Comment 34 for bug 585765

I had 3 freezes yesterday. In each case the last entry in kern.log was:

do_IRQ: 0.189 No irq handler for vector (irq -1)
(Sometimes the number is 0.175, 0.177).

(Interestingly, there have been no instances of the "BUG: CPU soft lockup" messages in the last two days.

I ran a memory test overnight and it was clean.

I also tested kernel 2.6.31-02063112-generic, 2.6.34-020634-generic, 2.6.35-020635rc1-generic today. All of these froze.

In exasperation, I booted this morning with "noapic nolapic acpi=off" on the grub kernel parameters with current 2.6.32-22-generic. This disables all power management (as I understand it), such that my system is running full-out all the time.

The system has been running solid ever since. Even with torture testing it with a DVD rip in Handbrake, Virtualbox, MythTV recording and commercial flagging, an h264 transcode, and Amarok playing.... all running simultaneously. The DVD rip alone would have been enough to lock the system up within a few minutes without "noapic nolapic acpi=off".

This is not desirable work-around, or a long-term solution, but it at least lets me work for the moment, and allows me to continue helping debug rather than reverting back to karmic.

To re-iterated: this system has been working fine for over a year with karmic and jaunty, with power management. This is a regression with Lucid, and the issue is present with upstream kernels as well.

It would be very nice if someone could assign a more appropriate "High Priority" to this issue and we could get some attention from some devs before this bug report becomes a confused with too many un-related freezes.