scsi: pmcraid: fix endianess sparse annotations

The use of le32_to_cpu() etc in this driver looks completely arbitrary.
It may have made sense at some point, but it is not applied consistently,
so this driver presumably won't work on big-endian kernel builds.

Unfortunately it's unclear whether the type names or the calls to
le32_to_cpu() are the correct ones. I'm taking educated guesses here
and assume that most of the __le32 and __le16 annotations are correct,
adding the conversion helpers whereever we access those fields.

The exceptions are the 'fw_version' field that is always accessed as
big-endian, so I'm changing the type here, and the 'hrrq' values that
are accessed as little-endian, so I'm changing those the other way.

None of these changes should have any effect on little-endian
architectures like x86, but it addresses the sparse warnings.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

drivers/scsi/pmcraid.h

@@ -554,7 +554,7 @@ struct pmcraid_inquiry_data {
 	__u8	add_page_len;
 	__u8	length;
 	__u8	reserved2;
-	__le16	fw_version;
+	__be16	fw_version;
 	__u8	reserved3[16];
 };
 
@@ -697,13 +697,13 @@ struct pmcraid_instance {
 	dma_addr_t hrrq_start_bus_addr[PMCRAID_NUM_MSIX_VECTORS];
 
 	/* Pointer to 1st entry of HRRQ */
-	__be32 *hrrq_start[PMCRAID_NUM_MSIX_VECTORS];
+	__le32 *hrrq_start[PMCRAID_NUM_MSIX_VECTORS];
 
 	/* Pointer to last entry of HRRQ */
-	__be32 *hrrq_end[PMCRAID_NUM_MSIX_VECTORS];
+	__le32 *hrrq_end[PMCRAID_NUM_MSIX_VECTORS];
 
 	/* Pointer to current pointer of hrrq */
-	__be32 *hrrq_curr[PMCRAID_NUM_MSIX_VECTORS];
+	__le32 *hrrq_curr[PMCRAID_NUM_MSIX_VECTORS];
 
 	/* Lock for HRRQ access */
 	spinlock_t hrrq_lock[PMCRAID_NUM_MSIX_VECTORS];