[SCSI] be2iscsi: Fix for case where task->sc was cleanedup earlier

This patch fixes a bug where the task->sc was cleaned up earlier on a
different thread, possibly abort, and the completion comes later.

This was causing a crash which has been reported in multiple places

Signed-off-by: Jayamohan Kallickal <jayamohan.kallickal@emulex.com>
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

drivers/scsi/be2iscsi/be_main.c

@@ -1123,7 +1123,12 @@ be_complete_io(struct beiscsi_conn *beiscsi_conn,
 						& SOL_STS_MASK) >> 8);
 	flags = ((psol->dw[offsetof(struct amap_sol_cqe, i_flags) / 32]
 					& SOL_FLAGS_MASK) >> 24) | 0x80;
+	if (!task->sc) {
+		if (io_task->scsi_cmnd)
+			scsi_dma_unmap(io_task->scsi_cmnd);
 
+		return;
+	}
 	task->sc->result = (DID_OK << 16) | status;
 	if (rsp != ISCSI_STATUS_CMD_COMPLETED) {
 		task->sc->result = DID_ERROR << 16;