Re: 2.6.22-rc1: Broken suspend on SMP with tifm

[Oleg Nesterov <oleg@tv-sign.ru>]

On 05/14, Rafael J. Wysocki wrote:
>
> Hmm, I guess we could add an additional mutex that would only be taken in
> flush_workqueue() and in _cpu_down()/_cpu_up() via workqueue_cpu_callback()
> with CPU_LOCK_ACQUIRE?

This will deadlock if work->func() does flush_workqueue(), because it may
run when _cpu_down() holds this lock (note that it doesn't help if we
re-introduce take_over_work()).

This is a reason why mutex_lock(&workqueue_mutex) was removed from
flush_workqueue().

> It doesn't seem to be a good idea to run flush_workqueue() while CPUs are being
> taken up and down anyway.

We can freeze all tasks :) Otherwise we can't forbid them to call
flush_workqueue().

flush_workqueue() is OK. create/destroy is a problem.

Oleg.