-----BEGIN PGP SIGNATURE-----

Version: GnuPG v2
 
 iQEtBAABCAAXBQJZQyPmEBxmYW16QHJlZGhhdC5jb20ACgkQyjViTGqRccaWrgf/
 SCAHpi4gzWbr7AN03jP16Qy/kqNik6F7LTNSqrRbvBPb3TNchDd4z44SAghK5m/r
 +IlYQc20sBZ60tRHIHAUSF2WNcea2pj1v3ZVgjrI7hiJ3DXPiqqt/dAR/W/BLIDO
 tAHAVF6Pnrjm9DC4d2zATLDHvcHMzWOsnePh7XcOm44REbwUr3GDg6bf2+j+5yfS
 9ewmXfh8z4w1IvSn+f5B+IeCvGvJNA1D55dqcGo8Ivlg9PnElziXFaXO2s7UiLIM
 mF3eTSIbJQNNN+E+0lpRpnqQiq+Txxggu61Q4f8bOTBhEOPa3etj1ydnXMVbvX25
 6SUuBfGh51tyOIZOJz3GtA==
 =9b+J
 -----END PGP SIGNATURE-----

Merge remote-tracking branch 'remotes/famz/tags/docker-and-block-pull-request' into staging

# gpg: Signature made Fri 16 Jun 2017 01:18:46 BST
# gpg:                using RSA key 0xCA35624C6A9171C6
# gpg: Good signature from "Fam Zheng <famz@redhat.com>"
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg:          It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 5003 7CB7 9706 0F76 F021  AD56 CA35 624C 6A91 71C6

* remotes/famz/tags/docker-and-block-pull-request: (23 commits)
  block: make accounting thread-safe
  block: split BlockAcctStats creation and setup
  block: introduce block_account_one_io
  block: protect modification of dirty bitmaps with a mutex
  migration/block: reset dirty bitmap before reading
  block: introduce dirty_bitmap_mutex
  block: protect tracked_requests and flush_queue with reqs_lock
  block: access write_gen with atomics
  block: use Stat64 for wr_highest_offset
  util: add stats64 module
  throttle-groups: protect throttled requests with a CoMutex
  throttle-groups: do not use qemu_co_enter_next
  throttle-groups: only start one coroutine from drained_begin
  block: access io_plugged with atomic ops
  block: access wakeup with atomic ops
  block: access serialising_in_flight with atomic ops
  block: access io_limits_disabled with atomic ops
  block: access quiesce_counter with atomic ops
  block: access copy_on_read with atomic ops
  docker: Add flex and bison to centos6 image
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

block.c

@@ -320,6 +320,8 @@ BlockDriverState *bdrv_new(void)
         QLIST_INIT(&bs->op_blockers[i]);
     }
     notifier_with_return_list_init(&bs->before_write_notifiers);
+    qemu_co_mutex_init(&bs->reqs_lock);
+    qemu_mutex_init(&bs->dirty_bitmap_mutex);
     bs->refcnt = 1;
     bs->aio_context = qemu_get_aio_context();
 
@@ -1300,7 +1302,9 @@ static int bdrv_open_common(BlockDriverState *bs, BlockBackend *file,
         goto fail_opts;
     }
 
-    assert(bs->copy_on_read == 0); /* bdrv_new() and bdrv_close() make it so */
+    /* bdrv_new() and bdrv_close() make it so */
+    assert(atomic_read(&bs->copy_on_read) == 0);
+
     if (bs->open_flags & BDRV_O_COPY_ON_READ) {
         if (!bs->read_only) {
             bdrv_enable_copy_on_read(bs);
@@ -3063,7 +3067,7 @@ static void bdrv_close(BlockDriverState *bs)
 
         g_free(bs->opaque);
         bs->opaque = NULL;
-        bs->copy_on_read = 0;
+        atomic_set(&bs->copy_on_read, 0);
         bs->backing_file[0] = '\0';
         bs->backing_format[0] = '\0';
         bs->total_sectors = 0;
@@ -3422,7 +3426,7 @@ int bdrv_truncate(BdrvChild *child, int64_t offset, Error **errp)
         ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
         bdrv_dirty_bitmap_truncate(bs);
         bdrv_parent_cb_resize(bs);
-        ++bs->write_gen;
+        atomic_inc(&bs->write_gen);
     }
     return ret;
 }

block/accounting.c

@@ -32,23 +32,28 @@
 static QEMUClockType clock_type = QEMU_CLOCK_REALTIME;
 static const int qtest_latency_ns = NANOSECONDS_PER_SECOND / 1000;
 
-void block_acct_init(BlockAcctStats *stats, bool account_invalid,
+void block_acct_init(BlockAcctStats *stats)
-                     bool account_failed)
 {
-    stats->account_invalid = account_invalid;
+    qemu_mutex_init(&stats->lock);
-    stats->account_failed = account_failed;
-
     if (qtest_enabled()) {
         clock_type = QEMU_CLOCK_VIRTUAL;
     }
 }
 
+void block_acct_setup(BlockAcctStats *stats, bool account_invalid,
+                      bool account_failed)
+{
+    stats->account_invalid = account_invalid;
+    stats->account_failed = account_failed;
+}
+
 void block_acct_cleanup(BlockAcctStats *stats)
 {
     BlockAcctTimedStats *s, *next;
     QSLIST_FOREACH_SAFE(s, &stats->intervals, entries, next) {
         g_free(s);
     }
+    qemu_mutex_destroy(&stats->lock);
 }
 
 void block_acct_add_interval(BlockAcctStats *stats, unsigned interval_length)
@@ -58,12 +63,15 @@ void block_acct_add_interval(BlockAcctStats *stats, unsigned interval_length)
 
     s = g_new0(BlockAcctTimedStats, 1);
     s->interval_length = interval_length;
+    s->stats = stats;
+    qemu_mutex_lock(&stats->lock);
     QSLIST_INSERT_HEAD(&stats->intervals, s, entries);
 
     for (i = 0; i < BLOCK_MAX_IOTYPE; i++) {
         timed_average_init(&s->latency[i], clock_type,
                            (uint64_t) interval_length * NANOSECONDS_PER_SECOND);
     }
+    qemu_mutex_unlock(&stats->lock);
 }
 
 BlockAcctTimedStats *block_acct_interval_next(BlockAcctStats *stats,
@@ -86,7 +94,8 @@ void block_acct_start(BlockAcctStats *stats, BlockAcctCookie *cookie,
     cookie->type = type;
 }
 
-void block_acct_done(BlockAcctStats *stats, BlockAcctCookie *cookie)
+static void block_account_one_io(BlockAcctStats *stats, BlockAcctCookie *cookie,
+                                 bool failed)
 {
     BlockAcctTimedStats *s;
     int64_t time_ns = qemu_clock_get_ns(clock_type);
@@ -98,31 +107,16 @@ void block_acct_done(BlockAcctStats *stats, BlockAcctCookie *cookie)
 
     assert(cookie->type < BLOCK_MAX_IOTYPE);
 
-    stats->nr_bytes[cookie->type] += cookie->bytes;
+    qemu_mutex_lock(&stats->lock);
-    stats->nr_ops[cookie->type]++;
-    stats->total_time_ns[cookie->type] += latency_ns;
-    stats->last_access_time_ns = time_ns;
 
-    QSLIST_FOREACH(s, &stats->intervals, entries) {
+    if (failed) {
-        timed_average_account(&s->latency[cookie->type], latency_ns);
+        stats->failed_ops[cookie->type]++;
+    } else {
+        stats->nr_bytes[cookie->type] += cookie->bytes;
+        stats->nr_ops[cookie->type]++;
     }
-}
-
-void block_acct_failed(BlockAcctStats *stats, BlockAcctCookie *cookie)
-{
-    assert(cookie->type < BLOCK_MAX_IOTYPE);
-
-    stats->failed_ops[cookie->type]++;
-
-    if (stats->account_failed) {
-        BlockAcctTimedStats *s;
-        int64_t time_ns = qemu_clock_get_ns(clock_type);
-        int64_t latency_ns = time_ns - cookie->start_time_ns;
-
-        if (qtest_enabled()) {
-            latency_ns = qtest_latency_ns;
-        }
 
+    if (!failed || stats->account_failed) {
         stats->total_time_ns[cookie->type] += latency_ns;
         stats->last_access_time_ns = time_ns;
 
@@ -130,29 +124,45 @@ void block_acct_failed(BlockAcctStats *stats, BlockAcctCookie *cookie)
             timed_average_account(&s->latency[cookie->type], latency_ns);
         }
     }
+
+    qemu_mutex_unlock(&stats->lock);
+}
+
+void block_acct_done(BlockAcctStats *stats, BlockAcctCookie *cookie)
+{
+    block_account_one_io(stats, cookie, false);
+}
+
+void block_acct_failed(BlockAcctStats *stats, BlockAcctCookie *cookie)
+{
+    block_account_one_io(stats, cookie, true);
 }
 
 void block_acct_invalid(BlockAcctStats *stats, enum BlockAcctType type)
 {
     assert(type < BLOCK_MAX_IOTYPE);
 
-    /* block_acct_done() and block_acct_failed() update
+    /* block_account_one_io() updates total_time_ns[], but this one does
-     * total_time_ns[], but this one does not. The reason is that
+     * not.  The reason is that invalid requests are accounted during their
-     * invalid requests are accounted during their submission,
+     * submission, therefore there's no actual I/O involved.
-     * therefore there's no actual I/O involved. */
+     */
-
+    qemu_mutex_lock(&stats->lock);
     stats->invalid_ops[type]++;
 
     if (stats->account_invalid) {
         stats->last_access_time_ns = qemu_clock_get_ns(clock_type);
     }
+    qemu_mutex_unlock(&stats->lock);
 }
 
 void block_acct_merge_done(BlockAcctStats *stats, enum BlockAcctType type,
                       int num_requests)
 {
     assert(type < BLOCK_MAX_IOTYPE);
+
+    qemu_mutex_lock(&stats->lock);
     stats->merged[type] += num_requests;
+    qemu_mutex_unlock(&stats->lock);
 }
 
 int64_t block_acct_idle_time_ns(BlockAcctStats *stats)
@@ -167,7 +177,9 @@ double block_acct_queue_depth(BlockAcctTimedStats *stats,
 
     assert(type < BLOCK_MAX_IOTYPE);
 
