Comment 5 for bug 1589563

.This is a new laptop with fresh install. I received it only a few days before noticing the swapping issue. Laptop model is previous generation meaning that this hardware has been around for nearly two years.

I haven't tried older kernels if I can find a stable one. Only relevant looking report about very similar problem pointed to bug being present already in 3.19 kernels. If this is same bug then his report about older kernel being stable means I would need to install quite old kernel to find a stable one. But I might setup kernel build and bisection for the broken commit in a few days. But I wanted first to learn how relevant code behaviors in my system before trying to find the bug. That means I need to read thousands lines of code and setup some kernel traces to see how thread, hw access and irq timings interact.

I tested mainline kernel package linux-image-4.7.0-040700rc2-generic_4.7.0-040700rc2.201606051831_amd64.deb

There is some changes to system behavior but still underlying bug is present that makes swapping not working on this laptop.

Changes in 4.7-rc2 kernel:
+ SLUB now reports actively about allocation problems
+ Kernel manages to use about 1G swap space before freezing (so far only one attempt to reproduce) when 4.4 kernel consistently used only about 100M swap before freeze
+ freeze triggers OOM killer very quickly and system automatically recovers after OOM killer removes the offending compiling process.
- Netconsole has become unreliable. There is allocation failures reported from netconsole context that results to lost kernel messages
+ Local kernel log recorded to disk is complete. There is no report from systemd about ring buffer overflows.

Attaching locally recorded kern.log from boot and reproduction attempt.