[Qemu-devel] [RFC PATCH 36/36] mirror: allow customizing the granularity

[Paolo Bonzini <pbonzini@redhat.com>]

The desired granularity may be very different depending on the kind of operation
(e.g. continous replication vs. collapse-to-raw) and whether the VM is expected
to perform lots of I/O while mirroring is in progress.  Allow the user to customize
it.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 block/mirror.c   |   38 +++++++++++++++++++-------------------
 block_int.h      |    3 ++-
 blockdev.c       |    7 ++++++-
 hmp.c            |    2 +-
 qapi-schema.json |    5 ++++-
 qmp-commands.hx  |    4 +++-
 6 files changed, 35 insertions(+), 24 deletions(-)

diff --git a/block/mirror.c b/block/mirror.c
index e09beaf..4e6aa81 100644
--- a/block/mirror.c
+++ b/block/mirror.c
@@ -17,9 +17,6 @@
 #include "qemu/ratelimit.h"
 #include "bitmap.h"
 
-#define BLOCK_SIZE                       (1 << 20)
-#define BDRV_SECTORS_PER_DIRTY_CHUNK     (BLOCK_SIZE >> BDRV_SECTOR_BITS)
-
 #define SLICE_TIME 100000000ULL /* ns */
 
 typedef struct MirrorBlockJob {
@@ -31,6 +28,7 @@ typedef struct MirrorBlockJob {
     bool synced;
     bool complete;
     int64_t sector_num;
+    int64_t granularity;
     int64_t buf_size;
     unsigned long *cow_bitmap;
     HBitmapIter hbi;
@@ -43,7 +41,7 @@ static int coroutine_fn mirror_iteration(MirrorBlockJob *s,
     BlockDriverState *source = s->common.bs;
     BlockDriverState *target = s->target;
     QEMUIOVector qiov;
-    int ret, nb_sectors;
+    int ret, nb_sectors, nb_sectors_chunk;
     int64_t end, sector_num, cluster_num;
     struct iovec iov;
 
@@ -58,24 +56,24 @@ static int coroutine_fn mirror_iteration(MirrorBlockJob *s,
      * is very large, we need to do COW ourselves.  The first time a cluster is
      * copied, copy it entirely.
      *
-     * Because both BDRV_SECTORS_PER_DIRTY_CHUNK and the cluster size are
-     * powers of two, the number of sectors to copy cannot exceed one cluster.
+     * Because both the granularity and the cluster size are powers of two, the
+     * number of sectors to copy cannot exceed one cluster.
      */
     sector_num = s->sector_num;
-    nb_sectors = BDRV_SECTORS_PER_DIRTY_CHUNK;
-    cluster_num = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
+    nb_sectors_chunk = nb_sectors = s->granularity >> BDRV_SECTOR_BITS;
+    cluster_num = sector_num / nb_sectors_chunk;
     if (s->cow_bitmap && !test_bit(cluster_num, s->cow_bitmap)) {
         bdrv_round_to_clusters(s->target,
-                               sector_num, BDRV_SECTORS_PER_DIRTY_CHUNK,
+                               sector_num, nb_sectors_chunk,
                                &sector_num, &nb_sectors);
-        bitmap_set(s->cow_bitmap, sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK,
-                   nb_sectors / BDRV_SECTORS_PER_DIRTY_CHUNK);
+        bitmap_set(s->cow_bitmap, sector_num / nb_sectors_chunk,
+                   nb_sectors / nb_sectors_chunk);
     }
 
     end = s->common.len >> BDRV_SECTOR_BITS;
     nb_sectors = MIN(nb_sectors, end - sector_num);
     trace_mirror_one_iteration(s, sector_num);
-    bdrv_reset_dirty(source, sector_num, BDRV_SECTORS_PER_DIRTY_CHUNK);
+    bdrv_reset_dirty(source, sector_num, nb_sectors);
 
     /* Copy the dirty cluster.  */
     iov.iov_base = s->buf;
@@ -107,7 +105,7 @@ static void coroutine_fn mirror_run(void *opaque)
 {
     MirrorBlockJob *s = opaque;
     BlockDriverState *bs = s->common.bs;
-    int64_t sector_num, end;
+    int64_t sector_num, end, nb_sectors_chunk;
     int ret = 0;
     int n;
 