+    qemu_mutex_lock(&stats->stats->lock);
     sum = timed_average_sum(&stats->latency[type], &elapsed);
+    qemu_mutex_unlock(&stats->stats->lock);
 
     return (double) sum / elapsed;
 }

block/block-backend.c

@@ -216,8 +216,10 @@ BlockBackend *blk_new(uint64_t perm, uint64_t shared_perm)
     blk->shared_perm = shared_perm;
     blk_set_enable_write_cache(blk, true);
 
+    qemu_co_mutex_init(&blk->public.throttled_reqs_lock);
     qemu_co_queue_init(&blk->public.throttled_reqs[0]);
     qemu_co_queue_init(&blk->public.throttled_reqs[1]);
+    block_acct_init(&blk->stats);
 
     notifier_list_init(&blk->remove_bs_notifiers);
     notifier_list_init(&blk->insert_bs_notifiers);
@@ -1953,7 +1955,7 @@ static void blk_root_drained_begin(BdrvChild *child)
     /* Note that blk->root may not be accessible here yet if we are just
      * attaching to a BlockDriverState that is drained. Use child instead. */
 
-    if (blk->public.io_limits_disabled++ == 0) {
+    if (atomic_fetch_inc(&blk->public.io_limits_disabled) == 0) {
         throttle_group_restart_blk(blk);
     }
 }
@@ -1964,7 +1966,7 @@ static void blk_root_drained_end(BdrvChild *child)
     assert(blk->quiesce_counter);
 
     assert(blk->public.io_limits_disabled);
-    --blk->public.io_limits_disabled;
+    atomic_dec(&blk->public.io_limits_disabled);
 
     if (--blk->quiesce_counter == 0) {
         if (blk->dev_ops && blk->dev_ops->drained_end) {

block/dirty-bitmap.c

@@ -37,6 +37,7 @@
  *     or enabled. A frozen bitmap can only abdicate() or reclaim().
  */
 struct BdrvDirtyBitmap {
+    QemuMutex *mutex;
     HBitmap *bitmap;            /* Dirty sector bitmap implementation */
     HBitmap *meta;              /* Meta dirty bitmap */
     BdrvDirtyBitmap *successor; /* Anonymous child; implies frozen status */
@@ -52,6 +53,27 @@ struct BdrvDirtyBitmapIter {
     BdrvDirtyBitmap *bitmap;
 };
 
+static inline void bdrv_dirty_bitmaps_lock(BlockDriverState *bs)
+{
+    qemu_mutex_lock(&bs->dirty_bitmap_mutex);
+}
+
+static inline void bdrv_dirty_bitmaps_unlock(BlockDriverState *bs)
+{
+    qemu_mutex_unlock(&bs->dirty_bitmap_mutex);
+}
+
+void bdrv_dirty_bitmap_lock(BdrvDirtyBitmap *bitmap)
+{
+    qemu_mutex_lock(bitmap->mutex);
+}
+
+void bdrv_dirty_bitmap_unlock(BdrvDirtyBitmap *bitmap)
+{
+    qemu_mutex_unlock(bitmap->mutex);
+}
+
+/* Called with BQL or dirty_bitmap lock taken.  */
 BdrvDirtyBitmap *bdrv_find_dirty_bitmap(BlockDriverState *bs, const char *name)
 {
     BdrvDirtyBitmap *bm;
@@ -65,6 +87,7 @@ BdrvDirtyBitmap *bdrv_find_dirty_bitmap(BlockDriverState *bs, const char *name)
     return NULL;
 }
 
+/* Called with BQL taken.  */
 void bdrv_dirty_bitmap_make_anon(BdrvDirtyBitmap *bitmap)
 {
     assert(!bdrv_dirty_bitmap_frozen(bitmap));
@@ -72,6 +95,7 @@ void bdrv_dirty_bitmap_make_anon(BdrvDirtyBitmap *bitmap)
     bitmap->name = NULL;
 }
 
+/* Called with BQL taken.  */
 BdrvDirtyBitmap *bdrv_create_dirty_bitmap(BlockDriverState *bs,
                                           uint32_t granularity,
                                           const char *name,
@@ -96,11 +120,14 @@ BdrvDirtyBitmap *bdrv_create_dirty_bitmap(BlockDriverState *bs,
         return NULL;
     }
     bitmap = g_new0(BdrvDirtyBitmap, 1);
+    bitmap->mutex = &bs->dirty_bitmap_mutex;
     bitmap->bitmap = hbitmap_alloc(bitmap_size, ctz32(sector_granularity));
     bitmap->size = bitmap_size;
     bitmap->name = g_strdup(name);
     bitmap->disabled = false;
+    bdrv_dirty_bitmaps_lock(bs);
     QLIST_INSERT_HEAD(&bs->dirty_bitmaps, bitmap, list);
+    bdrv_dirty_bitmaps_unlock(bs);
     return bitmap;
 }
 
@@ -119,20 +146,24 @@ void bdrv_create_meta_dirty_bitmap(BdrvDirtyBitmap *bitmap,
                                    int chunk_size)
 {
     assert(!bitmap->meta);
+    qemu_mutex_lock(bitmap->mutex);
     bitmap->meta = hbitmap_create_meta(bitmap->bitmap,
                                        chunk_size * BITS_PER_BYTE);
+    qemu_mutex_unlock(bitmap->mutex);
 }
 
 void bdrv_release_meta_dirty_bitmap(BdrvDirtyBitmap *bitmap)
 {
     assert(bitmap->meta);
+    qemu_mutex_lock(bitmap->mutex);
     hbitmap_free_meta(bitmap->bitmap);
     bitmap->meta = NULL;
+    qemu_mutex_unlock(bitmap->mutex);
 }
 
-int bdrv_dirty_bitmap_get_meta(BlockDriverState *bs,
+int bdrv_dirty_bitmap_get_meta_locked(BlockDriverState *bs,
-                               BdrvDirtyBitmap *bitmap, int64_t sector,
+                                      BdrvDirtyBitmap *bitmap, int64_t sector,
-                               int nb_sectors)
+                                      int nb_sectors)
 {
     uint64_t i;
     int sectors_per_bit = 1 << hbitmap_granularity(bitmap->meta);
@@ -147,11 +178,26 @@ int bdrv_dirty_bitmap_get_meta(BlockDriverState *bs,
     return false;
 }
 
+int bdrv_dirty_bitmap_get_meta(BlockDriverState *bs,
+                               BdrvDirtyBitmap *bitmap, int64_t sector,
+                               int nb_sectors)
+{
+    bool dirty;
+
+    qemu_mutex_lock(bitmap->mutex);
+    dirty = bdrv_dirty_bitmap_get_meta_locked(bs, bitmap, sector, nb_sectors);
+    qemu_mutex_unlock(bitmap->mutex);
+
+    return dirty;
+}
+
 void bdrv_dirty_bitmap_reset_meta(BlockDriverState *bs,
                                   BdrvDirtyBitmap *bitmap, int64_t sector,
                                   int nb_sectors)
 {
+    qemu_mutex_lock(bitmap->mutex);
     hbitmap_reset(bitmap->meta, sector, nb_sectors);
+    qemu_mutex_unlock(bitmap->mutex);
 }
 
 int64_t bdrv_dirty_bitmap_size(const BdrvDirtyBitmap *bitmap)
@@ -164,16 +210,19 @@ const char *bdrv_dirty_bitmap_name(const BdrvDirtyBitmap *bitmap)
     return bitmap->name;
 }
 
+/* Called with BQL taken.  */
 bool bdrv_dirty_bitmap_frozen(BdrvDirtyBitmap *bitmap)
 {
     return bitmap->successor;
 }
 
+/* Called with BQL taken.  */
 bool bdrv_dirty_bitmap_enabled(BdrvDirtyBitmap *bitmap)
 {
     return !(bitmap->disabled || bitmap->successor);
 }
 
+/* Called with BQL taken.  */
 DirtyBitmapStatus bdrv_dirty_bitmap_status(BdrvDirtyBitmap *bitmap)
 {
     if (bdrv_dirty_bitmap_frozen(bitmap)) {
@@ -188,6 +237,7 @@ DirtyBitmapStatus bdrv_dirty_bitmap_status(BdrvDirtyBitmap *bitmap)
 /**
  * Create a successor bitmap destined to replace this bitmap after an operation.
  * Requires that the bitmap is not frozen and has no successor.
+ * Called with BQL taken.
  */
 int bdrv_dirty_bitmap_create_successor(BlockDriverState *bs,
                                        BdrvDirtyBitmap *bitmap, Error **errp)
@@ -220,6 +270,7 @@ int bdrv_dirty_bitmap_create_successor(BlockDriverState *bs,
 /**
  * For a bitmap with a successor, yield our name to the successor,
  * delete the old bitmap, and return a handle to the new bitmap.
+ * Called with BQL taken.
  */
 BdrvDirtyBitmap *bdrv_dirty_bitmap_abdicate(BlockDriverState *bs,
                                             BdrvDirtyBitmap *bitmap,
@@ -247,6 +298,7 @@ BdrvDirtyBitmap *bdrv_dirty_bitmap_abdicate(BlockDriverState *bs,
  * In cases of failure where we can no longer safely delete the parent,
  * we may wish to re-join the parent and child/successor.
  * The merged parent will be un-frozen, but not explicitly re-enabled.
+ * Called with BQL taken.
  */
 BdrvDirtyBitmap *bdrv_reclaim_dirty_bitmap(BlockDriverState *bs,
                                            BdrvDirtyBitmap *parent,
@@ -271,25 +323,30 @@ BdrvDirtyBitmap *bdrv_reclaim_dirty_bitmap(BlockDriverState *bs,
 
