[PATCH 2/4 V2] block/rbd: implement bdrv_aio_compare_and_write interface

[Yaowei Bai]

This patch adds librbd's SCSI COMPARE_AND_WRITE command interface
support with bdrv_aio_compare_and_write function pointer. Note
currently when a miscompare happens a mismatch offset of 0 is
always reported rather than the actual mismatch offset. This
should not be a big issue contemporarily and will be fixed
accordingly in the future.

Signed-off-by: Yaowei Bai <address@hidden>
---
 block/raw-format.c |  3 ++-
 block/rbd.c        | 45 +++++++++++++++++++++++++++++++++++++++++++--
 2 files changed, 45 insertions(+), 3 deletions(-)

diff --git a/block/raw-format.c b/block/raw-format.c
index 3a76ec7..e87cd44 100644
--- a/block/raw-format.c
+++ b/block/raw-format.c
@@ -439,7 +439,8 @@ static int raw_open(BlockDriverState *bs, QDict *options, 
int flags,
 
     bs->sg = bs->file->bs->sg;
     bs->supported_write_flags = BDRV_REQ_WRITE_UNCHANGED |
-        (BDRV_REQ_FUA & bs->file->bs->supported_write_flags);
+        (BDRV_REQ_FUA & bs->file->bs->supported_write_flags) |
+        (BDRV_REQ_COMPARE_AND_WRITE & bs->file->bs->supported_write_flags);
     bs->supported_zero_flags = BDRV_REQ_WRITE_UNCHANGED |
         ((BDRV_REQ_FUA | BDRV_REQ_MAY_UNMAP | BDRV_REQ_NO_FALLBACK) &
             bs->file->bs->supported_zero_flags);
diff --git a/block/rbd.c b/block/rbd.c
index 027cbcc..0e45bc3 100644
--- a/block/rbd.c
+++ b/block/rbd.c
@@ -73,11 +73,18 @@
 #define LIBRBD_USE_IOVEC 0
 #endif
 
+#ifdef LIBRBD_SUPPORTS_COMPARE_AND_WRITE
+#define LIBRBD_HAVE_COMPARE_AND_WRITE 1
+#else
+#define LIBRBD_HAVE_COMPARE_AND_WRITE 0
+#endif
+
 typedef enum {
     RBD_AIO_READ,
     RBD_AIO_WRITE,
     RBD_AIO_DISCARD,
-    RBD_AIO_FLUSH
+    RBD_AIO_FLUSH,
+    RBD_AIO_COMPARE_AND_WRITE
 } RBDAIOCmd;
 
 typedef struct RBDAIOCB {
@@ -798,6 +805,10 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict 
*options, int flags,
         }
     }
 
+    if (LIBRBD_HAVE_COMPARE_AND_WRITE) {
+        bs->supported_write_flags = BDRV_REQ_COMPARE_AND_WRITE;
+    }
+
     r = 0;
     goto out;
 
@@ -933,7 +944,15 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
 
     rcb = g_new(RADOSCB, 1);
 
-    if (!LIBRBD_USE_IOVEC) {
+    if (cmd == RBD_AIO_COMPARE_AND_WRITE) {
+        acb->bounce = qemu_try_blockalign(bs, qiov->size);
+        if (acb->bounce == NULL) {
+            goto failed;
+        }
+
+        qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size);
+        rcb->buf = acb->bounce;
+    } else if (!LIBRBD_USE_IOVEC) {
         if (cmd == RBD_AIO_DISCARD || cmd == RBD_AIO_FLUSH) {
             acb->bounce = NULL;
         } else {
@@ -993,6 +1012,10 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
     case RBD_AIO_FLUSH:
         r = rbd_aio_flush_wrapper(s->image, c);
         break;
+    case RBD_AIO_COMPARE_AND_WRITE:
+        r = rbd_aio_compare_and_write(s->image, off, size / 2, rcb->buf,
+                                     (rcb->buf + size / 2), c, 0, 0);
+        break;
     default:
         r = -EINVAL;
     }
@@ -1056,6 +1079,20 @@ static int qemu_rbd_co_flush(BlockDriverState *bs)
 }
 #endif
 
+#ifdef LIBRBD_SUPPORTS_COMPARE_AND_WRITE
+static BlockAIOCB *qemu_rbd_aio_compare_and_write(BlockDriverState *bs,
+                                                  uint64_t offset,
+                                                  uint64_t bytes,
+                                                  QEMUIOVector *qiov,
+                                                  int flags,
+                                                  BlockCompletionFunc *cb,
+                                                  void *opaque)
+{
+    return rbd_start_aio(bs, offset, qiov, bytes, cb, opaque,
+                         RBD_AIO_COMPARE_AND_WRITE);
+}
+#endif
+
 static int qemu_rbd_getinfo(BlockDriverState *bs, BlockDriverInfo *bdi)
 {
     BDRVRBDState *s = bs->opaque;
@@ -1310,6 +1347,10 @@ static BlockDriver bdrv_rbd = {
     .bdrv_aio_pdiscard      = qemu_rbd_aio_pdiscard,
 #endif
 
+#ifdef LIBRBD_SUPPORTS_COMPARE_AND_WRITE
+    .bdrv_aio_compare_and_write = qemu_rbd_aio_compare_and_write,
+#endif
+
     .bdrv_snapshot_create   = qemu_rbd_snap_create,
     .bdrv_snapshot_delete   = qemu_rbd_snap_remove,
     .bdrv_snapshot_list     = qemu_rbd_snap_list,
-- 
1.8.3.1