APMD-List:
Archives

  From: Craig Markwardt <craigm@pcasun3.gsfc.nasa.gov>
  To  : Avery Pennarun <apenwarr@worldvisions.ca>
        <apmd-list@worldvisions.ca>
  Date: Mon, 28 Sep 1998 20:26:29 -0400 (EDT)

suspend rejections (was Re: apm proposed changes)

Avery Pennarun writes:
 > On Sun, Sep 27, 1998 at 12:45:59PM -0400, Craig Markwardt wrote:
 > 
 > >   * ability for user-level code to *reject* a pending standby or
 > >     suspend event.  Very handy if something nasty is going on.
 > 
 > Sounds interesting, although I don't know what kind of rules you'd use to
 > decide this.  As far as I know (and I'll be happy if you correct me) the
 > only reasons you might want to avoid suspending are that the buffer cache
 > hasn't been flushed, or you haven't finished cleaning up yet... and
 > preventing those situations should be a kernel concern.


Actually, there are lots of reasons that you might want to reject a
suspend at the user level:

   - if AC power is connected
   - if CPU load is high
   - if network load is high
   - if modem is on line

I think we should let the user decide their own power management
policy, if possible.  It shouldn't be built into the kernel.
 
To assuage your fears:

* Kernel-level drivers can already reject a pending standby or
  suspend.  User-level programs are notified but not given a chance to
  agree or disagree: they *must* agree. The extra rejection logic in
  my kernel is only activated if a user-level program specifically
  rejects it.  This option at least can give a user a way out.

* It's not in the official kernel yet, so let's see what Stephen has
  to say.  I haven't heard from him yet.

 > If we get too sneaky, we end up with laptops that refuse to shut down.  Mine
 > does sometimes, and it's really annoying :(

Hmmm.  My laptop does that too.  I actually did some intense snooping
around, and found that there is a bug in my APM BIOS which gets
tweaked under certain circumstances.  I think having the ability to
reject a suspend event might actually be good in those circumstances,
if for example, you know that a suspend would crash the machine.  [
This depends on a machine-to-machine basis. ]

Craig