Re: [PATCH] android: binder: Disable preemption while holding the global binder lock

From: Peter Zijlstra
Date: Wed Sep 14 2016 - 12:11:28 EST


On Tue, Sep 13, 2016 at 12:53:27PM -0700, Arve Hjønnevåg wrote:
> On Tue, Sep 13, 2016 at 12:32 AM, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> > cgroups should be irrelevant, PI is unaware of them.
>
> I don't think cgroups are irrelevant. PI being unaware of them
> explains the problem I described. If the task that holds the lock is
> in a cgroup that has a low cpu.shares value, then boosting the task's
> priority within that group does necessarily make it any more likely to
> run.

Thing is, for FIFO/DL the important parameters (prio and deadline resp.)
are not cgroup dependent.

For CFS you're right, and as per usual, cgroups will be a royal pain.
While we can compute the total weight in the block chain, getting that
back to a task which is stuck in a cgroup is just not going to work
well.

The only 'solution' I can come up with in a hurry is, when the task is
boosted, move it to the root cgroup. That of course has a whole heap of
problems all on its own.

/me curses @ cgroups.. bloody stupid things.