[Qemu-devel] [PATCH 03/20] block: implement dirty bitmap using HBitmap

[Paolo Bonzini <pbonzini@redhat.com>]

This actually uses the dirty bitmap in the block layer, and converts
mirroring to use an HBitmapIter.

Reviewed-by: Laszlo Ersek <lersek@redhat.com> (except block/mirror.c parts)
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 Makefile.objs  |  2 +-
 block.c        | 94 ++++++++++------------------------------------------------
 block.h        |  6 ++--
 block/mirror.c | 12 ++++++--
 block_int.h    |  4 +--
 trace-events   |  1 +
 6 files changed, 33 insertions(+), 86 deletions(-)

diff --git a/Makefile.objs b/Makefile.objs
index 3c7abca..9dc96c3 100644
--- a/Makefile.objs
+++ b/Makefile.objs
@@ -46,7 +46,7 @@ coroutine-obj-$(CONFIG_WIN32) += coroutine-win32.o
 # block-obj-y is code used by both qemu system emulation and qemu-img
 
 block-obj-y = iov.o cache-utils.o qemu-option.o module.o async.o
-block-obj-y += nbd.o block.o blockjob.o aes.o qemu-config.o
+block-obj-y += nbd.o block.o blockjob.o aes.o qemu-config.o hbitmap.o
 block-obj-y += thread-pool.o qemu-progress.o qemu-sockets.o uri.o notify.o
 block-obj-y += $(coroutine-obj-y) $(qobject-obj-y) $(version-obj-y)
 block-obj-$(CONFIG_POSIX) += event_notifier-posix.o aio-posix.o
diff --git a/block.c b/block.c
index c05875f..f033c1e 100644
--- a/block.c
+++ b/block.c
@@ -1273,7 +1273,6 @@ static void bdrv_move_feature_fields(BlockDriverState *bs_dest,
     bs_dest->iostatus           = bs_src->iostatus;
 
     /* dirty bitmap */
-    bs_dest->dirty_count        = bs_src->dirty_count;
     bs_dest->dirty_bitmap       = bs_src->dirty_bitmap;
 
     /* job */
@@ -2022,36 +2021,6 @@ int bdrv_read_unthrottled(BlockDriverState *bs, int64_t sector_num,
     return ret;
 }
 
-#define BITS_PER_LONG  (sizeof(unsigned long) * 8)
-
-static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
-                             int nb_sectors, int dirty)
-{
-    int64_t start, end;
-    unsigned long val, idx, bit;
-
-    start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
-    end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
-
-    for (; start <= end; start++) {
-        idx = start / BITS_PER_LONG;
-        bit = start % BITS_PER_LONG;
-        val = bs->dirty_bitmap[idx];
-        if (dirty) {
-            if (!(val & (1UL << bit))) {
-                bs->dirty_count++;
-                val |= 1UL << bit;
-            }
-        } else {
-            if (val & (1UL << bit)) {
-                bs->dirty_count--;
-                val &= ~(1UL << bit);
-            }
-        }
-        bs->dirty_bitmap[idx] = val;
-    }
-}
-
 /* Return < 0 if error. Important errors are:
   -EIO         generic I/O error (may happen for all errors)
   -ENOMEDIUM   No media inserted.
@@ -4254,18 +4223,15 @@ void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
 {
     int64_t bitmap_size;
 
-    bs->dirty_count = 0;
     if (enable) {
         if (!bs->dirty_bitmap) {
-            bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
-                    BDRV_SECTORS_PER_DIRTY_CHUNK * BITS_PER_LONG - 1;
-            bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * BITS_PER_LONG;
-
-            bs->dirty_bitmap = g_new0(unsigned long, bitmap_size);
+            bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS);
+            bs->dirty_bitmap = hbitmap_alloc(bitmap_size,
+                                             BDRV_LOG_SECTORS_PER_DIRTY_CHUNK);
         }
     } else {
         if (bs->dirty_bitmap) {
-            g_free(bs->dirty_bitmap);
+            hbitmap_free(bs->dirty_bitmap);
             bs->dirty_bitmap = NULL;
         }
     }
@@ -4273,67 +4239,37 @@ void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
 
 int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
 {
-    int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
-
-    if (bs->dirty_bitmap &&
-        (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
-        return !!(bs->dirty_bitmap[chunk / BITS_PER_LONG] &
-            (1UL << (chunk % BITS_PER_LONG)));
+    if (bs->dirty_bitmap) {
+        return hbitmap_get(bs->dirty_bitmap, sector);
     } else {
         return 0;
     }
 }
 
-int64_t bdrv_get_next_dirty(BlockDriverState *bs, int64_t sector)
+void bdrv_dirty_iter_init(BlockDriverState *bs, HBitmapIter *hbi)
 {
-    int64_t chunk;
-    int bit, elem;
-
-    /* Avoid an infinite loop.  */
-    assert(bs->dirty_count > 0);
-
-    sector = (sector | (BDRV_SECTORS_PER_DIRTY_CHUNK - 1)) + 1;
-    chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
-
-    QEMU_BUILD_BUG_ON(sizeof(bs->dirty_bitmap[0]) * 8 != BITS_PER_LONG);
-    elem = chunk / BITS_PER_LONG;
-    bit = chunk % BITS_PER_LONG;
-    for (;;) {
-        if (sector >= bs->total_sectors) {
-            sector = 0;
-            bit = elem = 0;
-        }
-        if (bit == 0 && bs->dirty_bitmap[elem] == 0) {
-            sector += BDRV_SECTORS_PER_DIRTY_CHUNK * BITS_PER_LONG;
-            elem++;
-        } else {
-            if (bs->dirty_bitmap[elem] & (1UL << bit)) {
-                return sector;
-            }
-            sector += BDRV_SECTORS_PER_DIRTY_CHUNK;
-            if (++bit == BITS_PER_LONG) {
-                bit = 0;
-                elem++;
-            }
-        }
-    }
+    hbitmap_iter_init(hbi, bs->dirty_bitmap, 0);
 }
 
