As reported by Rajat Jain, there are problems when ACPI operation
region handlers or similar, called at the ->resume_early() time, for
I2C client devices try to access an I2C controller that has already
been suspended at that point.  To avoid that, move the suspend/resume
of i2c-designware-platdrv to the late/early stages, respectively.

While at it, avoid resuming the device from runtime suspend in the
driver's ->suspend callback which isn't particularly nice.  [A better
approach would be to make the driver track the PM state of the device
so that it doesn't need to resume it in ->suspend, so implement it.]

First, drop dw_i2c_plat_suspend() added by commit a23318fe

 (i2c:
designware: Fix system suspend) and rename dw_i2c_plat_runtime_suspend()
back to dw_i2c_plat_suspend().

Second, point the driver's ->late_suspend and ->early_resume
callbacks, rather than its ->suspend and ->resume callbacks,
to dw_i2c_plat_suspend() and dw_i2c_plat_resume(), respectively,
so that they are not executed in parallel with each other, for
example if runtime resume of the device takes place during system
suspend.

Finally, add "suspended" and "skip_resume" flags to struct dw_i2c_dev
and make dw_i2c_plat_suspend() and dw_i2c_plat_resume() use them to
avoid suspending or resuming the device twice in a row and to avoid
resuming a previously runtime-suspended device during system resume.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Tested-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Tested-by: Johannes Stezenbach <js@sig21.net>
Tested-by: Rajat Jain <rajatja@google.com>
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>