Suspend/Resume problems: what is "hpet" and test of koji kernel 2.6.21-1.3241.fc7

Paul Johnson pauljohn32 at gmail.com
Wed Jul 4 17:26:44 UTC 2007


I have a Dell Latitude D820 that does not resume after suspend to RAM
in F7.  Suspend had been working great in FC6, it did not work in FC5,
it did work in FC4.  I've been through a lot of testing and bad
restarts.

The problem I have now is not the "black screen" problem--simply a
video issue--the system is entirely unresponsive.  Keyboard does
nothing. I believe it is a kernel crash upon resume.  The motherboad
is Intel Corporation Mobile 945GM/PM/GMS, Dual Core Intel 2.33MHZ,
Nvidia card, Intel IPW3945.

I know, from reviewing the pm-suspend quirks list, that the experts
are suspicious of the proprietary Nvidia modules and the intel ipw3945
driver, so I've removed them entirely for testing and I boot into
runlevel 3.  The system suspends just fine, but no resume.

The only way out is to hold down the power button until it restarts.
The network is not active, can't log in, and I believe the kernel has
died.  I can't find evidence of a kernel panic in
/var/log/messages--just evidence of a suspend and then the eventual
restart.   If you tell me what to look for, I can check and report
back.

I've been through all kinds of tests and pm-suspend quirks.  In
http://bugzilla.redhat.com, I was reading several reports about the
same problem and one author made a great deal of progress, removing
components and rebuilding kernels, and he isolated the restart problem
to the "hpet" support in F7.  Some people reported the issue is fixed
in some kernels, but I've not found one yet that fixes it.

Here's the specific buzilla that points to hpet
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=241310#c18

There are other informative bugs, so I don't forget, here are a few of
my favorites
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=240964
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=240964

You'll see that there are many many random variables.  Kernel options
vary from machine to machine, not all machines with same model number
have exact same MOBO and BIOS and BIOS configuration.  Everybody who
finds suspend/resume does not work gets messages saying "unload module
XYZ" or "get the right quirk".  I think we are not getting to the root
of the problem on my machine, at least.

Today I'm testing a kernel 2.6.21-1.3241.fc7 I found here
http://koji.fedoraproject.org/koji/buildinfo?buildID=10224.  I'm using
the newer pm-utils also offered on that website because it includes
vbetool

I believe hpet is at the root of the problem.

If I add the boot line hpet=disable, then I find I CAN suspend and
resume (runlevel 3) with this command:

pm-suspend --quirk-vbe-post  --quirk-vbemode-restore

Well, that's good news.  When I tried the same with Dave Jones kernel
kernel-2.6.21-1.3232.fc7, the kernel and system would come back to
life, but the laptop's keyboard was dead.

So, by disabling hpet, what am I losing?  If I understand it right,
I'm wasting power while on battery, because the kernel has to poll
devices frequently, rather than having the devices just tell the
kernel what they need when they need it.  Is that so?  How great is
the impact?

I'm skeptical that this will work inside X11, mostly because I have to
use the Nvidia proprietary driver (nv lacks support for the resolution
I need as well as 3d acceleration) has never played well with vbetool.
 That's disappointing.  But I still have hope!


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




More information about the fedora-test-list mailing list