@@ -123,13 +121,14 @@ static void coroutine_fn mirror_run(void *opaque)
 
     end = s->common.len >> BDRV_SECTOR_BITS;
     s->buf = qemu_blockalign(bs, s->buf_size);
+    nb_sectors_chunk = s->granularity >> BDRV_SECTOR_BITS;
 
     if (s->mode == MIRROR_SYNC_MODE_FULL || s->mode == MIRROR_SYNC_MODE_TOP) {
         /* First part, loop on the sectors and initialize the dirty bitmap.  */
         BlockDriverState *base;
         base = s->mode == MIRROR_SYNC_MODE_FULL ? NULL : bs->backing_hd;
         for (sector_num = 0; sector_num < end; ) {
-            int64_t next = (sector_num | (BDRV_SECTORS_PER_DIRTY_CHUNK - 1)) + 1;
+            int64_t next = (sector_num | (nb_sectors_chunk - 1)) + 1;
             ret = bdrv_co_is_allocated_above(bs, base,
                                              sector_num, next - sector_num, &n);
 
@@ -209,7 +208,7 @@ static void coroutine_fn mirror_run(void *opaque)
             s->common.offset = (end - cnt) * BDRV_SECTOR_SIZE;
 
             if (s->common.speed) {
-                delay_ns = ratelimit_calculate_delay(&s->limit, BDRV_SECTORS_PER_DIRTY_CHUNK);
+                delay_ns = ratelimit_calculate_delay(&s->limit, nb_sectors_chunk);
             } else {
                 delay_ns = 0;
             }
@@ -301,7 +300,7 @@ static BlockJobType mirror_job_type = {
 };
 
 void mirror_start(BlockDriverState *bs, BlockDriverState *target,
-                  int64_t speed, MirrorSyncMode mode,
+                  int64_t speed, int64_t granularity, MirrorSyncMode mode,
                   BlockdevOnError on_source_error,
                   BlockdevOnError on_target_error,
                   BlockDriverCompletionFunc *cb,
@@ -326,19 +325,20 @@ void mirror_start(BlockDriverState *bs, BlockDriverState *target,
      * the destination do COW.  Instead, we copy sectors around the
      * dirty data if needed.
      */
-    s->buf_size = BLOCK_SIZE;
+    s->granularity = granularity;
+    s->buf_size = granularity;
     bdrv_get_backing_filename(s->target, backing_filename,
                               sizeof(backing_filename));
     if (backing_filename[0] && !s->target->backing_hd) {
         bdrv_get_info(s->target, &bdi);
         if (s->buf_size < bdi.cluster_size) {
             s->buf_size = bdi.cluster_size;
-            length = (bdrv_getlength(bs) + BLOCK_SIZE - 1) / BLOCK_SIZE;
+            length = (bdrv_getlength(bs) + granularity - 1) / granularity;
             s->cow_bitmap = bitmap_new(length);
         }
     }
 
-    bdrv_set_dirty_tracking(bs, BDRV_SECTORS_PER_DIRTY_CHUNK);
+    bdrv_set_dirty_tracking(bs, granularity >> BDRV_SECTOR_BITS);
     bdrv_set_on_error(s->target, on_target_error, on_target_error);
     bdrv_iostatus_enable(s->target);
     s->common.co = qemu_coroutine_create(mirror_run);
diff --git a/block_int.h b/block_int.h
index 6d36fe7..8bbe479 100644
--- a/block_int.h
+++ b/block_int.h
@@ -311,6 +311,7 @@ void stream_start(BlockDriverState *bs, BlockDriverState *base,
  * @bs: Block device to operate on.
  * @target: Block device to write to.
  * @speed: The maximum speed, in bytes per second, or 0 for unlimited.
+ * @granularity: The chosen granularity for the dirty bitmap.
  * @mode: Whether to collapse all images in the chain to the target.
  * @on_source_error: The action to take upon error reading from the source.
  * @on_target_error: The action to take upon error writing to the target.
@@ -324,7 +325,7 @@ void stream_start(BlockDriverState *bs, BlockDriverState *base,
  * @bs will be switched to read from @target.
  */
 void mirror_start(BlockDriverState *bs, BlockDriverState *target,
-                  int64_t speed, MirrorSyncMode mode,
+                  int64_t speed, int64_t granularity, MirrorSyncMode mode,
                   BlockdevOnError on_source_error,
                   BlockdevOnError on_target_error,
                   BlockDriverCompletionFunc *cb,
diff --git a/blockdev.c b/blockdev.c
index f940e8f..a1b2b1b 100644
--- a/blockdev.c
+++ b/blockdev.c
@@ -836,6 +836,7 @@ void qmp_drive_mirror(const char *device, const char *target,
                       enum MirrorSyncMode sync,
                       bool has_mode, enum NewImageMode mode,
                       bool has_speed, int64_t speed,
+                      bool has_granularity, int64_t granularity,
                       bool has_on_source_error, BlockdevOnError on_source_error,
                       bool has_on_target_error, BlockdevOnError on_target_error,
                       Error **errp)
@@ -858,6 +859,9 @@ void qmp_drive_mirror(const char *device, const char *target,
     if (!has_mode) {
         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
     }
+    if (!has_granularity) {
+        granularity = 65536;
+    }
 
