PM/Suspend: Introduce two new platform callbacks to avoid breakage
authorRafael J. Wysocki <rjw@sisk.pl>
Sun, 19 Apr 2009 18:08:42 +0000 (20:08 +0200)
committerRafael J. Wysocki <rjw@sisk.pl>
Sun, 19 Apr 2009 18:08:42 +0000 (20:08 +0200)
commit6a7c7eaf71b636f197d73b381a2ab729ebdcfb2e
tree73c642bff623c2f1a87f6c353b21e3484bd9e899
parentff54250a0ebab7f90a5f848a0ba63f999830c872
PM/Suspend: Introduce two new platform callbacks to avoid breakage

Commit 900af0d973856d6feb6fc088c2d0d3fde57707d3 (PM: Change suspend
code ordering) changed the ordering of suspend code in such a way
that the platform .prepare() callback is now executed after the
device drivers' late suspend callbacks have run.  Unfortunately, this
turns out to break ARM platforms that need to talk via I2C to power
control devices during the .prepare() callback.

For this reason introduce two new platform suspend callbacks,
.prepare_late() and .wake(), that will be called just prior to
disabling non-boot CPUs and right after bringing them back on line,
respectively, and use them instead of .prepare() and .finish() for
ACPI suspend.  Make the PM core execute the .prepare() and .finish()
platform suspend callbacks where they were executed previously (that
is, right after calling the regular suspend methods provided by
device drivers and right before executing their regular resume
methods, respectively).

It is not necessary to make analogous changes to the hibernation
code and data structures at the moment, because they are only used
by ACPI platforms.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Reported-by: Russell King <rmk+kernel@arm.linux.org.uk>
Acked-by: Len Brown <len.brown@intel.com>
drivers/acpi/sleep.c
include/linux/suspend.h
kernel/power/main.c