F16 Kernels 3.6.* cause hibernate to freeze on Dell E6410

[aaronsloman]

(Sorry -- realised that the thread title should have mentioned kernels 3.2.*, not 3.6.* (the most recent one I tried was kernel 3.2.6-3)
Title now corrected.)

I am using 32bit Fedora 16 on a Dell Latitude E6410, with intel graphics and intel wireless.

I use hibernate a lot, to preserve contents of my various virtual desktops using the CTWM window manager (or OpenBox) sitting on an installation of Fedora LXDE. I start up in plain text mode (runlevel 3) then invoke X using startx, controlled by my ~/.xinitrc to specify window manager, etc.

Everything works fine on kernel 3.1.7-1.fc16.i686 but ever since 3.2.* kernels have appeared I have had troubles with pm-hibernate freezing.

I've noticed that with the newer kernels hibernation is very much faster, and it reports using 3 threads for compression. But after I have hibernated and resumed two or three times I find that hibernation freezes soon after reporting the number of threads. I then have to hold down the power button for several seconds to force a complete reboot. So each time after this happens a couple of times I remove the latest kernel and go back to 3.1.7-1.

This has happened in several different versions of 3.2, most recently with 3.2.6-3.fc16. So I am waiting for the next update, hoping the problem will be fixed.

I have noticed that there has been a greater frequency of kernel updates than usual in the last few weeks. I hope this message will get the attention of people working on the bugs.

I have a 500GB drive, with a 10G swap partition, and 4G RAM.

I hope this can be fixed because the much faster hibernate and resume is impressive!

This may also be relevant. I had problems with resume failing in the past which I solved by having two versions of the grub2 boot entry for each kernel. I use one for a fresh reboot, and the other (the default) for resume. The only difference is that the resume version has on the linux line acpi=off

I was forced into this scheme because without it resume occasionally failed and I would then get an automatic reboot. (I read the tip about acpi=off somewhere on a web page, though I can't find documentation about what it does! Using it at boot time stops some things working, but it's OK for resume.).

/boot is in a separate partition, making it easy to try new versions of linux in different partitions, sharing a boot partition.

Aaron Sloman
University of Birmingham UK

[christian_l]

3.2 seems to have created multiple problems for various users. Have you checked Bugzilla (https://bugzilla.redhat.com/) yet? On this forum here you might get advise if somebody has already figured out a workaround, but to really get things fixed, you might want to look at Bugzilla. First check whether somebody has already reported this problem. Maybe you can add something to it, and thus help the developers figure out the details. Or maybe they are not aware of it yet - then you should definitely file a new bug report to bring it to their attention.

[jims]

Greetings:

There are some items related to 3.2.6-3 kernel on bugzilla and one seems to reference your issue
https://bugzilla.redhat.com/buglist....search=3.2.6-3

[aaronsloman]

Thanks for the comment. The only directly related thing I found (to which I am unable to contribute) was this
https://bugzilla.redhat.com/show_bug.cgi?id=785384

But that stated (wrongly according to my experience) in comment 4: Hibernate works again since kernel 3.2.5-3.fc16.i686 (or possibly one of the other packages that came in the upgrade). Closing this bug.

Unfortunately, I still get (intermittently) freezing during pm-hibernate after something like this is printed out:
[paraphrase] using 3 threads for compression.
PM: Compressing and saving image data nn%

Some times it freezes immediately. Sometimes the nn% pauses and changes a few times then freezes.
(My E6410 has a four core intel i5 cpu, with intel graphic card.)

Normally it doesn't freeze the first time I use pm-hibernate. It may work fine two or three times, then freeze on the next one. Memory 4G, swap area 10G.

When it works, hibernation and resume are very fast. My hunch is that the code for multi-threaded compression does not cover enough cases and still has bugs.

Unfortunately I am not a system programmer and cannot dig into kernel software to experiment.

Could this have anything to do with the fact that I run neither gnome nor kde. I boot into runlevel 3 )multi-user text only) then use startx to run X11 with CTWM (or sometimes OpenBox as window manager). Then I don't reboot until I have a special reason e.g. new kernel, because hibernate/resume keeps my 10 virtual desktops running. I have tried several versions of the 3.2 kernel. Each works for a while, then freezes. But 3.1.7-1 works perfectly, though more slowly.

If I can do any specific tests or provide more information about my configuration I am willing to try, though I'll be going away tomorrow to give a talk.
Aaron
www.cs.bham.ac.uk/~axs

[aaronsloman]

I have just installed the latest 3.2 kernel on my E6410. When I tried pm-hibernate, it seemed freeze waiting for something (I forget the exact message printed on the terminal, but it reminded me of the SD card which I leave in the card reader as a device for backups when travelling.)

So I removed the card: and immediately after that the hibernation continued and the machine shut down as expected.
Could there be a bug hibernating with a card in the SD card reader, which has not been mounted?

[geoffleach]

Similar problems here. See Bugzilla https://bugzilla.redhat.com/show_bug.cgi?id=796516

There may be a movement starting in the kernel space to dump hibernate. See https://bugzilla.redhat.com/show_bug.cgi?id=781749

[aaronsloman]

I have submitted a (lengthy) plea for hibernate to be preserved, here https://bugzilla.redhat.com/show_bug.cgi?id=781749

[aaronsloman]

The problem of hibernate freezing seems to have been fixed in kernel 3.3.0-4.fc16.i686

There is still a problem with resume failing and causing a reboot. I have found that by having a grub menu entry for resume with acpi=off as boot flag allows resume to work reliably. This should not be necessary.

There is also a serious problem with 'yum install kernel....' or 'yum update kernel' : it gets the root drive ID wrong in /boot/grub2/grub.cfg, making it impossible to boot into the new kernel without first editing the drive ID. E.g. see https://bugzilla.redhat.com/show_bug.cgi?id=756559