IB core: Fix ib_sg_to_pages()
authorBart Van Assche <bart.vanassche@sandisk.com>
Fri, 4 Dec 2015 00:04:17 +0000 (16:04 -0800)
committerDoug Ledford <dledford@redhat.com>
Mon, 7 Dec 2015 22:20:12 +0000 (17:20 -0500)
On 12/03/2015 01:18 AM, Christoph Hellwig wrote:
> The patch looks good to me, but while we touch this area, how about
> throwing in a few cosmetic fixes as well?

How about the patch below ? In that version of the ib_sg_to_pages() fix
these concerns have been addressed and additionally to more bugs have been fixed.

------------

[PATCH] IB core: Fix ib_sg_to_pages()

Fix the code for detecting gaps. A gap occurs not only if the
second or later scatterlist element is not aligned but also if
any scatterlist element other than the last does not end at a
page boundary.

In the code for coalescing contiguous elements, ensure that
mr->length is correct and that last_page_addr is up-to-date.

Ensure that this function returns a negative
error code instead of zero if the first set_page() call fails.

Fixes: commit 4c67e2bfc8b7 ("IB/core: Introduce new fast registration API")
Reported-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
drivers/infiniband/core/verbs.c

index 043a60ee6836db72e498410ebd80e004ed220c7d..545906dec26dc5e936df37c635c68b94bb27e226 100644 (file)
@@ -1516,7 +1516,7 @@ EXPORT_SYMBOL(ib_map_mr_sg);
  * @sg_nents:      number of entries in sg
  * @set_page:      driver page assignment function pointer
  *
- * Core service helper for drivers to covert the largest
+ * Core service helper for drivers to convert the largest
  * prefix of given sg list to a page vector. The sg list
  * prefix converted is the prefix that meet the requirements
  * of ib_map_mr_sg.
@@ -1533,7 +1533,7 @@ int ib_sg_to_pages(struct ib_mr *mr,
        u64 last_end_dma_addr = 0, last_page_addr = 0;
        unsigned int last_page_off = 0;
        u64 page_mask = ~((u64)mr->page_size - 1);
-       int i;
+       int i, ret;
 
        mr->iova = sg_dma_address(&sgl[0]);
        mr->length = 0;
@@ -1544,27 +1544,29 @@ int ib_sg_to_pages(struct ib_mr *mr,
                u64 end_dma_addr = dma_addr + dma_len;
                u64 page_addr = dma_addr & page_mask;
 
-               if (i && page_addr != dma_addr) {
-                       if (last_end_dma_addr != dma_addr) {
-                               /* gap */
-                               goto done;
-
-                       } else if (last_page_off + dma_len <= mr->page_size) {
-                               /* chunk this fragment with the last */
-                               mr->length += dma_len;
-                               last_end_dma_addr += dma_len;
-                               last_page_off += dma_len;
-                               continue;
-                       } else {
-                               /* map starting from the next page */
-                               page_addr = last_page_addr + mr->page_size;
-                               dma_len -= mr->page_size - last_page_off;
-                       }
+               /*
+                * For the second and later elements, check whether either the
+                * end of element i-1 or the start of element i is not aligned
+                * on a page boundary.
+                */
+               if (i && (last_page_off != 0 || page_addr != dma_addr)) {
+                       /* Stop mapping if there is a gap. */
+                       if (last_end_dma_addr != dma_addr)
+                               break;
+
+                       /*
+                        * Coalesce this element with the last. If it is small
+                        * enough just update mr->length. Otherwise start
+                        * mapping from the next page.
+                        */
+                       goto next_page;
                }
 
                do {
-                       if (unlikely(set_page(mr, page_addr)))
-                               goto done;
+                       ret = set_page(mr, page_addr);
+                       if (unlikely(ret < 0))
+                               return i ? : ret;
+next_page:
                        page_addr += mr->page_size;
                } while (page_addr < end_dma_addr);
 
@@ -1574,7 +1576,6 @@ int ib_sg_to_pages(struct ib_mr *mr,
                last_page_off = end_dma_addr & ~page_mask;
        }
 
-done:
        return i;
 }
 EXPORT_SYMBOL(ib_sg_to_pages);