IB/srp: Fix infinite loop when FMR sg[0].offset != 0

Avoid that mapping an sg-list in which the first element has a
non-zero offset triggers an infinite loop when using FMR. This
patch makes the FMR mapping code similar to that of ib_sg_to_pages().

Note: older Mellanox HCAs do not support non-zero offsets for FMR.
See also commit 8c4037b ("IB/srp: always avoid non-zero offsets
into an FMR").

Reported-by: Alex Estrin <alex.estrin@intel.com>
Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>

drivers/infiniband/ulp/srp/ib_srp.c

@@ -1402,7 +1402,9 @@ static int srp_map_sg_entry(struct srp_map_state *state,
 
 	while (dma_len) {
 		unsigned offset = dma_addr & ~dev->mr_page_mask;
-		if (state->npages == dev->max_pages_per_mr || offset != 0) {
+
+		if (state->npages == dev->max_pages_per_mr ||
+		    (state->npages > 0 && offset != 0)) {
 			ret = srp_map_finish_fmr(state, ch);
 			if (ret)
 				return ret;
@@ -1419,12 +1421,12 @@ static int srp_map_sg_entry(struct srp_map_state *state,
 	}
 
 	/*
-	 * If the last entry of the MR wasn't a full page, then we need to
+	 * If the end of the MR is not on a page boundary then we need to
 	 * close it out and start a new one -- we can only merge at page
 	 * boundaries.
 	 */
 	ret = 0;
-	if (len != dev->mr_page_size)
+	if ((dma_addr & ~dev->mr_page_mask) != 0)
 		ret = srp_map_finish_fmr(state, ch);
 	return ret;
 }