So, I've had an exciting evening armed with ftrace and a kernel debugger. I'm questioning my own sanity a bit here, but I'm fairly sure at this point that setting ftrace going and loading the kernel debugging module (I'm using kgdboe here because I'm on AWS) stop whatever the issue is from happening - kswapd0's CPU usage doesn't even increase with them enabled, never mind run out of control. However, it *is* possible to trigger issue and then enable ftrace or load the debug module and break in. I haven't found the debugger super useful, though I did manage to break into kswapd0 while it was doing things at one point, so it's potentially an option if necessary. ftrace looked more useful; I've attached a trace of an arbitrary time slice taken with the issue happening. Happy to do extra work with either if these if anyone has further ideas.
** Attachment added: "ftrace of kswapd0 exhibiting issue" https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1518457/+attachment/4549568/+files/ftrace4 -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1518457 Title: kswapd0 100% CPU usage To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1518457/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