 /**
  * Truncates _all_ bitmaps attached to a BDS.
+ * Called with BQL taken.
  */
 void bdrv_dirty_bitmap_truncate(BlockDriverState *bs)
 {
     BdrvDirtyBitmap *bitmap;
     uint64_t size = bdrv_nb_sectors(bs);
 
+    bdrv_dirty_bitmaps_lock(bs);
     QLIST_FOREACH(bitmap, &bs->dirty_bitmaps, list) {
         assert(!bdrv_dirty_bitmap_frozen(bitmap));
         assert(!bitmap->active_iterators);
         hbitmap_truncate(bitmap->bitmap, size);
         bitmap->size = size;
     }
+    bdrv_dirty_bitmaps_unlock(bs);
 }
 
+/* Called with BQL taken.  */
 static void bdrv_do_release_matching_dirty_bitmap(BlockDriverState *bs,
                                                   BdrvDirtyBitmap *bitmap,
                                                   bool only_named)
 {
     BdrvDirtyBitmap *bm, *next;
+    bdrv_dirty_bitmaps_lock(bs);
     QLIST_FOREACH_SAFE(bm, &bs->dirty_bitmaps, list, next) {
         if ((!bitmap || bm == bitmap) && (!only_named || bm->name)) {
             assert(!bm->active_iterators);
@@ -301,15 +358,19 @@ static void bdrv_do_release_matching_dirty_bitmap(BlockDriverState *bs,
             g_free(bm);
 
             if (bitmap) {
-                return;
+                goto out;
             }
         }
     }
     if (bitmap) {
         abort();
     }
+
+out:
+    bdrv_dirty_bitmaps_unlock(bs);
 }
 
+/* Called with BQL taken.  */
 void bdrv_release_dirty_bitmap(BlockDriverState *bs, BdrvDirtyBitmap *bitmap)
 {
     bdrv_do_release_matching_dirty_bitmap(bs, bitmap, false);
@@ -318,18 +379,21 @@ void bdrv_release_dirty_bitmap(BlockDriverState *bs, BdrvDirtyBitmap *bitmap)
 /**
  * Release all named dirty bitmaps attached to a BDS (for use in bdrv_close()).
  * There must not be any frozen bitmaps attached.
+ * Called with BQL taken.
  */
 void bdrv_release_named_dirty_bitmaps(BlockDriverState *bs)
 {
     bdrv_do_release_matching_dirty_bitmap(bs, NULL, true);
 }
 
+/* Called with BQL taken.  */
 void bdrv_disable_dirty_bitmap(BdrvDirtyBitmap *bitmap)
 {
     assert(!bdrv_dirty_bitmap_frozen(bitmap));
     bitmap->disabled = true;
 }
 
+/* Called with BQL taken.  */
 void bdrv_enable_dirty_bitmap(BdrvDirtyBitmap *bitmap)
 {
     assert(!bdrv_dirty_bitmap_frozen(bitmap));
@@ -342,6 +406,7 @@ BlockDirtyInfoList *bdrv_query_dirty_bitmaps(BlockDriverState *bs)
     BlockDirtyInfoList *list = NULL;
     BlockDirtyInfoList **plist = &list;
 
+    bdrv_dirty_bitmaps_lock(bs);
     QLIST_FOREACH(bm, &bs->dirty_bitmaps, list) {
         BlockDirtyInfo *info = g_new0(BlockDirtyInfo, 1);
         BlockDirtyInfoList *entry = g_new0(BlockDirtyInfoList, 1);
@@ -354,12 +419,14 @@ BlockDirtyInfoList *bdrv_query_dirty_bitmaps(BlockDriverState *bs)
         *plist = entry;
         plist = &entry->next;
     }
+    bdrv_dirty_bitmaps_unlock(bs);
 
     return list;
 }
 
-int bdrv_get_dirty(BlockDriverState *bs, BdrvDirtyBitmap *bitmap,
+/* Called within bdrv_dirty_bitmap_lock..unlock */
-                   int64_t sector)
+int bdrv_get_dirty_locked(BlockDriverState *bs, BdrvDirtyBitmap *bitmap,
+                          int64_t sector)
 {
     if (bitmap) {
         return hbitmap_get(bitmap->bitmap, sector);
@@ -432,23 +499,42 @@ int64_t bdrv_dirty_iter_next(BdrvDirtyBitmapIter *iter)
     return hbitmap_iter_next(&iter->hbi);
 }
 
-void bdrv_set_dirty_bitmap(BdrvDirtyBitmap *bitmap,
+/* Called within bdrv_dirty_bitmap_lock..unlock */
-                           int64_t cur_sector, int64_t nr_sectors)
+void bdrv_set_dirty_bitmap_locked(BdrvDirtyBitmap *bitmap,
+                                  int64_t cur_sector, int64_t nr_sectors)
 {
     assert(bdrv_dirty_bitmap_enabled(bitmap));
     hbitmap_set(bitmap->bitmap, cur_sector, nr_sectors);
 }
 
-void bdrv_reset_dirty_bitmap(BdrvDirtyBitmap *bitmap,
+void bdrv_set_dirty_bitmap(BdrvDirtyBitmap *bitmap,
-                             int64_t cur_sector, int64_t nr_sectors)
+                           int64_t cur_sector, int64_t nr_sectors)
+{
+    bdrv_dirty_bitmap_lock(bitmap);
+    bdrv_set_dirty_bitmap_locked(bitmap, cur_sector, nr_sectors);
+    bdrv_dirty_bitmap_unlock(bitmap);
+}
+
+/* Called within bdrv_dirty_bitmap_lock..unlock */
+void bdrv_reset_dirty_bitmap_locked(BdrvDirtyBitmap *bitmap,
+                                    int64_t cur_sector, int64_t nr_sectors)
 {
     assert(bdrv_dirty_bitmap_enabled(bitmap));
     hbitmap_reset(bitmap->bitmap, cur_sector, nr_sectors);
 }
 
+void bdrv_reset_dirty_bitmap(BdrvDirtyBitmap *bitmap,
+                             int64_t cur_sector, int64_t nr_sectors)
+{
+    bdrv_dirty_bitmap_lock(bitmap);
+    bdrv_reset_dirty_bitmap_locked(bitmap, cur_sector, nr_sectors);
+    bdrv_dirty_bitmap_unlock(bitmap);
+}
+
 void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out)
 {
     assert(bdrv_dirty_bitmap_enabled(bitmap));
+    bdrv_dirty_bitmap_lock(bitmap);
     if (!out) {
         hbitmap_reset_all(bitmap->bitmap);
     } else {
@@ -457,6 +543,7 @@ void bdrv_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap **out)
                                        hbitmap_granularity(backup));
         *out = backup;
     }
+    bdrv_dirty_bitmap_unlock(bitmap);
 }
 
 void bdrv_undo_clear_dirty_bitmap(BdrvDirtyBitmap *bitmap, HBitmap *in)
@@ -508,12 +595,19 @@ void bdrv_set_dirty(BlockDriverState *bs, int64_t cur_sector,
                     int64_t nr_sectors)
 {
     BdrvDirtyBitmap *bitmap;
+
+    if (QLIST_EMPTY(&bs->dirty_bitmaps)) {
+        return;
+    }
+
+    bdrv_dirty_bitmaps_lock(bs);
     QLIST_FOREACH(bitmap, &bs->dirty_bitmaps, list) {
         if (!bdrv_dirty_bitmap_enabled(bitmap)) {
             continue;
         }
         hbitmap_set(bitmap->bitmap, cur_sector, nr_sectors);
     }
+    bdrv_dirty_bitmaps_unlock(bs);
 }
 
 /**

block/io.c

@@ -130,13 +130,13 @@ void bdrv_refresh_limits(BlockDriverState *bs, Error **errp)
  */
 void bdrv_enable_copy_on_read(BlockDriverState *bs)
 {
-    bs->copy_on_read++;
+    atomic_inc(&bs->copy_on_read);
 }
 
 void bdrv_disable_copy_on_read(BlockDriverState *bs)
 {
-    assert(bs->copy_on_read > 0);
+    int old = atomic_fetch_dec(&bs->copy_on_read);
-    bs->copy_on_read--;
+    assert(old >= 1);
 }
 
 /* Check if any requests are in-flight (including throttled requests) */
@@ -241,7 +241,7 @@ void bdrv_drained_begin(BlockDriverState *bs)
         return;
     }
 
-    if (!bs->quiesce_counter++) {
+    if (atomic_fetch_inc(&bs->quiesce_counter) == 0) {
         aio_disable_external(bdrv_get_aio_context(bs));
         bdrv_parent_drained_begin(bs);
     }
@@ -252,7 +252,7 @@ void bdrv_drained_begin(BlockDriverState *bs)
 void bdrv_drained_end(BlockDriverState *bs)
 {
     assert(bs->quiesce_counter > 0);
-    if (--bs->quiesce_counter > 0) {
+    if (atomic_fetch_dec(&bs->quiesce_counter) > 1) {
         return;
     }
 
@@ -375,11 +375,13 @@ void bdrv_drain_all(void)
 static void tracked_request_end(BdrvTrackedRequest *req)
 {
     if (req->serialising) {
-        req->bs->serialising_in_flight--;
+        atomic_dec(&req->bs->serialising_in_flight);
     }
 
+    qemu_co_mutex_lock(&req->bs->reqs_lock);
     QLIST_REMOVE(req, list);
     qemu_co_queue_restart_all(&req->wait_queue);
+    qemu_co_mutex_unlock(&req->bs->reqs_lock);
 }
 
