Re: [gpm] Is g-p-m (or pm-suspend) supposed to work on a desktop PC?

["Paul Johnson" <pauljohn32 gmail com>]

Dear Richard and everybody else:

This has been an exciting bug.  Just to summarize the problem, on a
Dell Optiplex GX280, the hard disk cannot wake up from pm-suspend.
The EXT3 journal aborts, the sata drive remounts as read only, and
most everything works wrong.

This morning, I updated the Dell System BIOS to the newest revision
and I installed the  newest g-p-m and other RPMS from the utopia
archive.

Now, you say, "show us the messages file", which sounds great in
theory, except that the disk is remounted read only and so the
messages file does not collect anything that appears.  Observe there
is nothing between the suspend and the eventual "power button
restart".

May 31 10:10:48 pols16 gconfd (pauljohn-2219): Resolved address
"xml:readonly:/etc/gconf/gconf.xml.mandatory" to a read-only
configuration source at position 0
May 31 10:10:48 pols16 gconfd (pauljohn-2219): Resolved address
"xml:readwrite:/home/pauljohn/.gconf" to a writable configuration
source at position 1
May 31 10:10:48 pols16 gconfd (pauljohn-2219): Resolved address
"xml:readonly:/etc/gconf/gconf.xml.defaults" to a read-only
configuration source at position 2
May 31 10:10:49 pols16 gconfd (pauljohn-2219): Resolved address
"xml:readwrite:/home/pauljohn/.gconf" to a writable configuration
source at position 0
May 31 10:12:49 pols16 ntpd[1745]: synchronized to 129.237.32.1, stratum 3
May 31 10:12:50 pols16 ntpd[1745]: time reset +0.578202 s
May 31 10:12:50 pols16 ntpd[1745]: kernel time sync disabled 0041
May 31 10:14:11 pols16 gnome-power-manager: Suspending computer
because the DBUS method Suspend() was invoked
May 31 10:14:19 pols16 ntpd[1745]: ntpd exiting on signal 15
May 31 10:23:44 pols16 syslogd 1.4.1: restart.
May 31 10:23:44 pols16 kernel: klogd 1.4.1, log source = /proc/kmsg started.
May 31 10:23:44 pols16 kernel: Linux version 2.6.16-1.2122_FC5
(bhcompile hs20-bc1-3 build redhat com) (gcc version 4.1.0 20060304
(Red Hat 4.1.0-3)) #1 Sun May 21 15:01:01 EDT 2006
May 31 10:23:44 pols16 kernel: BIOS-provided physical RAM map:
May 31 10:23:44 pols16 kernel:  BIOS-e820: 0000000000000000 -
00000000000a0000 (usable)
May 31 10:23:44 pols16 kernel:  BIOS-e820: 00000000000f0000 -
0000000000100000 (reserved)
May 31 10:23:44 pols16 kernel:  BIOS-e820: 0000000000100000 -
000000005fe86c00 (usable)
May 31 10:23:44 pols16 kernel:  BIOS-e820: 000000005fe86c00 -
000000005fe88c00 (ACPI NVS)
May 31 10:23:44 pols16 kernel:  BIOS-e820: 000000005fe88c00 -
000000005fe8ac00 (ACPI data)
May 31 10:23:44 pols16 kernel:  BIOS-e820: 000000005fe8ac00 -
0000000060000000 (reserved)
May 31 10:23:44 pols16 kernel:  BIOS-e820: 00000000e0000000 -
00000000f0000000 (reserved)
May 31 10:23:44 pols16 kernel:  BIOS-e820: 00000000fec00000 -
00000000fed00400 (reserved)
May 31 10:23:44 pols16 kernel:  BIOS-e820: 00000000fed20000 -
00000000feda0000 (reserved)
May 31 10:23:44 pols16 kernel:  BIOS-e820: 00000000fee00000 -
00000000fef00000 (reserved)
May 31 10:23:44 pols16 kernel:  BIOS-e820: 00000000ffb00000 -
0000000100000000 (reserved)


I can provide the whole messages file if you think it will help.

