On (11/19/18 16:18), Greg Kroah-Hartman wrote:
On Sun, Nov 11, 2018 at 08:20:45PM +0000, Sudip Mukherjee wrote:
Hi Greg,
This was not marked for stable but seems it should be in stable. Please apply to your queue of 4.14-stable.
Now queued up, thanks.
Very sorry for the late reply!
Greg, Sudip, the commit in question is known to be controversial. It does fix some lockups, but it also does make printk non-atomic in some cases: the printing task can get preempted which can cause printk stalls (no messages on serial consoles, until the printing task gets rescheduled again) in some dark-corner cases.
I think Tetsuo is the only person who ever reported printk stalls, probably because he is the only person who is testing very tough OOM-scenarios on a regular basis.
So, long story short, I call that commit "a mistake" and we have reverted it upstream, to make printk always atomic (just like it should be).
As of printk lockups, Steven Rostedt has contributed a much better solution.
-ss