 /**
@@ -404,7 +406,9 @@ static void tracked_request_begin(BdrvTrackedRequest *req,
 
     qemu_co_queue_init(&req->wait_queue);
 
+    qemu_co_mutex_lock(&bs->reqs_lock);
     QLIST_INSERT_HEAD(&bs->tracked_requests, req, list);
+    qemu_co_mutex_unlock(&bs->reqs_lock);
 }
 
 static void mark_request_serialising(BdrvTrackedRequest *req, uint64_t align)
@@ -414,7 +418,7 @@ static void mark_request_serialising(BdrvTrackedRequest *req, uint64_t align)
                                - overlap_offset;
 
     if (!req->serialising) {
-        req->bs->serialising_in_flight++;
+        atomic_inc(&req->bs->serialising_in_flight);
         req->serialising = true;
     }
 
@@ -501,7 +505,8 @@ static void dummy_bh_cb(void *opaque)
 
 void bdrv_wakeup(BlockDriverState *bs)
 {
-    if (bs->wakeup) {
+    /* The barrier (or an atomic op) is in the caller.  */
+    if (atomic_read(&bs->wakeup)) {
         aio_bh_schedule_oneshot(qemu_get_aio_context(), dummy_bh_cb, NULL);
     }
 }
@@ -519,12 +524,13 @@ static bool coroutine_fn wait_serialising_requests(BdrvTrackedRequest *self)
     bool retry;
     bool waited = false;
 
-    if (!bs->serialising_in_flight) {
+    if (!atomic_read(&bs->serialising_in_flight)) {
         return false;
     }
 
     do {
         retry = false;
+        qemu_co_mutex_lock(&bs->reqs_lock);
         QLIST_FOREACH(req, &bs->tracked_requests, list) {
             if (req == self || (!req->serialising && !self->serialising)) {
                 continue;
@@ -543,7 +549,7 @@ static bool coroutine_fn wait_serialising_requests(BdrvTrackedRequest *self)
                  * (instead of producing a deadlock in the former case). */
                 if (!req->waiting_for) {
                     self->waiting_for = req;
-                    qemu_co_queue_wait(&req->wait_queue, NULL);
+                    qemu_co_queue_wait(&req->wait_queue, &bs->reqs_lock);
                     self->waiting_for = NULL;
                     retry = true;
                     waited = true;
@@ -551,6 +557,7 @@ static bool coroutine_fn wait_serialising_requests(BdrvTrackedRequest *self)
                 }
             }
         }
+        qemu_co_mutex_unlock(&bs->reqs_lock);
     } while (retry);
 
     return waited;
@@ -1144,7 +1151,7 @@ int coroutine_fn bdrv_co_preadv(BdrvChild *child,
     bdrv_inc_in_flight(bs);
 
     /* Don't do copy-on-read if we read data before write operation */
-    if (bs->copy_on_read && !(flags & BDRV_REQ_NO_SERIALISING)) {
+    if (atomic_read(&bs->copy_on_read) && !(flags & BDRV_REQ_NO_SERIALISING)) {
         flags |= BDRV_REQ_COPY_ON_READ;
     }
 
@@ -1401,12 +1408,10 @@ static int coroutine_fn bdrv_aligned_pwritev(BdrvChild *child,
     }
     bdrv_debug_event(bs, BLKDBG_PWRITEV_DONE);
 
-    ++bs->write_gen;
+    atomic_inc(&bs->write_gen);
     bdrv_set_dirty(bs, start_sector, end_sector - start_sector);
 
-    if (bs->wr_highest_offset < offset + bytes) {
+    stat64_max(&bs->wr_highest_offset, offset + bytes);
-        bs->wr_highest_offset = offset + bytes;
-    }
 
     if (ret >= 0) {
         bs->total_sectors = MAX(bs->total_sectors, end_sector);
@@ -2292,14 +2297,17 @@ int coroutine_fn bdrv_co_flush(BlockDriverState *bs)
         goto early_exit;
     }
 
-    current_gen = bs->write_gen;
+    qemu_co_mutex_lock(&bs->reqs_lock);
+    current_gen = atomic_read(&bs->write_gen);
 
     /* Wait until any previous flushes are completed */
     while (bs->active_flush_req) {
-        qemu_co_queue_wait(&bs->flush_queue, NULL);
+        qemu_co_queue_wait(&bs->flush_queue, &bs->reqs_lock);
     }
 
+    /* Flushes reach this point in nondecreasing current_gen order.  */
     bs->active_flush_req = true;
+    qemu_co_mutex_unlock(&bs->reqs_lock);
 
     /* Write back all layers by calling one driver function */
     if (bs->drv->bdrv_co_flush) {
@@ -2371,9 +2379,12 @@ out:
     if (ret == 0) {
         bs->flushed_gen = current_gen;
     }
+
+    qemu_co_mutex_lock(&bs->reqs_lock);
     bs->active_flush_req = false;
     /* Return value is ignored - it's ok if wait queue is empty */
     qemu_co_queue_next(&bs->flush_queue);
+    qemu_co_mutex_unlock(&bs->reqs_lock);
 
 early_exit:
     bdrv_dec_in_flight(bs);
@@ -2517,7 +2528,7 @@ int coroutine_fn bdrv_co_pdiscard(BlockDriverState *bs, int64_t offset,
     }
     ret = 0;
 out:
-    ++bs->write_gen;
+    atomic_inc(&bs->write_gen);
     bdrv_set_dirty(bs, req.offset >> BDRV_SECTOR_BITS,
                    req.bytes >> BDRV_SECTOR_BITS);
     tracked_request_end(&req);
@@ -2644,7 +2655,7 @@ void bdrv_io_plug(BlockDriverState *bs)
         bdrv_io_plug(child->bs);
     }
 
-    if (bs->io_plugged++ == 0) {
+    if (atomic_fetch_inc(&bs->io_plugged) == 0) {
         BlockDriver *drv = bs->drv;
         if (drv && drv->bdrv_io_plug) {
             drv->bdrv_io_plug(bs);
@@ -2657,7 +2668,7 @@ void bdrv_io_unplug(BlockDriverState *bs)
     BdrvChild *child;
 
     assert(bs->io_plugged);
-    if (--bs->io_plugged == 0) {
+    if (atomic_fetch_dec(&bs->io_plugged) == 1) {
         BlockDriver *drv = bs->drv;
         if (drv && drv->bdrv_io_unplug) {
             drv->bdrv_io_unplug(bs);

block/mirror.c

@@ -342,6 +342,7 @@ static uint64_t coroutine_fn mirror_iteration(MirrorBlockJob *s)
     int max_io_sectors = MAX((s->buf_size >> BDRV_SECTOR_BITS) / MAX_IN_FLIGHT,
                              MAX_IO_SECTORS);
 
+    bdrv_dirty_bitmap_lock(s->dirty_bitmap);
     sector_num = bdrv_dirty_iter_next(s->dbi);
     if (sector_num < 0) {
         bdrv_set_dirty_iter(s->dbi, 0);
@@ -349,6 +350,7 @@ static uint64_t coroutine_fn mirror_iteration(MirrorBlockJob *s)
         trace_mirror_restart_iter(s, bdrv_get_dirty_count(s->dirty_bitmap));
         assert(sector_num >= 0);
     }
+    bdrv_dirty_bitmap_unlock(s->dirty_bitmap);
 
     first_chunk = sector_num / sectors_per_chunk;
     while (test_bit(first_chunk, s->in_flight_bitmap)) {
@@ -360,12 +362,13 @@ static uint64_t coroutine_fn mirror_iteration(MirrorBlockJob *s)
 
     /* Find the number of consective dirty chunks following the first dirty
      * one, and wait for in flight requests in them. */
+    bdrv_dirty_bitmap_lock(s->dirty_bitmap);
     while (nb_chunks * sectors_per_chunk < (s->buf_size >> BDRV_SECTOR_BITS)) {
         int64_t next_dirty;
         int64_t next_sector = sector_num + nb_chunks * sectors_per_chunk;
         int64_t next_chunk = next_sector / sectors_per_chunk;
         if (next_sector >= end ||
-            !bdrv_get_dirty(source, s->dirty_bitmap, next_sector)) {
+            !bdrv_get_dirty_locked(source, s->dirty_bitmap, next_sector)) {
             break;
         }
         if (test_bit(next_chunk, s->in_flight_bitmap)) {
@@ -386,8 +389,10 @@ static uint64_t coroutine_fn mirror_iteration(MirrorBlockJob *s)
      * calling bdrv_get_block_status_above could yield - if some blocks are
      * marked dirty in this window, we need to know.
      */
-    bdrv_reset_dirty_bitmap(s->dirty_bitmap, sector_num,
+    bdrv_reset_dirty_bitmap_locked(s->dirty_bitmap, sector_num,
-                            nb_chunks * sectors_per_chunk);
+                                  nb_chunks * sectors_per_chunk);
+    bdrv_dirty_bitmap_unlock(s->dirty_bitmap);
+
     bitmap_set(s->in_flight_bitmap, sector_num / sectors_per_chunk, nb_chunks);
     while (nb_chunks > 0 && sector_num < end) {
         int64_t ret;
@@ -506,6 +511,8 @@ static void mirror_exit(BlockJob *job, void *opaque)
     BlockDriverState *mirror_top_bs = s->mirror_top_bs;
     Error *local_err = NULL;
 
