error handling in sd_suspend()

Hi,

I've encountered a problem when sd_suspend() races disconnect handling of SATA. This race can lead to needlessly aborted suspensions. To avoid this problem. I suggest evaluating the error returns from the lower layers, to tell us whether a problem is transient or totally unrecoverable. If a transient problem occurs we should not go to sleep and in some cases suggest that the upper layer retry. An unrecoverable error, such as a device going away can be ignored.

Regards
Oliver