power off (again)

Hi!

I'm still fighting the problem that power off doesn't work with one of our machines since moving from 2.2.19 to 2.4.7 kernel.

I compared both apm.c versions (1.13ac and 1.14) but there are no 'real' changes to the functions involved in the shutdown process, except some minor changes to macro definitions.

'cat /proc/apm' says '1.14 1.2 0x03 0x01 0xff 0x80 -1% -1 ?' which are all normal values for non-notebooks.

I'm a little bit stuck with this and I have no idea what to look for so hints would be highly appreciated!

Christian Schoenebeck

P.S. Yes, apm is supported by the bios, I have enabled apm, I tried real mode power off, I tried using apm as module and also tried acpi instead

Just out of curiosity, have you changed your power off scripts to
reflect: "halt -p". I can't remember where you make this change on RedHat, but I had to make it to make 2.4.x work for power down (Actually it was 2.3.xx where the change was made, I believe).

Trever

Am Freitag, 19. April 2002 06:56 schrieb [email protected]:

And ours append="apm=on", but I also tried apm=power-off - didn't help

Am Donnerstag, 18. April 2002 22:46 schrieb Marcelo de Paula Bezerra:

I tried: APM only, ACPI only and I'm not really sure, but I think I also tried ACPI & APM, but AFAIK this automatically enables just one of them anyway.

Am Donnerstag, 18. April 2002 23:02 schrieb Trever L. Adams:

Yes, this is not the problem

Oh, forgot to mention earlier...

The machines that refuse to actually switch off all SUSPEND just fine. They'll power down most of the way and stay suspended on battery power for a week.

It's just that when you tell them to actually switch off, use NO battery, and reboot when you come back up... Then they won't even suspend, they just stay on but spin the hard drive down.

Weird, eh?

Rob

Just thought I'd give a "me too" response. The Red Hat 7.2 kernel powers down all three systems I've tried it on (a Dell Inspiron 3500 laptop, a Toshiba Tecra 8000, and an SIS chipset motherboard). The 2.4.18 and 2.4.17 kernels do NOT power down any of those systems (It will spins down the hard drive instead, but the system power stays on. Yes, I'm compiling in the right APM support. I've tried it both with and without the "use APM bios to power down" switch.)

The only difference between boots is the kernel. Haven't had time to really look into it, but a quick check of the code didn't reveal any obviously cuplable differences between the apm power down functions under the arch directory. It seems like something else in the system state is getting twiddled that turns a turn-the-machine-off request into a "reduce power but stay on" request that just spins down the hard drive...

Dunno.

Rob

Am Freitag, 19. April 2002 17:52 schrieb Rob Landley:

And I already thought I was the only one having that problem.

Meanwhile Wolfgang Loeffler told me the possibility that power off also won't work if you enabled SMP on a uniprocessor machine. It wasn't a solution for my machine as I haven't compiled the kernel with SMP support, but maybe it's one for yours.

Hi,

Should be /etc/rc.d/init.d/halt. See appended patch.

Regards,
Thunder

                                                  Thunder from the hill.
Not a citizen of any town.                   Not a citizen of any state.
Not a citizen of any country.               Not a citizen of any planet.
                         Citizen of our universe.

@@ -234,3 +234,3 @@
 
-HALTARGS="-i -d -p"
+HALTARGS="-i -d"
 if [ -f /halt ]; then

Nope. SMP is off.

It might be a configuration thing. Maybe. I suppose I could compare the red hat and 2.4.18 .config files to see if anything obvious jumps out at me.

But it does suspend just fine. That's the odd part...

Rob

Unless your patch is reversed, that's what I've got (on the red hat systems, the linux from scratch systems use BSD style init scripts because I'm not THAT masochistic). It doesn't help.

I've also wandered through arch/i386/kernel/apm.c enough to confirm that apm_power_off() does seem to be getting called. And the hard drive does audibly spin down on the systems that have the case off, it just doesn't stop throwing out a video signal to the monitor, using the processor to heat the room, etc...

I might get some time to thump onn it more this weekend, but it's pretty far down on the to-do list...

Rob

It is reversed. If you want power off, you do need the -p.

Trever

Hi,

Sure. I accidently mixed those files... As you could see, the halt was removed, the halt~ inserted...

Regards,
Thunder

                                                  Thunder from the hill.
Not a citizen of any town.                   Not a citizen of any state.
Not a citizen of any country.               Not a citizen of any planet.
                         Citizen of our universe.

In any case, I just re-verified: "halt -p", as root, from the command line, takes the machine down to the "Power down" message, parks the hard drive, but leaves the rest of the system on. This is on both a linux from scratch system and a Red Hat system, both of which have been known to power down before (with a different kernel).

It still might be my .config, although a config that produces a kernel that powers down for suspend but won't power down on halt, on three radically different systems (dell, toshiba, sis)...

I'll thump on it some more later.

Rob

Another data point: VIA KT266 chipset (MSI mobo) and ALI Magick1 chipset (Soyo mobo), kernels 2.4.7, 2.4.17 and 2.4.18: enabling ACPI makes the system power down correctly. It doesn't matter whether APM is enabled, nor whether "use real mode APM to power down" is enabled. Single processor, no SMP support in the kernel. (Distro kernels tested:
stock Redhat 7.1, 7.2, and Mandrake 8.1 kernels do not power the system down; the stock SuSE 7.3 kernel does.)

	...Akkana