Regression: Fan does not come on due to ACPI problems

Please retest against 2.6.20-2 when it is available in the feisty archive.

That kernel doesn't boot at all for me, bug filed: https://launchpad.net/bugs/76204 .

76204 is fixed in 2.6.20-3, please retest against it.

Thanks

Tested on 2.6.20-3-generic.

No better. Ripping a CD causes my CPU temperature to jump up to roughly 80°C, causing no change in the fan noise and making my laptop touchpad all jumpy and unresponsive.

This bug is still there in 2.6.20-4-generic, version 2.6.20-4.6.

Please include this patch: http://bugzilla.kernel.org/attachment.cgi?id=9885&action=view .

And while I'm in wish list mode, this patch is needed for the fans to work properly after suspend/resume: http://bugzilla.kernel.org/attachment.cgi?id=9757&action=view

I downloaded linux-source-2.6.20, patched the kernel with
http://bugzilla.kernel.org/attachment.cgi?id=9885&action=view (for working fans)

and

http://bugzilla.kernel.org/attachment.cgi?id=10022&action=view (for working fans after resume)

and rebuilt the kernel.
The result was that pre-suspend, the fans worked as they should so the first patch helped.

The computer behaved strangely after resume so it was hard to tell if the second patch worked as it should. 2.6.17 and earlier Ubuntu kernels work fine with suspend to both ram and disk (modulo the fan problem) so it appears that there's been some suspend regressions in the 2.6.20 Ubuntu kernels, unrelated to this particular bug. I noticed Ben Collins saying in IRC that he'd dropped quite a few ACPI patches for the newer kernels, so that may be the cause.

I've included the first patch. Can you get me some context on the second patch?

Sure. it's from comment #69 in http://bugzilla.kernel.org/show_bug.cgi?id=7122 . I've done some more testing and that patch doesn't seem to help anyway. To sum up, the situation is this:

With Ubuntu kernels 2.6.17 and earlier:
* Before suspend, the fans come on as they should
* Suspend to both ram and disk works
* After suspend, usually (but not always) the fans get "stuck" at whichever speed they were running just before suspend.

With Ubuntu 2.6.20 kernels:
* Before suspend, the fans only come on if I run a script that does "acpi -t" every few seconds.
* Suspend to ram does not work. Most often, the fans come on full blast when I resume and the display never returns. Sometimes the system does come up but it's very unstable and X soon crashes.
* Suspend to disk works, but the fans have the same problem as in 2.6.17.

With my patched Ubuntu 2.6.20 kernel:
* Before suspend, the fans work as they should.
* Suspend to ram has the same problem as the unpatched kernel.
* Suspend to disk works, but the fans behave as with the unpatched kernel, i.e. the second patch does not seem to help.

I'll do some more testing. I know Matthew Garrett has an NX6125, maybe he can help.

I've tried various patches over the last week or so and at first didn't have much success with any combination but I'm now running 2.6.20-rc5 with the patches from http://www.sisk.pl/kernel/patches/2.6.20-rc5/ and I have working fans both before and after resume. For context, see comments 70-74 of http://bugzilla.kernel.org/show_bug.cgi?id=7122 .

situation on my nx6325 with the current kernel (2.6.20-6) and the latest hp BIOS (F.06):

* thermal events get processed out of the box, the fans spin up and down correctly

* rebooting without unloading "psmouse" results in "bad state" so that the boot takes slightly longer and no battery events get processed. fans still work.

* suspend-to-ram works, but after wakeup fans are permanently off. thermal events seem to get processed, since "cat /proc/acpi/thermal_zone/TZ*/*" says it is switching states.

* suspend-to-disk works, but has the same problems like suspend-to-ram

slightly OT:
where can I see which patches are applied to the ubuntu kernel? and which patches mentioned in this bug report were applied?

I upgraded my Edgy machine to Feisty yesterday. Kernel is 2.6.20.6-generic. The fan is always on before and after resume, though not a maximum speed.
With edgy it worked all right before and after resume, AFAICT.

*cough* fans are still of after resume *cough*

according to the comments in:
http://bugzilla.kernel.org/show_bug.cgi?id=7122

you need to apply the patches from comment #38 and #69 in order to make resume work.

just tried it - after applying the two mentioned patches, everything works fine.

after some more testing I found out that the "fan-problem-fix.patch" is also needed - otherwise the fans dont behave correctly after resume from disk. (skips state 3)

all patches can be found here:
http://www.sisk.pl/kernel/patches/2.6.20/

here is a kernel with those three patches additionally applied. nothing else changed:
http://uploader.polorix.net//files/140/linux-image-2.6.20-rc6-ubuntu1-m2k_2.6.20-rc6-ubuntu1-m2k_i386.deb

just in case someone wants to try it out.

I'm not sure how to test it quantitatively, but it seems to be better for me.

Bug # 7122 has been closed.

It is patched in 2.6.21-rc1.

Tried it, and now everything works great! :-)

commit 0a6139027f3986162233adc17285151e78b39cac

I have this problem with 2.6.20-9-generic. My system is an HP NC4200. Fans work normally before hibernate (suspend is broken due to another bug), but the fan does not work after, and the system overheats.

