Most recent kernel where this bug did not occur: 2.6.12-6 Distribution: vanilla kernel on FC4 system Hardware Environment: Dell Latitude LS H500ST Software Environment: Problem Description: Suspend to RAM worked on this machine in kernels 2.6.11 and 2.6.12. In 2.6.13 and 2.6.14, when I echo mem > /sys/power/state, the display turns off, the fan and the hard drive spin down, and the machine stops responding to network pings, but the LED that indicates power status remains in the "on" state, rather than entering the "asleep" state, and the machine refuses to wake up again. Using git bisect, I identified the following commit as the culprit: http://kernel.org/git/?p=linux/kernel/git/lenb/linux-acpi-2.6.git;a=commit;h=87bec66b9691522414862dd8d41e430b063735ef
Do you think the commit is still the root cause in 2.6.14? The possible violation (disable link device at suspend) has been removed (the code it's commented). Could you please revert the commits in pci_link.c, and see if your problem disappears?
It's not possible to revert the commit only in pci_link.c, because it defines new functions in pci_link.c that it calls from other files. I confirm that reverting the entire commit gets rid of the problem. What are you saying is commented, exactly? I don't see anything that comes from this commit that's commented out in 2.6.14.
I'm still interested in getting this fixed. I'm available to test patches and otherwise help debugging, but I don't understand what's going on enough to do it myself. Is there any hope? I tried the acpi kernel from git last night, and the behavior is the same except that now the hard drive continues spinning when the machine goes into sleep mode.
In kernel 2.6.19 on an FC4 system, S3 works after booting in single-user mode. If I then start up services one at a time, suspend continues to work until I start klogd. Running klogd has some effect that persists even after klogd has been stopped--if I start klogd, then kill it, then try to suspend, suspend doesn't work. This bug is a moving target--the symptoms seem to change with every new kernel version. The last one I tried was the Fedora kernel 2.6.17-1.2142_FC4. In that version suspend never works, even in single-user mode with no services started.
hmmm, what is the state of things in 2.6.22?
Looks like suspend to RAM works fine in 2.6.22.
thanks. closed.