+    bdrv_release_dirty_bitmap(src, s->dirty_bitmap);
+
     /* Make sure that the source BDS doesn't go away before we called
      * block_job_completed(). */
     bdrv_ref(src);
@@ -904,7 +911,6 @@ immediate_exit:
     g_free(s->cow_bitmap);
     g_free(s->in_flight_bitmap);
     bdrv_dirty_iter_free(s->dbi);
-    bdrv_release_dirty_bitmap(bs, s->dirty_bitmap);
 
     data = g_malloc(sizeof(*data));
     data->ret = ret;

block/nfs.c

@@ -730,7 +730,9 @@ nfs_get_allocated_file_size_cb(int ret, struct nfs_context *nfs, void *data,
     if (task->ret < 0) {
         error_report("NFS Error: %s", nfs_get_error(nfs));
     }
-    task->complete = 1;
+
+    /* Set task->complete before reading bs->wakeup.  */
+    atomic_mb_set(&task->complete, 1);
     bdrv_wakeup(task->bs);
 }
 

block/qapi.c

@@ -441,7 +441,7 @@ static BlockStats *bdrv_query_bds_stats(const BlockDriverState *bs,
         s->node_name = g_strdup(bdrv_get_node_name(bs));
     }
 
-    s->stats->wr_highest_offset = bs->wr_highest_offset;
+    s->stats->wr_highest_offset = stat64_get(&bs->wr_highest_offset);
 
     if (bs->file) {
         s->has_parent = true;

block/sheepdog.c

@@ -698,7 +698,8 @@ out:
 
     srco->co = NULL;
     srco->ret = ret;
-    srco->finished = true;
+    /* Set srco->finished before reading bs->wakeup.  */
+    atomic_mb_set(&srco->finished, true);
     if (srco->bs) {
         bdrv_wakeup(srco->bs);
     }

block/throttle-groups.c

@@ -240,7 +240,7 @@ static bool throttle_group_schedule_timer(BlockBackend *blk, bool is_write)
     ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
     bool must_wait;
 
-    if (blkp->io_limits_disabled) {
+    if (atomic_read(&blkp->io_limits_disabled)) {
         return false;
     }
 
@@ -260,6 +260,25 @@ static bool throttle_group_schedule_timer(BlockBackend *blk, bool is_write)
     return must_wait;
 }
 
+/* Start the next pending I/O request for a BlockBackend.  Return whether
+ * any request was actually pending.
+ *
+ * @blk:       the current BlockBackend
+ * @is_write:  the type of operation (read/write)
+ */
+static bool coroutine_fn throttle_group_co_restart_queue(BlockBackend *blk,
+                                                         bool is_write)
+{
+    BlockBackendPublic *blkp = blk_get_public(blk);
+    bool ret;
+
+    qemu_co_mutex_lock(&blkp->throttled_reqs_lock);
+    ret = qemu_co_queue_next(&blkp->throttled_reqs[is_write]);
+    qemu_co_mutex_unlock(&blkp->throttled_reqs_lock);
+
+    return ret;
+}
+
 /* Look for the next pending I/O request and schedule it.
  *
  * This assumes that tg->lock is held.
@@ -287,12 +306,12 @@ static void schedule_next_request(BlockBackend *blk, bool is_write)
     if (!must_wait) {
         /* Give preference to requests from the current blk */
         if (qemu_in_coroutine() &&
-            qemu_co_queue_next(&blkp->throttled_reqs[is_write])) {
+            throttle_group_co_restart_queue(blk, is_write)) {
             token = blk;
         } else {
             ThrottleTimers *tt = &blk_get_public(token)->throttle_timers;
             int64_t now = qemu_clock_get_ns(tt->clock_type);
-            timer_mod(tt->timers[is_write], now + 1);
+            timer_mod(tt->timers[is_write], now);
             tg->any_timer_armed[is_write] = true;
         }
         tg->tokens[is_write] = token;
@@ -326,7 +345,10 @@ void coroutine_fn throttle_group_co_io_limits_intercept(BlockBackend *blk,
     if (must_wait || blkp->pending_reqs[is_write]) {
         blkp->pending_reqs[is_write]++;
         qemu_mutex_unlock(&tg->lock);
-        qemu_co_queue_wait(&blkp->throttled_reqs[is_write], NULL);
+        qemu_co_mutex_lock(&blkp->throttled_reqs_lock);
+        qemu_co_queue_wait(&blkp->throttled_reqs[is_write],
+                           &blkp->throttled_reqs_lock);
+        qemu_co_mutex_unlock(&blkp->throttled_reqs_lock);
         qemu_mutex_lock(&tg->lock);
         blkp->pending_reqs[is_write]--;
     }
@@ -340,15 +362,50 @@ void coroutine_fn throttle_group_co_io_limits_intercept(BlockBackend *blk,
     qemu_mutex_unlock(&tg->lock);
 }
 
+typedef struct {
+    BlockBackend *blk;
+    bool is_write;
+} RestartData;
+
+static void coroutine_fn throttle_group_restart_queue_entry(void *opaque)
+{
+    RestartData *data = opaque;
+    BlockBackend *blk = data->blk;
+    bool is_write = data->is_write;
+    BlockBackendPublic *blkp = blk_get_public(blk);
+    ThrottleGroup *tg = container_of(blkp->throttle_state, ThrottleGroup, ts);
+    bool empty_queue;
+
+    empty_queue = !throttle_group_co_restart_queue(blk, is_write);
+
+    /* If the request queue was empty then we have to take care of
+     * scheduling the next one */
+    if (empty_queue) {
+        qemu_mutex_lock(&tg->lock);
+        schedule_next_request(blk, is_write);
+        qemu_mutex_unlock(&tg->lock);
+    }
+}
+
+static void throttle_group_restart_queue(BlockBackend *blk, bool is_write)
+{
+    Coroutine *co;
+    RestartData rd = {
+        .blk = blk,
+        .is_write = is_write
+    };
+
+    co = qemu_coroutine_create(throttle_group_restart_queue_entry, &rd);
+    aio_co_enter(blk_get_aio_context(blk), co);
+}
+
 void throttle_group_restart_blk(BlockBackend *blk)
 {
     BlockBackendPublic *blkp = blk_get_public(blk);
-    int i;
 
-    for (i = 0; i < 2; i++) {
+    if (blkp->throttle_state) {
-        while (qemu_co_enter_next(&blkp->throttled_reqs[i])) {
+        throttle_group_restart_queue(blk, 0);
-            ;
+        throttle_group_restart_queue(blk, 1);
-        }
     }
 }
 
@@ -376,8 +433,7 @@ void throttle_group_config(BlockBackend *blk, ThrottleConfig *cfg)
     throttle_config(ts, tt, cfg);
     qemu_mutex_unlock(&tg->lock);
 
-    qemu_co_enter_next(&blkp->throttled_reqs[0]);
+    throttle_group_restart_blk(blk);
-    qemu_co_enter_next(&blkp->throttled_reqs[1]);
 }
 
 /* Get the throttle configuration from a particular group. Similar to
@@ -408,7 +464,6 @@ static void timer_cb(BlockBackend *blk, bool is_write)
     BlockBackendPublic *blkp = blk_get_public(blk);
     ThrottleState *ts = blkp->throttle_state;
     ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
-    bool empty_queue;
 
     /* The timer has just been fired, so we can update the flag */
     qemu_mutex_lock(&tg->lock);
@@ -416,17 +471,7 @@ static void timer_cb(BlockBackend *blk, bool is_write)
     qemu_mutex_unlock(&tg->lock);
 
     /* Run the request that was waiting for this timer */
-    aio_context_acquire(blk_get_aio_context(blk));
+    throttle_group_restart_queue(blk, is_write);
-    empty_queue = !qemu_co_enter_next(&blkp->throttled_reqs[is_write]);
-    aio_context_release(blk_get_aio_context(blk));
-
-    /* If the request queue was empty then we have to take care of
-     * scheduling the next one */
-    if (empty_queue) {
-        qemu_mutex_lock(&tg->lock);
-        schedule_next_request(blk, is_write);
-        qemu_mutex_unlock(&tg->lock);
-    }
 }
 
 static void read_timer_cb(void *opaque)

blockdev.c

@@ -595,7 +595,7 @@ static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
             autostart = 0;
         }
 
-        block_acct_init(blk_get_stats(blk), account_invalid, account_failed);
+        block_acct_setup(blk_get_stats(blk), account_invalid, account_failed);
 
         if (!parse_stats_intervals(blk_get_stats(blk), interval_list, errp)) {
             blk_unref(blk);
@@ -1362,12 +1362,10 @@ out_aio_context:
 static BdrvDirtyBitmap *block_dirty_bitmap_lookup(const char *node,
                                                   const char *name,
                                                   BlockDriverState **pbs,
-                                                  AioContext **paio,
                                                   Error **errp)
 {
     BlockDriverState *bs;
     BdrvDirtyBitmap *bitmap;
-    AioContext *aio_context;
 
     if (!node) {
         error_setg(errp, "Node cannot be NULL");
@@ -1383,29 +1381,17 @@ static BdrvDirtyBitmap *block_dirty_bitmap_lookup(const char *node,
         return NULL;
     }
 
-    aio_context = bdrv_get_aio_context(bs);
-    aio_context_acquire(aio_context);
-
     bitmap = bdrv_find_dirty_bitmap(bs, name);
     if (!bitmap) {
         error_setg(errp, "Dirty bitmap '%s' not found", name);
-        goto fail;
+        return NULL;
     }
 
     if (pbs) {
         *pbs = bs;
     }
-    if (paio) {
-        *paio = aio_context;
-    } else {
-        aio_context_release(aio_context);
-    }
 
     return bitmap;
-
- fail:
-    aio_context_release(aio_context);
-    return NULL;
 }
 
 /* New and old BlockDriverState structs for atomic group operations */
@@ -1791,7 +1777,7 @@ static void external_snapshot_commit(BlkActionState *common)
     /* We don't need (or want) to use the transactional
      * bdrv_reopen_multiple() across all the entries at once, because we
      * don't want to abort all of them if one of them fails the reopen */
-    if (!state->old_bs->copy_on_read) {
+    if (!atomic_read(&state->old_bs->copy_on_read)) {
         bdrv_reopen(state->old_bs, state->old_bs->open_flags & ~BDRV_O_RDWR,
                     NULL);
     }
@@ -2025,7 +2011,6 @@ static void block_dirty_bitmap_clear_prepare(BlkActionState *common,
     state->bitmap = block_dirty_bitmap_lookup(action->node,
                                               action->name,
                                               &state->bs,
-                                              &state->aio_context,
                                               errp);
     if (!state->bitmap) {
         return;
@@ -2733,7 +2718,6 @@ void qmp_block_dirty_bitmap_add(const char *node, const char *name,
                                 bool has_granularity, uint32_t granularity,
                                 Error **errp)
 {
-    AioContext *aio_context;
     BlockDriverState *bs;
 
     if (!name || name[0] == '\0') {
@@ -2746,14 +2730,11 @@ void qmp_block_dirty_bitmap_add(const char *node, const char *name,
         return;
     }
 
