Date: Sat, 13 Jun 1998 23:28:14 -0600 From: Warner Losh <imp@village.org> To: mobile@FreeBSD.ORG Subject: apm vs apmconf Message-ID: <199806140528.XAA11752@harmony.village.org>
next in thread | raw e-mail | index | archive | help
OK. I'm putting some finishing touches on some APM 1.2 enhancements
to FreeBSD. I have a couple of questions, so I thought I'd share what
I have done so far then ask the questions. This is a little long, so
please bear with me. I'm very interested in hearing what others who
have looked into this problem have to think about what I've done and
where I'd like to take it.
First, I have added a "go into standby mode" IOCTL to the apm driver.
I've also tweaked syscons to have a STBY key just like it has a SUSP
key now. I've tweaked kbdconfig to grok this key. I've tweaked a
couple places in the driver to issue a standby command rather than a
suspend (like for a user requested standby). This is accessible to
the user in apm(8) via -Z. Seemed like a good thing to me. This is
in the kernel because it needs to know when we go into STANDBY mode so
it can call the right suspend/resume hooks. I'm using this in
conjunction with X screen savers because it turns off the display (of
at least my libretto). While the libretto 50CT won't charge in this
mode (more's the pitty), it does crank the power consumption way down.
I've modified the APMIO_BIOS ioctl to not return EIO when an error
happens. Rather, it clears the ah portion of the eax value returned
on success. This isn't bad because all of the calls that return stuff
in ah are APM driver only calls (that's what the APM standard calls
them). You would need to check ah to see if it was non-zero to see if
the ioctl succeeded.
I've also implemented the APM 1.2 set/get/disable resume timer
functionality in apm(8) (although I'm starting to think this might be
better in apmconf(8), which is what motivated this message). apm(8)
reports the time that the computer will resume from {suspend,standby}
mode. For the moment, you can only specify the number of seconds from
now to have the timer expire, but I plan to convert to using strptime.
This is in userland using APMIO_BIOS.
I've added the resume on ring indicator, but have no modem connected
to my libretto to test it. It reports that this is disabled on my
box. This is in userland using APMIO_BIOS.
I've expanded the information that APMIO_GETINFO to include the number
of batteries installed on the laptop as well as the APM capabilities
that exist for this box (eg can resume from timer for standby, for
suspend, etc).
I've also tweaked the relevant man pages.
What's left to be done before I commit:
o resume timer in absolute time (strptime) format
o reporting wakeup timer in the right locale format
o have a way to report just the resume time
o have a way to report just the resume on ring indicator
status
o enhance apm(4) man page
o better trapping of the 1.2 only features of my
enhancements.
o better APM error reporting in apm(8)
o GMT issue (see questions)
o proof read my man page tweaks
What's waiting on PnP BIOS code access:
o Fine grain power control (eg display having different
brightnesses, CPU having different speeds, etc).
o Even more reporting on power management things that apm 1.2
supports.
Nice to have, but not yet:
o OEM specific function from the command line, at least to a
limited extent (giving number args is easy, but giving
pointers to things is much harder for anything but strings).
o The ability to get notified when new capabilities happen to
the system, eg docking at a docking port. Since I don't
think the libretto 50CT generates these, I may not implement
them for some time.
Long term goal:
o have the ability to write a powerd that sits and gets APM
events and deals with them in the way that it sees to be
most fit. This would allow for easier configuration of
power management policies. This likely won't happen for
quite some time.
o move more of the driver into userland. Maybe have a libapm
to facilitate sharing of common code between modules (apm,
apmconf and the currently mythical powerd). This would also
make it possible for apm(8) to be less dependent on ioctls
that go into the kernel for info since it can get this
information itself with the APMIO_BIOS ioctl.
My question:
o does configuring the APM BIOS's wakeup timer belong in apm
or apmconf?
o ditto with ring indicator resume
o ditto with the OEM functions
o the timer values that I use MUST be in RTC time. how do I
tell if that is GMT or localtime, my guess is that it is one
of the following sysctls: machdep.adjerntz or
machdep.wall_cmos_clock? There doesn't seem to be docs for
them in any of the usual places.
o should I commit well tested portions of this work, or wait
for all pieces to in place first?
Comments in general?
Warner
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-mobile" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199806140528.XAA11752>