 void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector,
                     int nr_sectors)
 {
-    set_dirty_bitmap(bs, cur_sector, nr_sectors, 1);
+    hbitmap_set(bs->dirty_bitmap, cur_sector, nr_sectors);
 }
 
 void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
                       int nr_sectors)
 {
-    set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
+    hbitmap_reset(bs->dirty_bitmap, cur_sector, nr_sectors);
 }
 
 int64_t bdrv_get_dirty_count(BlockDriverState *bs)
 {
-    return bs->dirty_count;
+    if (bs->dirty_bitmap) {
+        return hbitmap_count(bs->dirty_bitmap) >> BDRV_LOG_SECTORS_PER_DIRTY_CHUNK;
+    } else {
+        return 0;
+    }
 }
 
 void bdrv_set_in_use(BlockDriverState *bs, int in_use)
diff --git a/block.h b/block.h
index 722c620..6a84350 100644
--- a/block.h
+++ b/block.h
@@ -350,13 +350,15 @@ int bdrv_img_create(const char *filename, const char *fmt,
 void bdrv_set_buffer_alignment(BlockDriverState *bs, int align);
 void *qemu_blockalign(BlockDriverState *bs, size_t size);
 
-#define BDRV_SECTORS_PER_DIRTY_CHUNK 2048
+#define BDRV_SECTORS_PER_DIRTY_CHUNK     (1 << BDRV_LOG_SECTORS_PER_DIRTY_CHUNK)
+#define BDRV_LOG_SECTORS_PER_DIRTY_CHUNK 11
 
+struct HBitmapIter;
 void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable);
 int bdrv_get_dirty(BlockDriverState *bs, int64_t sector);
 void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector, int nr_sectors);
 void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector, int nr_sectors);
-int64_t bdrv_get_next_dirty(BlockDriverState *bs, int64_t sector);
+void bdrv_dirty_iter_init(BlockDriverState *bs, struct HBitmapIter *hbi);
 int64_t bdrv_get_dirty_count(BlockDriverState *bs);
 
 void bdrv_enable_copy_on_read(BlockDriverState *bs);
diff --git a/block/mirror.c b/block/mirror.c
index d6618a4..30bb267 100644
--- a/block/mirror.c
+++ b/block/mirror.c
@@ -36,6 +36,7 @@ typedef struct MirrorBlockJob {
     bool synced;
     bool should_complete;
     int64_t sector_num;
+    HBitmapIter hbi;
     uint8_t *buf;
 } MirrorBlockJob;
 
@@ -62,8 +63,15 @@ static int coroutine_fn mirror_iteration(MirrorBlockJob *s,
     int64_t end;
     struct iovec iov;
 
+    s->sector_num = hbitmap_iter_next(&s->hbi);
+    if (s->sector_num < 0) {
+        bdrv_dirty_iter_init(source, &s->hbi);
+        s->sector_num = hbitmap_iter_next(&s->hbi);
+        trace_mirror_restart_iter(s, bdrv_get_dirty_count(source));
+        assert(s->sector_num >= 0);
+    }
+
     end = s->common.len >> BDRV_SECTOR_BITS;
-    s->sector_num = bdrv_get_next_dirty(source, s->sector_num);
     nb_sectors = MIN(BDRV_SECTORS_PER_DIRTY_CHUNK, end - s->sector_num);
     bdrv_reset_dirty(source, s->sector_num, nb_sectors);
 
@@ -136,7 +144,7 @@ static void coroutine_fn mirror_run(void *opaque)
         }
     }
 
-    s->sector_num = -1;
+    bdrv_dirty_iter_init(bs, &s->hbi);
     for (;;) {
         uint64_t delay_ns;
         int64_t cnt;
diff --git a/block_int.h b/block_int.h
index 9deedb8..364ab2d 100644
--- a/block_int.h
+++ b/block_int.h
@@ -32,6 +32,7 @@
 #include "qapi-types.h"
 #include "qerror.h"
 #include "monitor.h"
+#include "hbitmap.h"
 
 #define BLOCK_FLAG_ENCRYPT          1
 #define BLOCK_FLAG_COMPAT6          4
@@ -269,8 +270,7 @@ struct BlockDriverState {
     bool iostatus_enabled;
     BlockDeviceIoStatus iostatus;
     char device_name[32];
-    unsigned long *dirty_bitmap;
-    int64_t dirty_count;
+    HBitmap *dirty_bitmap;
     int in_use; /* users other than guest access, eg. block migration */
     QTAILQ_ENTRY(BlockDriverState) list;
 
diff --git a/trace-events b/trace-events
index 93e8901..b868ac6 100644
--- a/trace-events
+++ b/trace-events
@@ -79,6 +79,7 @@ commit_start(void *bs, void *base, void *top, void *s, void *co, void *opaque) "
 
 # block/mirror.c
 mirror_start(void *bs, void *s, void *co, void *opaque) "bs %p s %p co %p opaque %p"
+mirror_restart_iter(void *s, int64_t cnt) "s %p dirty count %"PRId64
 mirror_before_flush(void *s) "s %p"
 mirror_before_drain(void *s, int64_t cnt) "s %p dirty count %"PRId64
 mirror_before_sleep(void *s, int64_t cnt, int synced) "s %p dirty count %"PRId64" synced %d"
-- 
1.8.0.1