-    aio_context = bdrv_get_aio_context(bs);
-    aio_context_acquire(aio_context);
-
     if (has_granularity) {
         if (granularity < 512 || !is_power_of_2(granularity)) {
             error_setg(errp, "Granularity must be power of 2 "
                              "and at least 512");
-            goto out;
+            return;
         }
     } else {
         /* Default to cluster size, if available: */
@@ -2761,19 +2742,15 @@ void qmp_block_dirty_bitmap_add(const char *node, const char *name,
     }
 
     bdrv_create_dirty_bitmap(bs, granularity, name, errp);
-
- out:
-    aio_context_release(aio_context);
 }
 
 void qmp_block_dirty_bitmap_remove(const char *node, const char *name,
                                    Error **errp)
 {
-    AioContext *aio_context;
     BlockDriverState *bs;
     BdrvDirtyBitmap *bitmap;
 
-    bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
+    bitmap = block_dirty_bitmap_lookup(node, name, &bs, errp);
     if (!bitmap || !bs) {
         return;
     }
@@ -2782,13 +2759,10 @@ void qmp_block_dirty_bitmap_remove(const char *node, const char *name,
         error_setg(errp,
                    "Bitmap '%s' is currently frozen and cannot be removed",
                    name);
-        goto out;
+        return;
     }
     bdrv_dirty_bitmap_make_anon(bitmap);
     bdrv_release_dirty_bitmap(bs, bitmap);
-
- out:
-    aio_context_release(aio_context);
 }
 
 /**
@@ -2798,11 +2772,10 @@ void qmp_block_dirty_bitmap_remove(const char *node, const char *name,
 void qmp_block_dirty_bitmap_clear(const char *node, const char *name,
                                   Error **errp)
 {
-    AioContext *aio_context;
     BdrvDirtyBitmap *bitmap;
     BlockDriverState *bs;
 
-    bitmap = block_dirty_bitmap_lookup(node, name, &bs, &aio_context, errp);
+    bitmap = block_dirty_bitmap_lookup(node, name, &bs, errp);
     if (!bitmap || !bs) {
         return;
     }
@@ -2811,18 +2784,15 @@ void qmp_block_dirty_bitmap_clear(const char *node, const char *name,
         error_setg(errp,
                    "Bitmap '%s' is currently frozen and cannot be modified",
                    name);
-        goto out;
+        return;
     } else if (!bdrv_dirty_bitmap_enabled(bitmap)) {
         error_setg(errp,
                    "Bitmap '%s' is currently disabled and cannot be cleared",
                    name);
-        goto out;
+        return;
     }
 
     bdrv_clear_dirty_bitmap(bitmap, NULL);
-
- out:
-    aio_context_release(aio_context);
 }
 
 void hmp_drive_del(Monitor *mon, const QDict *qdict)

include/block/accounting.h

@@ -26,8 +26,10 @@
 #define BLOCK_ACCOUNTING_H
 
 #include "qemu/timed-average.h"
+#include "qemu/thread.h"
 
 typedef struct BlockAcctTimedStats BlockAcctTimedStats;
+typedef struct BlockAcctStats BlockAcctStats;
 
 enum BlockAcctType {
     BLOCK_ACCT_READ,
@@ -37,12 +39,14 @@ enum BlockAcctType {
 };
 
 struct BlockAcctTimedStats {
+    BlockAcctStats *stats;
     TimedAverage latency[BLOCK_MAX_IOTYPE];
     unsigned interval_length; /* in seconds */
     QSLIST_ENTRY(BlockAcctTimedStats) entries;
 };
 
-typedef struct BlockAcctStats {
+struct BlockAcctStats {
+    QemuMutex lock;
     uint64_t nr_bytes[BLOCK_MAX_IOTYPE];
     uint64_t nr_ops[BLOCK_MAX_IOTYPE];
     uint64_t invalid_ops[BLOCK_MAX_IOTYPE];
@@ -53,7 +57,7 @@ typedef struct BlockAcctStats {
     QSLIST_HEAD(, BlockAcctTimedStats) intervals;
     bool account_invalid;
     bool account_failed;
-} BlockAcctStats;
+};
 
 typedef struct BlockAcctCookie {
     int64_t bytes;
@@ -61,7 +65,8 @@ typedef struct BlockAcctCookie {
     enum BlockAcctType type;
 } BlockAcctCookie;
 
-void block_acct_init(BlockAcctStats *stats, bool account_invalid,
+void block_acct_init(BlockAcctStats *stats);
+void block_acct_setup(BlockAcctStats *stats, bool account_invalid,
                      bool account_failed);
 void block_acct_cleanup(BlockAcctStats *stats);
 void block_acct_add_interval(BlockAcctStats *stats, unsigned interval_length);

include/block/block.h

@@ -402,7 +402,8 @@ void bdrv_drain_all(void);
          * block_job_defer_to_main_loop for how to do it). \
          */                                                \
         assert(!bs_->wakeup);                              \
-        bs_->wakeup = true;                                \
+        /* Set bs->wakeup before evaluating cond.  */      \
+        atomic_mb_set(&bs_->wakeup, true);                 \
         while (busy_) {                                    \
             if ((cond)) {                                  \
                 waited_ = busy_ = true;                    \
@@ -414,7 +415,7 @@ void bdrv_drain_all(void);
                 waited_ |= busy_;                          \
             }                                              \
         }                                                  \
-        bs_->wakeup = false;                               \
+        atomic_set(&bs_->wakeup, false);                   \
     }                                                      \
     waited_; })
 

include/block/block_int.h

@@ -29,6 +29,7 @@
 #include "qemu/option.h"
 #include "qemu/queue.h"
 #include "qemu/coroutine.h"
+#include "qemu/stats64.h"
 #include "qemu/timer.h"
 #include "qapi-types.h"
 #include "qemu/hbitmap.h"
@@ -595,11 +596,6 @@ struct BlockDriverState {
 
     /* Protected by AioContext lock */
 
-    /* If true, copy read backing sectors into image.  Can be >1 if more
-     * than one client has requested copy-on-read.
-     */
-    int copy_on_read;
-
     /* If we are reading a disk image, give its size in sectors.
      * Generally read-only; it is written to by load_snapshot and
      * save_snaphost, but the block layer is quiescent during those.
@@ -609,34 +605,57 @@ struct BlockDriverState {
     /* Callback before write request is processed */
     NotifierWithReturnList before_write_notifiers;
 
-    /* number of in-flight requests; overall and serialising */
-    unsigned int in_flight;
-    unsigned int serialising_in_flight;
-
-    bool wakeup;
-
-    /* Offset after the highest byte written to */
-    uint64_t wr_highest_offset;
-
     /* threshold limit for writes, in bytes. "High water mark". */
     uint64_t write_threshold_offset;
     NotifierWithReturn write_threshold_notifier;
 
-    /* counter for nested bdrv_io_plug */
+    /* Writing to the list requires the BQL _and_ the dirty_bitmap_mutex.
-    unsigned io_plugged;
+     * Reading from the list can be done with either the BQL or the
-
+     * dirty_bitmap_mutex.  Modifying a bitmap only requires
-    QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
+     * dirty_bitmap_mutex.  */
-    CoQueue flush_queue;                  /* Serializing flush queue */
+    QemuMutex dirty_bitmap_mutex;
-    bool active_flush_req;                /* Flush request in flight? */
-    unsigned int write_gen;               /* Current data generation */
-    unsigned int flushed_gen;             /* Flushed write generation */
-
     QLIST_HEAD(, BdrvDirtyBitmap) dirty_bitmaps;
 
+    /* Offset after the highest byte written to */
+    Stat64 wr_highest_offset;
+
+    /* If true, copy read backing sectors into image.  Can be >1 if more
+     * than one client has requested copy-on-read.  Accessed with atomic
+     * ops.
+     */
+    int copy_on_read;
+
+    /* number of in-flight requests; overall and serialising.
+     * Accessed with atomic ops.
+     */
+    unsigned int in_flight;
+    unsigned int serialising_in_flight;
+
+    /* Internal to BDRV_POLL_WHILE and bdrv_wakeup.  Accessed with atomic
+     * ops.
+     */
+    bool wakeup;
+
+    /* counter for nested bdrv_io_plug.
+     * Accessed with atomic ops.
+    */
+    unsigned io_plugged;
+
     /* do we need to tell the quest if we have a volatile write cache? */
     int enable_write_cache;
 
