From: Paul Mackerras <paulus@samba.org>

Currently the device_pm_foo() functions are rather prone to deadlocks
during suspend/resume.  This is because the dpm_sem is held for the
duration of device_suspend() and device_resume() as well as device_pm_add()
and device_pm_remove().  If for any reason you get a device addition or
removal triggered by a device's suspend or resume code, you get a deadlock.
 (The classic example is a USB host adaptor resuming and discovering that
the mouse you used to have plugged in has gone away.)

This patch fixes the problem by using a separate semaphore, called
dpm_list_sem, to cover the places where we need the device pm lists to be
stable, and by being careful about how we traverse the lists on suspend and
resume.  I have analysed the various cases that can occur and I am
confident that I have handled them all correctly.  I posted this patch
together with a detailed analysis 10 days ago.

Signed-off-by Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <greg@kroah.com>