On Fri, 2013-04-26 at 10:41 +0200, Sebastian Andrzej Siewior wrote:

* Steven Rostedt | 2013-04-11 14:33:34 [-0400]:

quoted

As wait queue locks are notorious for long hold times, we can not
convert them to raw_spin_locks without causing issues with -rt. But
Thomas has created a "simple-wait" structure that uses raw spin locks
which may have been a good fit.

Unfortunately, wait queues are not the only issue, as the mce_notify_irq
also does a schedule_work(), which grabs the workqueue spin locks that
have the exact same issue.

mce_notify_irq() can use simple_waitqueue, no?

Yeah, and I went down that path.

But it also schedules work, which has the issue.

The other issue is that mce_report_event() is scheduling a per-cpu
workqueue (mce_schedule_work) in case of a memory fault. This has the
same issue.

Yeah, that looks like it can be an issue too. I wonder if we can use the
same thread and use flags check what to do. Atomically set the flag for
the function to perform, and then have the thread clear it before doing
the function and only go to sleep when all flags are cleared.

-- Steve