I will try the package above and report back.

tarek : )

I can confirm that the package posted by Pavel Rojtberg fixed this problem for me. Fans worked as expected after hibernate with that kernel running.

tarek : )

http://git.kernel.org/?p=linux/kernel/git/lenb/linux-acpi-2.6.git;a=commit;h=0a6139027f3986162233adc17285151e78b39cac

I'm going to test this against the current feisty kernel source.

I tested various combinations of patches. When i only apply the patches from comments 38 & 69 in http://bugzilla.kernel.org/show_bug.cgi?id=7122 then the fans after suspend only work after i execute

$ echo 0 > /proc/acpi/fan/C34F/state
$ echo 0 > /proc/acpi/fan/C350/state
$ echo 0 > /proc/acpi/fan/C351/state
$ echo 0 > /proc/acpi/fan/C352/state

The correct behavior like in 2.6.21-rc1 is only obtained when the patches in http://bugzilla.kernel.org/show_bug.cgi?id=7570 are included. I applied the patches from comment 12 & 14.

It would be nice if someone can confirm that.

Correct behavior is only obtained when the patches from comments 38 & 69 in http://bugzilla.kernel.org/show_bug.cgi?id=7122 and the patches from comment 12 & 14 in http://bugzilla.kernel.org/show_bug.cgi?id=7570 are included.

The fans come only up after suspend when the patches from comments 38 & 69 in http://bugzilla.kernel.org/show_bug.cgi?id=7122 and the patches from comments 12 & 14 in http://bugzilla.kernel.org/show_bug.cgi?id=7570 are included.

Applying all of the patches discussed by Markus Brechtel in this bug, and by others in Bug #74877 and Bug #89779, gives me the attached deb files. A note is that the following command must also be executed to make this package usable:

sudo ln -s /lib/firmware/2.6.20-10-generic /lib/firmware/2.6.20.1-ubuntu1-custom2

I shall try to get a diff from the tree and post that in the near future.

tarek : )

I apologize, but I can't upload the linux-image file. Is there someplace where I can put it? It is apparently 226 megs.

tarek : )

I think if you just post the diff against the unpatched tree, that should be quite sufficient. Hopefully with several people having confirmed that the patches fix the bug, the patches can go into the next kernel build.

A Diff of all of the patches is included. Also, here are instructions for doing it for lay folks. It is assumed that the diff file is in the home directory:

1. sudo apt-get install kernel-package libncurses5-dev fakeroot bzip2 linux-source-2.6.20

2. cd /usr/src

3. sudo tar jxvf linux-source-2.6.20.tar.bz2

4. sudo cp /boot/config-2.6.20-10-generic linux-source-2.6.20/.config

5. sudo patch -p0 < ~/NC4200-acpi.diff

6. cd linux-source-2.6.20

7. sudo make-kpkg clean

8. sudo fakeroot make-kpkg --initrd --append-to-version=-custom kernel_image kernel_headers

This should produce a usable deb package in /usr/src

tarek : )

Ok, I really want to take in these patches, but I can't take them as one huge hunk. I need it broken down, preferably with SHA's for upstream's git which allows me to git-cherry-pick instead of munging my tree, plus it shows me where the patch came from, the author, and feedback about the patch.

I've done this work on my own for the majority of this patch, except for the part that changes root_suspend/root_resume to acpi_root_suspend/acpi_root_resume. Can you tell me where this patch came from and how it helps here?

Hi Ben!

Thank you very much for your work, and thank you for your patience with me. With regards to that particular patch, the usage of the function comes from Comment #38, Bug #7122, here:

http://bugzilla.kernel.org/show_bug.cgi?id=7122#c38

The function itself is from the same bug, Comment #69, here:

http://bugzilla.kernel.org/show_bug.cgi?id=7122#c69

I am not good with these things, so I am not positive why this function was itself necessary, but as indicated in Comment #93 of that bug, that patch shipped with 2.6.21-rc1 (and above?).

Please let me know if there are any other parts you want me to hunt.

thank you,

tarek : )

Here is the SHA as requested:

0a6139027f3986162233adc17285151e78b39cac (Comment 69 patch)

I have applied the attached patch on my HP/Compaq nc6220 - this patch doesn't quite apply cleanly to the current source as it looks like parts of it have made their way in already.

After the patch I have a working hard drive after suspend BUT no built-in sound. This is a regression back to the current behavior in Dapper/Edgy on this hardware. 2.6.20-11 with the patch included in bug #74877 gives me both working suspend AND working sound.

This is not fixed for me on HP nw8240: after resume from hibernate, ac_adapter is not properly detected, fan's do not start at all and CPU temperature quickly goes to 100C if I start some build. Also note that "acpi -t" command takes some 5 seconds to report temperatures:

$ acpi -t
     Battery 1: discharging, 60%, 01:10:50 remaining
     Battery 2: charged, 0%
     Thermal 1: active[3], 104.0 degrees C
     Thermal 2: ok, 82.0 degrees C
     Thermal 3: ok, 54.0 degrees C
     Thermal 4: ok, 0.0 degrees C

Should I re-open this bug or open new bug?

I will leave this one, and have re-opened bug 63123 which I initially reported.

This issue appears completely resolved for me in 2.6.20-14 on an HP NC4200 and HP NC6230

tarek : )

Works for me too. Many thanks!

Sometimes the Fans on first stage don't come up after resume on HP nx6325.

This script makes them come up again.

oh shit, i released the script too early. Shame on me! :-(

Here is the correct one