1 files changed, 33 insertions, 24 deletions

diff --git a/drivers/block/xen-blkfront.c b/drivers/block/xen-blkfront.c
index 06e2812ba124..657873e4328d 100644
--- a/drivers/block/xen-blkfront.c
+++ b/drivers/block/xen-blkfront.c
@@ -65,14 +65,14 @@ enum blkif_state {
 
 struct blk_shadow {
         struct blkif_request req;
-        unsigned long request;
+        struct request *request;
         unsigned long frame[BLKIF_MAX_SEGMENTS_PER_REQUEST];
 };
 
 static DEFINE_MUTEX(blkfront_mutex);
 static const struct block_device_operations xlvbd_block_fops;
 
-#define BLK_RING_SIZE __RING_SIZE((struct blkif_sring *)0, PAGE_SIZE)
+#define BLK_RING_SIZE __CONST_RING_SIZE(blkif, PAGE_SIZE)
 
 /*
  * We have one of these per vbd, whether ide, scsi or 'other'.  They
@@ -136,7 +136,7 @@ static void add_id_to_freelist(struct blkfront_info *info,
                                unsigned long id)
 {
         info->shadow[id].req.id  = info->shadow_free;
-        info->shadow[id].request = 0;
+        info->shadow[id].request = NULL;
         info->shadow_free = id;
 }
 
@@ -245,14 +245,11 @@ static int blkif_ioctl(struct block_device *bdev, fmode_t mode,
 }
 
 /*
- * blkif_queue_request
+ * Generate a Xen blkfront IO request from a blk layer request.  Reads
+ * and writes are handled as expected.  Since we lack a loose flush
+ * request, we map flushes into a full ordered barrier.
  *
- * request block io
- *
- * id: for guest use only.
- * operation: BLKIF_OP_{READ,WRITE,PROBE}
- * buffer: buffer to read/write into. this should be a
- *   virtual address in the guest os.
+ * @req: a request struct
  */
 static int blkif_queue_request(struct request *req)
 {
@@ -281,7 +278,7 @@ static int blkif_queue_request(struct request *req)
         /* Fill out a communications ring structure. */
         ring_req = RING_GET_REQUEST(&info->ring, info->ring.req_prod_pvt);
         id = get_id_from_freelist(info);
-        info->shadow[id].request = (unsigned long)req;
+        info->shadow[id].request = req;
 
         ring_req->id = id;
         ring_req->sector_number = (blkif_sector_t)blk_rq_pos(req);
@@ -289,8 +286,18 @@ static int blkif_queue_request(struct request *req)
 
         ring_req->operation = rq_data_dir(req) ?
                 BLKIF_OP_WRITE : BLKIF_OP_READ;
-        if (req->cmd_flags & REQ_HARDBARRIER)
+
+        if (req->cmd_flags & (REQ_FLUSH | REQ_FUA)) {
+                /*
+                 * Ideally we could just do an unordered
+                 * flush-to-disk, but all we have is a full write
+                 * barrier at the moment.  However, a barrier write is
+                 * a superset of FUA, so we can implement it the same
+                 * way.  (It's also a FLUSH+FUA, since it is
+                 * guaranteed ordered WRT previous writes.)
+                 */
                 ring_req->operation = BLKIF_OP_WRITE_BARRIER;
+        }
 
         ring_req->nr_segments = blk_rq_map_sg(req->q, req, info->sg);
         BUG_ON(ring_req->nr_segments > BLKIF_MAX_SEGMENTS_PER_REQUEST);
@@ -636,7 +643,7 @@ static irqreturn_t blkif_interrupt(int irq, void *dev_id)
 
                 bret = RING_GET_RESPONSE(&info->ring, i);
                 id   = bret->id;
-                req  = (struct request *)info->shadow[id].request;
+                req  = info->shadow[id].request;
 
                 blkif_completion(&info->shadow[id]);
 
@@ -649,6 +656,16 @@ static irqreturn_t blkif_interrupt(int irq, void *dev_id)
                                 printk(KERN_WARNING "blkfront: %s: write barrier op failed\n",
                                        info->gd->disk_name);
                                 error = -EOPNOTSUPP;
+                        }
+                        if (unlikely(bret->status == BLKIF_RSP_ERROR &&
+                                     info->shadow[id].req.nr_segments == 0)) {
+                                printk(KERN_WARNING "blkfront: %s: empty write barrier op failed\n",
+                                       info->gd->disk_name);
+                                error = -EOPNOTSUPP;
+                        }
+                        if (unlikely(error)) {
+                                if (error == -EOPNOTSUPP)
+                                        error = 0;
                                 info->feature_flush = 0;
                                 xlvbd_flush(info);
                         }
@@ -901,7 +918,7 @@ static int blkif_recover(struct blkfront_info *info)
         /* Stage 3: Find pending requests and requeue them. */
         for (i = 0; i < BLK_RING_SIZE; i++) {
                 /* Not in use? */
-                if (copy[i].request == 0)
+                if (!copy[i].request)
                         continue;
 
                 /* Grab a request slot and copy shadow state into it. */
@@ -918,9 +935,7 @@ static int blkif_recover(struct blkfront_info *info)
                                 req->seg[j].gref,
                                 info->xbdev->otherend_id,
                                 pfn_to_mfn(info->shadow[req->id].frame[j]),
-                                rq_data_dir(
-                                        (struct request *)
-                                        info->shadow[req->id].request));
+                                rq_data_dir(info->shadow[req->id].request));
                 info->shadow[req->id].req = *req;
 
                 info->ring.req_prod_pvt++;
@@ -1069,14 +1084,8 @@ static void blkfront_connect(struct blkfront_info *info)
          */
         info->feature_flush = 0;
 
-        /*
-         * The driver doesn't properly handled empty flushes, so
-         * lets disable barrier support for now.
-         */
-#if 0
         if (!err && barrier)
-                info->feature_flush = REQ_FLUSH;
-#endif
+                info->feature_flush = REQ_FLUSH | REQ_FUA;
 
         err = xlvbd_alloc_gendisk(sectors, info, binfo, sector_size);
         if (err) {