Alan Coopersmith wrote:
> Garrett D'Amore wrote:
>
>> Someday we need to get RCM working for x86 suspend/resume. (It already
>> works for suspend/resume on SPARC -- at least on the E10K class of
>> systems.) Then applications would be able to register interest and
>> assertively elect to handle or reject a suspend/resume request just as
>> if they were kernel drivers.
>>
>
> Sounds like the /dev/srn (SRN=Suspend/Resume Notification) interface that
> already exists in x86 suspend/resume for the X server to use (though it has
> the bug that the X server can't report "suspend failed, abort" - I think
> that's on Randy's to do list, since he found it).
>
>
The RCM interface is general purpose... and its not just notification.
Its an interpose interface, so that programs can refuse a suspend
request. It also has the ability to spawn a program (e.g. a shell
script) to do things too....
The RCM stuff was developed nigh a decade ago for the E10k. I
participated in the design of it, along with Scott Carter and others.
Sometimes you're doing something that is timing critical in userland,
and you *really* don't want to suspend/resume. So its potentially quite
useful. xntpd could use this, for example.
-- Garrett
