On 03/17/2015 09:10 PM, Paul Gortmaker wrote:

[[PATCH RT 26/39] scheduling while atomic in cgroup code] On 12/03/2015 (Thu 15:13) Steven Rostedt wrote:

quoted

3.14.34-rt32-rc1 stable review patch.
If anyone has any objections, please let me know.

------------------

From: Mike Galbraith <redacted>

mm, memcg: make refill_stock() use get_cpu_light()

This looks like only 1/2 of Mike's original patch:

https://lkml.org/lkml/2014/6/21/11 

I suspect that is because 3.18 could only use 1/2 of it, and based on
the SOB lines, this is backported from 3.18.

no. I didn't take the upper chunk because get_cpu_var() does only a
preempt_disable() and in that section there is not point of preemption
or a lock in involved. So it is fine the way it is.
The suggest change does a migrate_disable() which is a little more code.

The lower part (the chunk I applied) invokes drain_stock() and that is
where the sleeping-while-atomic warning came from.

So is the upper half really required and if so, why?

The other half applies to 3.14 -- testing in progress; not sure about
the 3.10-rt and earlier....

P.

Sebastian