#1 Monitoring interrupts

closed-accepted
None
5
2001-09-10
2001-07-02
No

This is a proposal for a new feature. Actually it
doesn't fit my needs but since it is written, I submit
it.

The goal is to prevent the disks from spinning down
when the machine is not idle (i.e. for a desktop
machine with a main user on the console).

The idea is to monitor interrupts (keyboard and mouse)
and to add an additional checking before spinning down
the disk. Interrupts are monitored via
/proc/interrupts.

Actually I would like also to stop some daemons. It is
out of the scope of noflushd I guess.

Laurent Pelecq

PS: I'm not subscribed to the mailing list (yet).

Discussion

  • Daniel Kobras

    Daniel Kobras - 2001-07-03

    Logged In: YES
    user_id=7832

    Nice idea. Thanks! I'll have time to look at the patch in
    detail only after LinuxTag (i.e. next week), but from your
    description I really like the feature. Stay tuned.

     
  • Daniel Kobras

    Daniel Kobras - 2001-07-03
    • assigned_to: nobody --> nold
     
  • Laurent Pelecq

    Laurent Pelecq - 2001-07-04

    Logged In: YES
    user_id=252538

    This an improvementfor the first patch.

    I forgot to say how to use it. There is a new option -i
    which takes a comma separated list of interrupts ids
    (ex: -i 1,12). Interrupts are monitored according to
    the default timeout (-n).
    
    With the second patch (note: it doesn't include the first
    one), when all monitored interrupts are still idle after
    the timeout, the program /etc/noflushd/do_idle is run if it
    exists. When the machine becomes active again, the program
    /etc/noflushd/do_active is run.So you can stop some
    daemons.

    The side effect is that if the timeout is 30 minutes and
    nobody is working on, after 30 minutes the script do_idle
    is run. It may cause read io. So you have to wait 30 minutes
    again for the disks to be spun down.

     
  • Laurent Pelecq

    Laurent Pelecq - 2001-07-13

    Logged In: YES
    user_id=252538

    Another patch. The problem is that the mouse interrupt
    may not appear when noflushd is started at boot time.
    However it may not be a real problem.

     
  • Laurent Pelecq

    Laurent Pelecq - 2001-07-13

    Logged In: YES
    user_id=252538

    boot time patch.

     
  • Laurent Pelecq

    Laurent Pelecq - 2001-08-12

    Logged In: YES
    user_id=252538

    Here is the patch for version 2.5. It is a single patch.
    The first three ones were for version 2.4. I have deleted
    them to avoid confusion.

     
  • Laurent Pelecq

    Laurent Pelecq - 2001-08-12

    Logged In: YES
    user_id=252538

    Correction. I'm not allowed to delete any files so only
    the latest is the correct one. Other patches were for v2.4.

     
  • Daniel Kobras

    Daniel Kobras - 2001-09-10

    Logged In: YES
    user_id=7832

    Whoops. For some reason, I didn't receive the usual e-mails
    when you added the new files, so I completely missed them.
    Patch is in the current CVS now with only a few minor
    cosmetic done by me. I might tweak it a little bit stil
    before final release. Many thanks for your contribution!

     
  • Daniel Kobras

    Daniel Kobras - 2001-09-10
    • status: open --> closed-accepted
     

Log in to post a comment.