RDMA/umem: Revert broken 'off by one' fix

The previous attempted bug fix overlooked the fact that
ib_umem_odp_map_dma_single_page() was doing a put_page() upon hitting an
error. So there was not really a bug there.

Therefore, this reverts the off-by-one change, but keeps the change to use
release_pages() in the error path.

Fixes: 75a3e6a3c129 ("RDMA/umem: minor bug fix in error handling path")
Suggested-by: Artemy Kovalyov <artemyko@mellanox.com>
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

diff --git a/drivers/infiniband/core/umem_odp.c b/drivers/infiniband/core/umem_odp.c
index 577f1b1..e6ec79a 100644 (file)
--- a/drivers/infiniband/core/umem_odp.c
+++ b/drivers/infiniband/core/umem_odp.c
@@ -687,10 +687,13 @@ int ib_umem_odp_map_dma_pages(struct ib_umem_odp *umem_odp, u64 user_virt,
 
                if (ret < 0) {
                        /*
-                        * Release pages, starting at the the first page
-                        * that experienced an error.
+                        * Release pages, remembering that the first page
+                        * to hit an error was already released by
+                        * ib_umem_odp_map_dma_single_page().
                         */
-                       release_pages(&local_page_list[j], npages - j);
+                       if (npages - (j + 1) > 0)
+                               release_pages(&local_page_list[j+1],
+                                             npages - (j + 1));
                        break;
                }
        }