Binary package hint: linux-image-generic
The following problem started with kernel 2.6.26-20 and still persists with 2.6.26-21 on (K)Ubuntu Hardy 8.04.1, both installed via apt from the official "proposed" repository (currently linux-image-generic 2.6.24.21.23).
Returning to my desktop in the morning, I'm not able to unlock it: While the system still runs, everything works - but the keyboard is dead. No input possible. Logging in remotely via SSH is still possible, so I was able to capture the dmesg ringbuffer (see attachment: crash.log).
What I tried to avoid a reboot:
* disconnected keyboard (PS2) and reconnected it
* disconnected keyboard (PS2) and connected another one (PS2,USB)
* switched completely to USB keyboard, on crash tried rmmod usbhid && rmmod hid && modprobe usbhid
None of these revived the keyboard, since it was the code built into the kernel that crashed. Since according to the log Bluetooth was involved (BlueProximity was running to be able to [un]lock desktop with my phone, which only worked occasionally), I now closed BlueProximity and removed the BT stick to see whether that helps. Will report results then.
As the summary says, these Oopses happen almost daily - but there have been a few days they didn't occur. So I will leave BT off for at least a week or until the next Oops, whatever happens first, and update this bug with the results then.
PS: I tried to switch back to the latest "stable" version (2.6.24.19.21 from hardy-security), but when I booted that the graphic srewed up: It could no longer handle the nvidia card, though the suitable linux-restricted-modules are installed. I would be happy to get rid of the "proposed" repo stuff (it's only in there since the Wiki said it's needed for dist-upgrade from gutsy), so advices for this are also welcome.
As I wrote above, I closed all BT apps and even removed the Bluetooth stick to check whether Bluetooth may be the cause. Now, 7 days later, it looks like I can approve this: Since I've done so, no more Oops appeared, and the keyboard is still working fine - so the bug must have to do with Bluetooth.
Would be nice if somebody can confirm this - I'm not in kernel development at all, so I don't know where and what to look for.