+    /* Accessed with atomic ops.  */
     int quiesce_counter;
+    unsigned int write_gen;               /* Current data generation */
+
+    /* Protected by reqs_lock.  */
+    CoMutex reqs_lock;
+    QLIST_HEAD(, BdrvTrackedRequest) tracked_requests;
+    CoQueue flush_queue;                  /* Serializing flush queue */
+    bool active_flush_req;                /* Flush request in flight? */
+
+    /* Only read/written by whoever has set active_flush_req to true.  */
+    unsigned int flushed_gen;             /* Flushed write generation */
 };
 
 struct BlockBackendRootState {

include/block/dirty-bitmap.h

@@ -36,8 +36,6 @@ bool bdrv_dirty_bitmap_frozen(BdrvDirtyBitmap *bitmap);
 const char *bdrv_dirty_bitmap_name(const BdrvDirtyBitmap *bitmap);
 int64_t bdrv_dirty_bitmap_size(const BdrvDirtyBitmap *bitmap);
 DirtyBitmapStatus bdrv_dirty_bitmap_status(BdrvDirtyBitmap *bitmap);
-int bdrv_get_dirty(BlockDriverState *bs, BdrvDirtyBitmap *bitmap,
-                   int64_t sector);
 void bdrv_set_dirty_bitmap(BdrvDirtyBitmap *bitmap,
                            int64_t cur_sector, int64_t nr_sectors);
 void bdrv_reset_dirty_bitmap(BdrvDirtyBitmap *bitmap,
@@ -45,6 +43,9 @@ void bdrv_reset_dirty_bitmap(BdrvDirtyBitmap *bitmap,
 int bdrv_dirty_bitmap_get_meta(BlockDriverState *bs,
                                BdrvDirtyBitmap *bitmap, int64_t sector,
                                int nb_sectors);
+int bdrv_dirty_bitmap_get_meta_locked(BlockDriverState *bs,
+                                      BdrvDirtyBitmap *bitmap, int64_t sector,
+                                      int nb_sectors);
 void bdrv_dirty_bitmap_reset_meta(BlockDriverState *bs,
                                   BdrvDirtyBitmap *bitmap, int64_t sector,
                                   int nb_sectors);
@@ -52,11 +53,6 @@ BdrvDirtyBitmapIter *bdrv_dirty_meta_iter_new(BdrvDirtyBitmap *bitmap);
 BdrvDirtyBitmapIter *bdrv_dirty_iter_new(BdrvDirtyBitmap *bitmap,
                                          uint64_t first_sector);
 void bdrv_dirty_iter_free(BdrvDirtyBitmapIter *iter);
-int64_t bdrv_dirty_iter_next(BdrvDirtyBitmapIter *iter);
-void bdrv_set_dirty_iter(BdrvDirtyBitmapIter *hbi, int64_t sector_num);
-int64_t bdrv_get_dirty_count(BdrvDirtyBitmap *bitmap);
-int64_t bdrv_get_meta_dirty_count(BdrvDirtyBitmap *bitmap);
-void bdrv_dirty_bitmap_truncate(BlockDriverState *bs);
 
 uint64_t bdrv_dirty_bitmap_serialization_size(const BdrvDirtyBitmap *bitmap,
                                               uint64_t start, uint64_t count);
@@ -72,4 +68,19 @@ void bdrv_dirty_bitmap_deserialize_zeroes(BdrvDirtyBitmap *bitmap,
                                           bool finish);
 void bdrv_dirty_bitmap_deserialize_finish(BdrvDirtyBitmap *bitmap);
 
+/* Functions that require manual locking.  */
+void bdrv_dirty_bitmap_lock(BdrvDirtyBitmap *bitmap);
+void bdrv_dirty_bitmap_unlock(BdrvDirtyBitmap *bitmap);
+int bdrv_get_dirty_locked(BlockDriverState *bs, BdrvDirtyBitmap *bitmap,
+                          int64_t sector);
+void bdrv_set_dirty_bitmap_locked(BdrvDirtyBitmap *bitmap,
+                                  int64_t cur_sector, int64_t nr_sectors);
+void bdrv_reset_dirty_bitmap_locked(BdrvDirtyBitmap *bitmap,
+                                    int64_t cur_sector, int64_t nr_sectors);
+int64_t bdrv_dirty_iter_next(BdrvDirtyBitmapIter *iter);
+void bdrv_set_dirty_iter(BdrvDirtyBitmapIter *hbi, int64_t sector_num);
+int64_t bdrv_get_dirty_count(BdrvDirtyBitmap *bitmap);
+int64_t bdrv_get_meta_dirty_count(BdrvDirtyBitmap *bitmap);
+void bdrv_dirty_bitmap_truncate(BlockDriverState *bs);
+
 #endif

include/qemu/stats64.h

@@ -0,0 +1,193 @@
+/*
+ * Atomic operations on 64-bit quantities.
+ *
+ * Copyright (C) 2017 Red Hat, Inc.
+ *
+ * Author: Paolo Bonzini <pbonzini@redhat.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ */
+
+#ifndef QEMU_STATS64_H
+#define QEMU_STATS64_H 1
+
+#include "qemu/atomic.h"
+
+/* This provides atomic operations on 64-bit type, using a reader-writer
+ * spinlock on architectures that do not have 64-bit accesses.  Even on
+ * those architectures, it tries hard not to take the lock.
+ */
+
+typedef struct Stat64 {
+#ifdef CONFIG_ATOMIC64
+    uint64_t value;
+#else
+    uint32_t low, high;
+    uint32_t lock;
+#endif
+} Stat64;
+
+#ifdef CONFIG_ATOMIC64
+static inline void stat64_init(Stat64 *s, uint64_t value)
+{
+    /* This is not guaranteed to be atomic! */
+    *s = (Stat64) { value };
+}
+
+static inline uint64_t stat64_get(const Stat64 *s)
+{
+    return atomic_read__nocheck(&s->value);
+}
+
+static inline void stat64_add(Stat64 *s, uint64_t value)
+{
+    atomic_add(&s->value, value);
+}
+
+static inline void stat64_min(Stat64 *s, uint64_t value)
+{
+    uint64_t orig = atomic_read__nocheck(&s->value);
+    while (orig > value) {
+        orig = atomic_cmpxchg__nocheck(&s->value, orig, value);
+    }
+}
+
+static inline void stat64_max(Stat64 *s, uint64_t value)
+{
+    uint64_t orig = atomic_read__nocheck(&s->value);
+    while (orig < value) {
+        orig = atomic_cmpxchg__nocheck(&s->value, orig, value);
+    }
+}
+#else
+uint64_t stat64_get(const Stat64 *s);
+bool stat64_min_slow(Stat64 *s, uint64_t value);
+bool stat64_max_slow(Stat64 *s, uint64_t value);
+bool stat64_add32_carry(Stat64 *s, uint32_t low, uint32_t high);
+
+static inline void stat64_init(Stat64 *s, uint64_t value)
+{
+    /* This is not guaranteed to be atomic! */
+    *s = (Stat64) { .low = value, .high = value >> 32, .lock = 0 };
+}
+
+static inline void stat64_add(Stat64 *s, uint64_t value)
+{
+    uint32_t low, high;
+    high = value >> 32;
+    low = (uint32_t) value;
+    if (!low) {
+        if (high) {
+            atomic_add(&s->high, high);
+        }
+        return;
+    }
+
+    for (;;) {
+        uint32_t orig = s->low;
+        uint32_t result = orig + low;
+        uint32_t old;
+
+        if (result < low || high) {
+            /* If the high part is affected, take the lock.  */
+            if (stat64_add32_carry(s, low, high)) {
+                return;
+            }
+            continue;
+        }
+
+        /* No carry, try with a 32-bit cmpxchg.  The result is independent of
+         * the high 32 bits, so it can race just fine with stat64_add32_carry
+         * and even stat64_get!
+         */
+        old = atomic_cmpxchg(&s->low, orig, result);
+        if (orig == old) {
+            return;
+        }
+    }
+}
+
+static inline void stat64_min(Stat64 *s, uint64_t value)
+{
+    uint32_t low, high;
+    uint32_t orig_low, orig_high;
+
+    high = value >> 32;
+    low = (uint32_t) value;
+    do {
+        orig_high = atomic_read(&s->high);
+        if (orig_high < high) {
+            return;
+        }
+
+        if (orig_high == high) {
+            /* High 32 bits are equal.  Read low after high, otherwise we
+             * can get a false positive (e.g. 0x1235,0x0000 changes to
+             * 0x1234,0x8000 and we read it as 0x1234,0x0000). Pairs with
+             * the write barrier in stat64_min_slow.
+             */
+            smp_rmb();
+            orig_low = atomic_read(&s->low);
+            if (orig_low <= low) {
+                return;
+            }
+
+            /* See if we were lucky and a writer raced against us.  The
+             * barrier is theoretically unnecessary, but if we remove it
+             * we may miss being lucky.
+             */
+            smp_rmb();
+            orig_high = atomic_read(&s->high);
+            if (orig_high < high) {
+                return;
+            }
+        }
+
+        /* If the value changes in any way, we have to take the lock.  */
+    } while (!stat64_min_slow(s, value));
+}
+
+static inline void stat64_max(Stat64 *s, uint64_t value)
+{
+    uint32_t low, high;
+    uint32_t orig_low, orig_high;
+
+    high = value >> 32;
+    low = (uint32_t) value;
+    do {
+        orig_high = atomic_read(&s->high);
+        if (orig_high > high) {
+            return;
+        }
+
+        if (orig_high == high) {
+            /* High 32 bits are equal.  Read low after high, otherwise we
+             * can get a false positive (e.g. 0x1234,0x8000 changes to
+             * 0x1235,0x0000 and we read it as 0x1235,0x8000). Pairs with
+             * the write barrier in stat64_max_slow.
+             */
+            smp_rmb();
+            orig_low = atomic_read(&s->low);
+            if (orig_low >= low) {
+                return;
+            }
+
+            /* See if we were lucky and a writer raced against us.  The
+             * barrier is theoretically unnecessary, but if we remove it
+             * we may miss being lucky.
+             */
+            smp_rmb();
+            orig_high = atomic_read(&s->high);
+            if (orig_high > high) {
+                return;
+            }
+        }
+
+        /* If the value changes in any way, we have to take the lock.  */
+    } while (!stat64_max_slow(s, value));
+}
+
+#endif
+
+#endif

include/sysemu/block-backend.h

@@ -72,15 +72,13 @@ typedef struct BlockDevOps {
  * fields that must be public. This is in particular for QLIST_ENTRY() and
  * friends so that BlockBackends can be kept in lists outside block-backend.c */
 typedef struct BlockBackendPublic {
-    /* I/O throttling has its own locking, but also some fields are
+    /* throttled_reqs_lock protects the CoQueues for throttled requests.  */
-     * protected by the AioContext lock.
+    CoMutex      throttled_reqs_lock;
-     */
-
-    /* Protected by AioContext lock.  */
     CoQueue      throttled_reqs[2];
 
