[Qemu-devel] [PATCH v2 41/45] 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. continuous 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, while providing a sane default so that
in general there will be no extra allocated space in the target compared
to the source.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
        v1->v2: pick granularity in mirror_start, not qmp_drive_mirror.

 block/mirror.c   | 50 ++++++++++++++++++++++++++++++++------------------
 block_int.h      |  3 ++-
 blockdev.c       | 15 ++++++++++++++-
 hmp.c            |  2 +-
 qapi-schema.json |  8 +++++++-
 qmp-commands.hx  |  8 +++++++-
 6 file modificati, 63 inserzioni(+), 23 rimozioni(-)

diff --git a/block/mirror.c b/block/mirror.c
index eba6259..335f17c 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;
     size_t buf_size;
     unsigned long *cow_bitmap;
     HBitmapIter hbi;
@@ -56,7 +54,7 @@ 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;
 
@@ -72,19 +70,19 @@ 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)) {
         trace_mirror_cow(s, sector_num);
         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;
@@ -120,7 +118,7 @@ static void coroutine_fn mirror_run(void *opaque)
 {
     MirrorBlockJob *s = opaque;
     BlockDriverState *bs = s->common.bs;
-    int64_t sector_num, end, length;
+    int64_t sector_num, end, nb_sectors_chunk, length;
     BlockDriverInfo bdi;
     char backing_filename[1024];
     int ret = 0;
@@ -146,20 +144,21 @@ static void coroutine_fn mirror_run(void *opaque)
         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) + s->granularity - 1) / s->granularity;
             s->cow_bitmap = bitmap_new(length);
         }
     }
 
     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_NONE) {
         /* 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);
 
@@ -239,7 +238,7 @@ 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;
             }
@@ -326,7 +325,7 @@ 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,
@@ -334,6 +333,20 @@ void mirror_start(BlockDriverState *bs, BlockDriverState *target,
 {
     MirrorBlockJob *s;
 
+    if (granularity == 0) {
+        /* Choose the default granularity based on the target file's cluster
+         * size, clamped between 4k and 64k.  */
+        BlockDriverInfo bdi;
+        if (bdrv_get_info(target, &bdi) >= 0 && bdi.cluster_size != 0) {
+            granularity = MAX(4096, bdi.cluster_size);
+            granularity = MIN(65536, granularity);
+        } else {
+            granularity = 65536;
+        }
+    }
+
+    assert ((granularity & (granularity - 1)) == 0);
+
     if ((on_source_error == BLOCKDEV_ON_ERROR_STOP ||
          on_source_error == BLOCKDEV_ON_ERROR_ENOSPC) &&
         !bdrv_iostatus_is_enabled(bs)) {
@@ -350,9 +363,10 @@ void mirror_start(BlockDriverState *bs, BlockDriverState *target,
     s->on_target_error = on_target_error;
     s->target = target;
     s->mode = mode;
-    s->buf_size = BLOCK_SIZE;
+    s->granularity = granularity;
+    s->buf_size = granularity;
 
-    bdrv_set_dirty_tracking(bs, BLOCK_SIZE);
+    bdrv_set_dirty_tracking(bs, granularity);
     bdrv_set_enable_write_cache(s->target, true);
     bdrv_set_on_error(s->target, on_target_error, on_target_error);
     bdrv_iostatus_enable(s->target);
diff --git a/block_int.h b/block_int.h
index c09a679..032db74 100644
--- a/block_int.h
+++ b/block_int.h
@@ -320,6 +320,7 @@ 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.
@@ -333,7 +334,7 @@ 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 c989ce6..c53e861 100644
--- a/blockdev.c
+++ b/blockdev.c
@@ -1122,6 +1122,7 @@ 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, uint32_t granularity,
                       bool has_on_source_error, BlockdevOnError on_source_error,
                       bool has_on_target_error, BlockdevOnError on_target_error,
                       Error **errp)
@@ -1147,6 +1148,17 @@ void qmp_drive_mirror(const char *device, const char *target,
     if (!has_mode) {
         mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
     }
+    if (!has_granularity) {
+        granularity = 0;
+    }
+    if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
+        error_set(errp, QERR_INVALID_PARAMETER, device);
+        return;
+    }
+    if (granularity & (granularity - 1)) {
+        error_set(errp, QERR_INVALID_PARAMETER, device);
+        return;
+    }
 
     bs = bdrv_find(device);
     if (!bs) {
@@ -1228,7 +1240,8 @@ 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 b4d2736..12d1413 100644
--- a/hmp.c
+++ b/hmp.c
@@ -783,7 +783,7 @@ 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 dc28685..1b9a962 100644
--- a/qapi-schema.json
+++ b/qapi-schema.json
@@ -1553,6 +1553,11 @@
 #        (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
+#               if the image format doesn't have clusters, 4K if the clusters
+#               are smaller than that, else the cluster size.  Must be a
+#               power of 2 between 512 and 64M.
+#
 # @on-source-error: #optional the action to take on an error on the source,
 #                   default 'report'.  'stop' and 'enospc' can only be used
 #                   if the block device supports io-status (see BlockInfo).
@@ -1569,7 +1574,8 @@
 { 'command': 'drive-mirror',
   'data': { 'device': 'str', 'target': 'str', '*format': 'str',
             'sync': 'MirrorSyncMode', '*mode': 'NewImageMode',
-            '*speed': 'int', '*on-source-error': 'BlockdevOnError',
+            '*speed': 'int', '*granularity': 'uint32',
+            '*on-source-error': 'BlockdevOnError',
             '*on-target-error': 'BlockdevOnError' } }
 
 ##
diff --git a/qmp-commands.hx b/qmp-commands.hx
index ec97eaa..e48ef5e 100644
--- a/qmp-commands.hx
+++ b/qmp-commands.hx
@@ -908,7 +908,8 @@ EQMP
     {
         .name       = "drive-mirror",
         .args_type  = "sync:s,device:B,target:s,speed:i?,mode: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,
     },
 
@@ -932,6 +933,7 @@ 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, optional)
 - "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
@@ -941,6 +943,10 @@ Arguments:
 - "on-target-error": the action to take on an error on the target
   (BlockdevOnError, default 'report')
 
+The default value of the granularity is, if the image format defines
+a cluster size, the cluster size or 4096, whichever is larger.  If it
+does not define a cluster size, the default value of the granularity
+is 65536.
 
 
 Example:
-- 
1.7.12