The datasheet mentions that the suspend mode is toggled by reading the
suspend register. The reading returns value 0xFF if the system was in
suspend mode, otherwise it returns value 0x00.

The bma220_deinit() function does up to 2 reads, in case the device was in
suspend mode, which suggests a level of paranoia that makes the logic in
bma220_suspend() and bma220_resume() look insufficient.

This change implements a bma220_power() function which does up to 2 reads
of the suspend register to make sure that the chip enters a desired
(suspended or normal) mode.

If the transition fails, then -EBUSY is returned.

Since only a reference to SPI device is required, we can remove the
spi_set_drvdata() call and get the SPI device object from the base device
object in the suspend/resume routines.
Signed-off-by: Alexandru Ardelean <aardelean@deviqon.com>
Link: https://lore.kernel.org/r/20210625140137.362282-2-aardelean@deviqon.comSigned-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>