tech-userlevel archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Making powerd=YES default



On Jul 25, 2011, at 7:02 AM, Adam Hoka wrote:

> On Mon, 25 Jul 2011 15:52:21 +0200
> Marc Balmer <mbalmer%NetBSD.org@localhost> wrote:
> 
>> Am 25.07.11 14:38, schrieb Jean-Yves Migeon:
>>> Dear all,
>>> 
>>> Would it be acceptable to turn powerd on by default? Rationale follows:
>>> 
>>> My current save/restore code for Xen makes use of it. The way the Xen
>>> API is designed does not permit sending back notifications to dom0 to
>>> inform that suspension cannot be done, and this ends up with a xm(1)
>>> command sleeping ~forever without the user's knowing why.
>>> 
>>> IMHO this is quite inconvenient from an admin's perspective, and the
>>> only ways around are:
>>> 1 - to have powerd(8) enabled by default,
>>> 2 - let the domU start it (assuming he knows that he has to) before
>>> firing suspend events. Assuming that the dom0 admin is not necessarily
>>> the domU's admin, this can be problematic.
>>> 3 - completely circumvent powerd(8) by managing the event in kernel, and
>>> let the pmf(9) threads perform the sleep for us.
>>> 
>>> IMHO, 1) looks like the best trade off. 3) feels awkward, because it's a
>>> choice that says powerd(8) might not be necessary for suspension, which
>>> is a departure from delegating power management to userland.
>>> 
>>> Opinions welcome. if no one objects, I'll go for 1) in a week or so :)
>> 
>> My opinion is:  It should *not* be enabled by default, but rather during
>> system install time, i.e. in sysinst, towards the end of the
>> installation, the user should be asked a few questions:
>> 
>> - Do you want to run powerd by default?
> 
> It's like: Do you install NetBSD on a machine not from the garbage?
> powerd on i386/amd64 pretty much part of the hardware support
> (although some of us would like to have less of its functionality in
> userspace, eg possibility to use aps, without touching the disk)

Not all arches support power management.  Maybe there should be a sysctl which 
indicates whether the kernel support power management.  /etc/rc.d/powerd could 
check this to see if starting powerd makes sense.  How does powerd determine 
the power states that are supported?


Home | Main Index | Thread Index | Old Index