     bs = bdrv_find(device);
     if (!bs) {
@@ -936,7 +940,8 @@ void qmp_drive_mirror(const char *device, const char *target,
         return;
     }
 
-    mirror_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
+    mirror_start(bs, target_bs, speed, granularity, sync,
+                 on_source_error, on_target_error,
                  block_job_cb, bs, &local_err);
     if (local_err != NULL) {
         bdrv_delete(target_bs);
diff --git a/hmp.c b/hmp.c
index ef0b87f..f01d608 100644
--- a/hmp.c
+++ b/hmp.c
@@ -717,7 +717,7 @@ void hmp_drive_mirror(Monitor *mon, const QDict *qdict)
 
     qmp_drive_mirror(device, filename, !!format, format,
                      full ? MIRROR_SYNC_MODE_FULL : MIRROR_SYNC_MODE_TOP,
-                     true, mode, false, 0,
+                     true, mode, false, 0, false, 0,
                      false, 0, false, 0, &errp);
     hmp_handle_error(mon, &errp);
 }
diff --git a/qapi-schema.json b/qapi-schema.json
index bbcfa0e..b579581 100644
--- a/qapi-schema.json
+++ b/qapi-schema.json
@@ -1383,6 +1383,8 @@
 #        (all the disk, only the sectors allocated in the topmost image, or
 #        only new I/O).
 #
+# @granularity: #optional granularity of the dirty bitmap, default is 64K.
+#
 # @on_source_error: #optional the action to take on an error on the source
 #
 # @on_target_error: #optional the action to take on an error on the target
@@ -1397,7 +1399,8 @@
 { 'command': 'drive-mirror',
   'data': { 'device': 'str', 'target': 'str', '*format': 'str',
             'sync': 'MirrorSyncMode', '*mode': 'NewImageMode',
-            '*speed': 'int', '*on_source_error': 'BlockdevOnError',
+            '*speed': 'int', '*granularity': 'int',
+            '*on_source_error': 'BlockdevOnError',
             '*on_target_error': 'BlockdevOnError' } }
 
 ##
diff --git a/qmp-commands.hx b/qmp-commands.hx
index f1b0f90..40ef5ab 100644
--- a/qmp-commands.hx
+++ b/qmp-commands.hx
@@ -839,7 +839,8 @@ EQMP
     {
         .name       = "drive-mirror",
         .args_type  = "sync:s,device:B,target:s,sync:s?,format:s?,"
-                      "on_source_error:s?,on_target_error:s?",
+                      "on_source_error:s?,on_target_error:s?,"
+                      "granularity:i?",
         .mhandler.cmd_new = qmp_marshal_input_drive_mirror,
     },
 
@@ -862,6 +863,7 @@ Arguments:
   file/device (NewImageMode, optional, default 'absolute-paths')
 - "speed": maximum speed of the streaming job, in bytes per second
   (json-int)
+- "granularity": granularity of the dirty bitmap (json-int, default 64k)
 - "sync": what parts of the disk image should be copied to the destination;
   possibilities include "full" for all the disk, "top" for only the sectors
   allocated in the topmost image, or "none" to only replicate new I/O
-- 
1.7.10.2