Re: [PATCH v2 0/7] CPU hotplug, cpusets: Fix issues with cpusets handling upon CPU hotplug

[Peter Zijlstra <peterz@infradead.org>]

On Fri, 2012-05-04 at 22:14 +0200, Peter Zijlstra wrote:
> > Also, we need to fix this problem at the CPU Hotplug level itself, and
> > not just for the suspend/resume case. Because, we have had numerous bug
> > reports and people complaining about this issue, in various scenarios,
> > including those that didn't involve suspend/resume.
> 
> NO, absolutely not and I will NAK any and all such nonsense. WTF is a
> cpuset worth if you can run on random other cpus? 

Sorting your cpuset 'problem' isn't nowhere near enough to make hotplug
'safe'. unplug also destroys task_struct::cpus_allowed.

Try it:

 # schedtool -a 2 $$
 # grep Cpus_allowed /proc/self/status
 Cpus_allowed:   000004
 # echo 0 > /sys/devices/system/cpu/cpu2/online
 # grep Cpus_allowed /proc/self/status
 Cpus_allowed:   ffffff


See, hotplug is destructive, it has to be, there's no saying the cpu
will every come back.

So mucking about trying to make cpusets non-destructive is pointless.

The real bug is people using hotplug (for all kinds of stupid stuff) and
expecting anything different.