set # ...

+ Reply to Thread
Results 1 to 4 of 4

Thread: Bug#311185: linux-2.6: Does the Debian Kernel Team still have concerns about CONFIG_PREEMPT ?

  1. Bug#311185: linux-2.6: Does the Debian Kernel Team still have concerns about CONFIG_PREEMPT ?

    Package: linux-2.6
    Followup-For: Bug #311185

    If I understand the following output correctly, the current (testing) Linux
    kernel does *not* have any preemption activated:

    $ grep -i EMPT /boot/config-2.6.22-3-amd64
    CONFIG_PREEMPT_NONE=y
    # CONFIG_PREEMPT_VOLUNTARY is not set
    # CONFIG_PREEMPT is not set
    # CONFIG_PREEMPT_BKL is not set

    By reading the bug log, I seem to understand that this is due to stability
    concerns by the Debian Kernel Team.
    Do these concerns still hold, now that more than 2 years have passed?

    If I read /usr/share/doc/jackd/README.Debian correctly, activating
    preemption would benefit audio applications, especially:

    | If you want better (or even only decent performance (fewer/no xruns, lower
    | latency...), you will probably have to do a bit of work. This includes:
    | 1. get recent alsa-drivers running well
    | 2. set up an .asoundrc if you have a professional and
    | powerful soundcard whose settings/channels/mixers you'd
    | like to have finer control over.
    | 3. use a 2.6 kernel with the realtime-preempt patch applied
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^
    | 5. run in realtime mode


    I would love to see Debian as ready as possible for setting up an audio
    processing workstation with minimal configuration (and without recompiling
    kernels or other packages).



    -- System Information:
    Debian Release: lenny/sid
    APT prefers testing
    APT policy: (500, 'testing')
    Architecture: amd64 (x86_64)

    Kernel: Linux 2.6.22-3-amd64 (SMP w/2 CPU cores)
    Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8)
    Shell: /bin/sh linked to /bin/bash



    --
    To UNSUBSCRIBE, email to debian-bugs-dist-REQUEST@lists.debian.org
    with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

  2. Bug#311185: linux-2.6: Does the Debian Kernel Team still have concerns about CONFIG_PREEMPT ?

    hello francesco,

    nice to read you!

    On Thu, Dec 13, 2007 at 12:10:04AM +0100, Francesco Poli wrote:
    >
    > If I understand the following output correctly, the current (testing) Linux
    > kernel does *not* have any preemption activated:
    >
    > $ grep -i EMPT /boot/config-2.6.22-3-amd64
    > CONFIG_PREEMPT_NONE=y
    > # CONFIG_PREEMPT_VOLUNTARY is not set
    > # CONFIG_PREEMPT is not set
    > # CONFIG_PREEMPT_BKL is not set


    yep, this is current state.

    > By reading the bug log, I seem to understand that this is due to stability
    > concerns by the Debian Kernel Team.
    > Do these concerns still hold, now that more than 2 years have passed?


    well partialy
    but due to shipment in all major distribution volountary preemption
    is beeing reconsidered as well preempt_bkl
    http://lists.debian.org/debian-kerne.../msg00496.html
    tickless kernel is an environmental necessity and thus enabled
    http://lists.debian.org/debian-kerne.../msg00014.html
    hz_1000 needs a benchmark run.

    the stability != preempt seems no longer true as the preempt
    path gets more testing out there.

    > If I read /usr/share/doc/jackd/README.Debian correctly, activating
    > preemption would benefit audio applications, especially:
    >
    > | 3. use a 2.6 kernel with the realtime-preempt patch applied
    >
    > I would love to see Debian as ready as possible for setting up an audio
    > processing workstation with minimal configuration (and without recompiling
    > kernels or other packages).


    well the realtime patch already helped a lot,
    but is not ready for default.
    it is currently quite marketing hyped by rh and suse.

    warm greetings

    --
    maks



    --
    To UNSUBSCRIBE, email to debian-bugs-dist-REQUEST@lists.debian.org
    with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

  3. Bug#311185: linux-2.6: Does the Debian Kernel Team still have concerns about CONFIG_PREEMPT ?

    On Thu, 13 Dec 2007 02:24:15 +0100 maximilian attems wrote:

    > hello francesco,
    >
    > nice to read you!


    Hi Maximilian!
    Thanks for your quick reply.

    >
    > On Thu, Dec 13, 2007 at 12:10:04AM +0100, Francesco Poli wrote:

    [...]
    > > By reading the bug log, I seem to understand that this is due to
    > > stability concerns by the Debian Kernel Team.
    > > Do these concerns still hold, now that more than 2 years have
    > > passed?

    >
    > well partialy
    > but due to shipment in all major distribution volountary preemption
    > is beeing reconsidered as well preempt_bkl
    > http://lists.debian.org/debian-kerne.../msg00496.html
    > tickless kernel is an environmental necessity and thus enabled
    > http://lists.debian.org/debian-kerne.../msg00014.html
    > hz_1000 needs a benchmark run.


    This *sounds* encouraging.
    Let's make the *Universal OS* suitable for intensive audio processing
    too! ;-)

    [...]
    > > If I read /usr/share/doc/jackd/README.Debian correctly, activating
    > > preemption would benefit audio applications, especially:
    > >
    > > | 3. use a 2.6 kernel with the realtime-preempt patch applied
    > >
    > > I would love to see Debian as ready as possible for setting up an
    > > audio processing workstation with minimal configuration (and without
    > > recompiling kernels or other packages).

    >
    > well the realtime patch already helped a lot,
    > but is not ready for default.
    > it is currently quite marketing hyped by rh and suse.


    This means that I cannot use the realtime option with jackd on a pure
    Debian box (testing or unstable), at present.
    Do I understand correctly?


    --
    http://frx.netsons.org/doc/nanodocs/...n_install.html
    Need to read a Debian testing installation walk-through?
    .................................................. .... Francesco Poli .
    GnuPG key fpr == C979 F34B 27CE 5CD8 DC12 31B5 78F4 279B DD6D FCF4

    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.4.1 (GNU/Linux)

    iD8DBQFHYcgiePQnm91t/PQRAo17AJ9puKzOt7h3giDSOLcb/qmvGaUpjwCfZ5Cy
    nyOHFP60rmskTLJcPeFVzZQ=
    =BsQ3
    -----END PGP SIGNATURE-----


  4. Bug#311185: linux-2.6: Does the Debian Kernel Team still have concerns about CONFIG_PREEMPT ?



    > the stability != preempt seems no longer true as the preempt
    > path gets more testing out there.


    I'm running my notebook with full preemption since years now, without
    any problems. I had to use the -ck patches, though, as the normal
    preemption did not work well enough for me. I didn't check how the new
    scheduler compares with the latest -ck patches yet...
    Also I'm not sure how well kernels with activated preemption will work
    on servers, would be interesting to see some benchmarks....

    > well the realtime patch already helped a lot,
    > but is not ready for default.


    I've given the RT patch a try some months ago as I thought about
    creating jacklab[1]-like CDD, but the Kernel was just not stable enough
    for the daily use (like suspend/resume...) - but I think for a pure
    audio system the RT patch should be fine today, at least it would be
    interesting to give it a try. While having a RT kernel as default is not
    an option imho, it would be great to have an installable RT kernel for
    the latest hardware (nobody wants to do realtime sound processing on old
    machines imho....).


    Cheers,

    Bernd


    [1]: http://jacklab.org/
    --
    Bernd Zeimetz




    --
    To UNSUBSCRIBE, email to debian-bugs-dist-REQUEST@lists.debian.org
    with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

+ Reply to Thread