resume fails after suspend in Lenovo 3000 v100

[Felipe Figueiredo]

Using Hardy (x86-32), up to date with hardy-update, I can't resume after suspend on a Lenovo 300 v100 laptop.

I can resume after hibernate.

I'm attaching debug information as per https://wiki.ubuntu.com/DebuggingKernelSuspend and some extra stuff that might be useful.

The information from dmesg (complete output attached) is

--
[ 21.348639] Magic number: 0:731:232
[ 21.348649] hash matches /build/buildd/linux-2.6.24/drivers/base/power/main.c:76
[ 21.348678] hash matches device ttyv2
--

I can't find information anywhere on how to disable whatever that is ttyv2. In an earlier hardy kernel, the message was about ttyv4, and then I didn't have any such device in /dev. Now I do have /dev/ttyv[0-f], but I have no idea what these tty's are for. There are nothing about any of them in a normal dmesg, or /var/log/kern.log.

[Felipe Figueiredo (philsf)]

Subscribing kernel team, and acpi team. Hope this is not inapropriate.

[Felipe Figueiredo (philsf)]

As noted in http://linux.derkeiler.com/Mailing-Lists/Kernel/2008-04/msg02644.html the upstream bug 10319 might be related ( http://bugzilla.kernel.org/show_bug.cgi?id=10319 ). Since I'm not sure of that, I'll not add it to the bugwatch list now.

[Felipe Figueiredo (philsf)]

I found a way to make it work! A workaround is to unload the ehci_hcd module. The following command (all in one line) accomplishes this for gnome-power-manager.

echo 'SUSPEND_MODULES="ehci_hcd"' | sudo tee /etc/pm/config.d/lenovo3000v100

Now, I don't understand why this is necessary, since according to /etc/default/acpi-support "network cards and USB controllers will automatically be unloaded unless they're listed in MODULES_WHITELIST". This should already be automatically unloaded, and it's not. Maybe this is another bug.

[TJ (tj)]

Felipe, well done for discovering this. I've confirmed that ehci_hcd is the cause of the failed suspends. It looks as if recent changes to that module have caused this. Bug # 212660 seems to be the best place to track the issue.

A note: although /etc/default/acpi-support says:

# Note that network cards and USB controllers will automatically be unloaded
# unless they're listed in MODULES_WHITELIST
MODULES=""

that functionality was removed (but obviously the comment in the configuration file was mistakenly left in place) by:

acpi-support (0.59) dapper; urgency=low

  * Disable USB module unloading - the kernel seems to manage it better
    now

 -- Matthew Garrett <email address hidden> Sat, 18 Feb 2006 14:25:11 +0000