     /* Nonzero if the I/O limits are currently being ignored; generally
-     * it is zero.  */
+     * it is zero.  Accessed with atomic operations.
+     */
     unsigned int io_limits_disabled;
 
     /* The following fields are protected by the ThrottleGroup lock.

migration/block.c

@@ -341,10 +341,8 @@ static int set_dirty_tracking(void)
     int ret;
 
     QSIMPLEQ_FOREACH(bmds, &block_mig_state.bmds_list, entry) {
-        aio_context_acquire(blk_get_aio_context(bmds->blk));
         bmds->dirty_bitmap = bdrv_create_dirty_bitmap(blk_bs(bmds->blk),
                                                       BLOCK_SIZE, NULL, NULL);
-        aio_context_release(blk_get_aio_context(bmds->blk));
         if (!bmds->dirty_bitmap) {
             ret = -errno;
             goto fail;
@@ -355,9 +353,7 @@ static int set_dirty_tracking(void)
 fail:
     QSIMPLEQ_FOREACH(bmds, &block_mig_state.bmds_list, entry) {
         if (bmds->dirty_bitmap) {
-            aio_context_acquire(blk_get_aio_context(bmds->blk));
             bdrv_release_dirty_bitmap(blk_bs(bmds->blk), bmds->dirty_bitmap);
-            aio_context_release(blk_get_aio_context(bmds->blk));
         }
     }
     return ret;
@@ -370,9 +366,7 @@ static void unset_dirty_tracking(void)
     BlkMigDevState *bmds;
 
     QSIMPLEQ_FOREACH(bmds, &block_mig_state.bmds_list, entry) {
-        aio_context_acquire(blk_get_aio_context(bmds->blk));
         bdrv_release_dirty_bitmap(blk_bs(bmds->blk), bmds->dirty_bitmap);
-        aio_context_release(blk_get_aio_context(bmds->blk));
     }
 }
 
@@ -531,13 +525,16 @@ static int mig_save_device_dirty(QEMUFile *f, BlkMigDevState *bmds,
         } else {
             blk_mig_unlock();
         }
-        if (bdrv_get_dirty(bs, bmds->dirty_bitmap, sector)) {
+        bdrv_dirty_bitmap_lock(bmds->dirty_bitmap);
-
+        if (bdrv_get_dirty_locked(bs, bmds->dirty_bitmap, sector)) {
             if (total_sectors - sector < BDRV_SECTORS_PER_DIRTY_CHUNK) {
                 nr_sectors = total_sectors - sector;
             } else {
                 nr_sectors = BDRV_SECTORS_PER_DIRTY_CHUNK;
             }
+            bdrv_reset_dirty_bitmap_locked(bmds->dirty_bitmap, sector, nr_sectors);
+            bdrv_dirty_bitmap_unlock(bmds->dirty_bitmap);
+
             blk = g_new(BlkMigBlock, 1);
             blk->buf = g_malloc(BLOCK_SIZE);
             blk->bmds = bmds;
@@ -570,12 +567,12 @@ static int mig_save_device_dirty(QEMUFile *f, BlkMigDevState *bmds,
                 g_free(blk);
             }
 
-            bdrv_reset_dirty_bitmap(bmds->dirty_bitmap, sector, nr_sectors);
             sector += nr_sectors;
             bmds->cur_dirty = sector;
-
             break;
         }
+
+        bdrv_dirty_bitmap_unlock(bmds->dirty_bitmap);
         sector += BDRV_SECTORS_PER_DIRTY_CHUNK;
         bmds->cur_dirty = sector;
     }

tests/docker/Makefile.include

@@ -126,7 +126,7 @@ docker-run: docker-qemu-src
 			"  COPYING $(EXECUTABLE) to $(IMAGE)"))
 	$(call quiet-command,						\
 		$(SRC_PATH)/tests/docker/docker.py run 			\
-			-t 						\
+			$(if $(NOUSER),,-u $(shell id -u)) -t 		\
 			$(if $V,,--rm) 					\
 			$(if $(DEBUG),-i,--net=none) 			\
 			-e TARGET_LIST=$(TARGET_LIST) 			\

tests/docker/dockerfiles/centos6.docker

@@ -1,7 +1,7 @@
 FROM centos:6
 RUN yum install -y epel-release
 ENV PACKAGES libfdt-devel ccache \
-    tar git make gcc g++ \
+    tar git make gcc g++ flex bison \
     zlib-devel glib2-devel SDL-devel pixman-devel \
     epel-release
 RUN yum install -y $PACKAGES

tests/docker/dockerfiles/fedora.docker

@@ -1,8 +1,8 @@
 FROM fedora:latest
 ENV PACKAGES \
-    ccache git tar PyYAML sparse flex bison python2 \
+    ccache git tar PyYAML sparse flex bison python2 bzip2 hostname \
     glib2-devel pixman-devel zlib-devel SDL-devel libfdt-devel \
-    gcc gcc-c++ clang make perl which bc findutils \
+    gcc gcc-c++ clang make perl which bc findutils libaio-devel \
     mingw32-pixman mingw32-glib2 mingw32-gmp mingw32-SDL mingw32-pkg-config \
     mingw32-gtk2 mingw32-gtk3 mingw32-gnutls mingw32-nettle mingw32-libtasn1 \
     mingw32-libjpeg-turbo mingw32-libpng mingw32-curl mingw32-libssh2 \

util/Makefile.objs

@@ -42,4 +42,5 @@ util-obj-y += log.o
 util-obj-y += qdist.o
 util-obj-y += qht.o
 util-obj-y += range.o
+util-obj-y += stats64.o
 util-obj-y += systemd.o

util/stats64.c

@@ -0,0 +1,137 @@
+/*
+ * Atomic operations on 64-bit quantities.
+ *
+ * Copyright (C) 2017 Red Hat, Inc.
+ *
+ * Author: Paolo Bonzini <pbonzini@redhat.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ */
+
+#include "qemu/osdep.h"
+#include "qemu/atomic.h"
+#include "qemu/stats64.h"
+#include "qemu/processor.h"
+
+#ifndef CONFIG_ATOMIC64
+static inline void stat64_rdlock(Stat64 *s)
+{
+    /* Keep out incoming writers to avoid them starving us. */
+    atomic_add(&s->lock, 2);
+
+    /* If there is a concurrent writer, wait for it.  */
+    while (atomic_read(&s->lock) & 1) {
+        cpu_relax();
+    }
+}
+
+static inline void stat64_rdunlock(Stat64 *s)
+{
+    atomic_sub(&s->lock, 2);
+}
+
+static inline bool stat64_wrtrylock(Stat64 *s)
+{
+    return atomic_cmpxchg(&s->lock, 0, 1) == 0;
+}
+
+static inline void stat64_wrunlock(Stat64 *s)
+{
+    atomic_dec(&s->lock);
+}
+
+uint64_t stat64_get(const Stat64 *s)
+{
+    uint32_t high, low;
+
+    stat64_rdlock((Stat64 *)s);
+
+    /* 64-bit writes always take the lock, so we can read in
+     * any order.
+     */
+    high = atomic_read(&s->high);
+    low = atomic_read(&s->low);
+    stat64_rdunlock((Stat64 *)s);
+
+    return ((uint64_t)high << 32) | low;
+}
+
+bool stat64_add32_carry(Stat64 *s, uint32_t low, uint32_t high)
+{
+    uint32_t old;
+
+    if (!stat64_wrtrylock(s)) {
+        cpu_relax();
+        return false;
+    }
+
+    /* 64-bit reads always take the lock, so they don't care about the
+     * order of our update.  By updating s->low first, we can check
+     * whether we have to carry into s->high.
+     */
+    old = atomic_fetch_add(&s->low, low);
+    high += (old + low) < old;
+    atomic_add(&s->high, high);
+    stat64_wrunlock(s);
+    return true;
+}
+
+bool stat64_min_slow(Stat64 *s, uint64_t value)
+{
+    uint32_t high, low;
+    uint64_t orig;
+
+    if (!stat64_wrtrylock(s)) {
+        cpu_relax();
+        return false;
+    }
+
+    high = atomic_read(&s->high);
+    low = atomic_read(&s->low);
+
+    orig = ((uint64_t)high << 32) | low;
+    if (orig < value) {
+        /* We have to set low before high, just like stat64_min reads
+         * high before low.  The value may become higher temporarily, but
+         * stat64_get does not notice (it takes the lock) and the only ill
+         * effect on stat64_min is that the slow path may be triggered
+         * unnecessarily.
+         */
+        atomic_set(&s->low, (uint32_t)value);
+        smp_wmb();
+        atomic_set(&s->high, value >> 32);
+    }
+    stat64_wrunlock(s);
+    return true;
+}
+
+bool stat64_max_slow(Stat64 *s, uint64_t value)
+{
+    uint32_t high, low;
+    uint64_t orig;
+
+    if (!stat64_wrtrylock(s)) {
+        cpu_relax();
+        return false;
+    }
+
+    high = atomic_read(&s->high);
+    low = atomic_read(&s->low);
+
+    orig = ((uint64_t)high << 32) | low;
+    if (orig > value) {
+        /* We have to set low before high, just like stat64_max reads
+         * high before low.  The value may become lower temporarily, but
+         * stat64_get does not notice (it takes the lock) and the only ill
+         * effect on stat64_max is that the slow path may be triggered
+         * unnecessarily.
+         */
+        atomic_set(&s->low, (uint32_t)value);
+        smp_wmb();
+        atomic_set(&s->high, value >> 32);
+    }
+    stat64_wrunlock(s);
+    return true;
+}
+#endif