Before I upgraded the bios, I could read the dmesg output after
resuming and it said that the journal had aborted and so the drive was
remounted as read only.  THe messages I got were just the same ones
that are reported in this bug report on redhat:
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=181474. In that
one, the author shows a log like:

EXT3-fs error (device sdc1): ext3_journal_start_sb: Detected aborted journal
Remounting filesystem read-only

After I upgraded the bios, the symptom of the problem is different.  I
decided to get a pad of paper and copy all of this down to report to
you.  Unfortunately, it flashes by on the screen so fast I can't get
it all.

When pm-suspend runs, the screen goes black and I see a message

ata_command_failed (but I can't write down the rest).

After the system resumes, the screen stays dark, except for the
cursor.  but if I do ALT-CTL-F1 and ALT-CTL-F7, then the display shows
the windows that were open when the system went into suspend.

I type "dmesg" and see messages about the disk, they start like this:

end_request: I/O error, dev sda, sector 18586287

As I copy that down, a whole raft of new messages of that sort fill up
the xterm, and I can't scroll back to the beginning.  If there is
something about the journal aborting, I cannot see it in the dmesg
output.  All I can see are the end_request lines.

Any effort to run a program like /sbin/shutdown fails, as it was when
I first posted about this desktop trouble.

If you have any ideas about how I might capture dmesg output when the
hard disk is read only, I will try.

Also, it is important that I learn how to remove the "suspend" option
from the system menu.  I can't have users accidentally hanging the
computer when they try to log out.

Thanks for your help.


pj


On 5/29/06, Richard Hughes <hughsient gmail com> wrote:
> On Thu, 2006-05-25 at 23:25 -0500, Paul Johnson wrote:
>
> Sorry for delay, just got back...
>
> > In our computer lab, we have a bunch of Dell Optiplex 280 pc's that
> > are 1 year old.  They have 3.2 GH pentium processors (FC5 installs an
> > SMP kernel by default) with ATI x300 video drivers.  There is one SATA
> > drive in each PC.  FC5 installed the g-p-m program and I was a little
> > surprised, because I thought that was just for laptops.  But these
> > machines are hot as hell all the time and so I considered the idea
> > that suspend to ram overnight might cool things off a little bit.  I'm
> > willing to try anthing, actually, but the CPU speed is not scalable...
>
> Yes, g-p-m should be used on desktops too, as it can be used with UPS's
> or wireless mice, and also handles DPMS screen power management.
>
> > Anyway, I tried the suspend in the gnome menu and the system did
> > suspend and the power button goes slowly on and off, and when I push
> > the button, the video does wake up
>
> Is that using the proprietary ati drivers or the free ones?
>
> >  and I can type in any open x
> > terminals.  however, no programs will run.  Commands like "/sbin/fdisk
> > /dev/sda" just return "command not found." The hard disk appears not
> > to be mounted anymore.
>
> I think (kernel issue) SATA drives could be the issue here. Have you got
> any more details or dmesg output after suspend please.
>
> > By coincidence, a friend mentioned to me tha the 2.6.16 kernels
> > included a SATA patch that allowed the used of hdparm commands on SATA
> > drives. So I tested that out.  I am able to set the power management
> > settings and put the hard disk into standby mode. (don't know yet if
> > that makes it cooler).  I am also able to put the disk into sleep
> > mode, but it never wakes up.  And while the disk is still sleeping,
> > the system behaves just like it did after the failed restart after
> > pm-suspend.
> >
> > So then I wondered, maybe I should not try to do this at all with a
> > desktop PC.  Maybe g-p-m is intended only for laptops?  Since FC5
> > installs wireless tools on systems that have no wireless devices, and
> > lots of other unneeded crap, I can't rule out the possibility that
> > g-p-m is there by mistake, not intention.
>
> You need g-p-m to do other stuff, see above. The auto-suspend stuff is
> just as valid for desktops as laptops IMO.
>
> Richard.


--
Paul E. Johnson
Professor, Political Science
1541 Lilac Lane, Room 504
University of Kansas