block: Move throttling fields from BDS to BB

This patch changes where the throttling state is stored (used to be the
BlockDriverState, now it is the BlockBackend), but it doesn't actually
make it a BB level feature yet. For example, throttling is still
disabled when the BDS is detached from the BB.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>

4 files changed, 102 insertions, 78 deletions

diff --git a/block/block-backend.c b/block/block-backend.c
index 964a205d38..6880659665 100644
--- a/block/block-backend.c
+++ b/block/block-backend.c
@@ -107,8 +107,12 @@ BlockBackend *blk_new(Error **errp)
 
     blk = g_new0(BlockBackend, 1);
     blk->refcnt = 1;
+    qemu_co_queue_init(&blk->public.throttled_reqs[0]);
+    qemu_co_queue_init(&blk->public.throttled_reqs[1]);
+
     notifier_list_init(&blk->remove_bs_notifiers);
     notifier_list_init(&blk->insert_bs_notifiers);
+
     QTAILQ_INSERT_TAIL(&block_backends, blk, link);
     return blk;
 }
@@ -437,7 +441,7 @@ void blk_remove_bs(BlockBackend *blk)
     notifier_list_notify(&blk->remove_bs_notifiers, blk);
 
     blk_update_root_state(blk);
-    if (blk->root->bs->throttle_state) {
+    if (blk->public.throttle_state) {
         bdrv_io_limits_disable(blk->root->bs);
     }
 
@@ -795,7 +799,6 @@ static int blk_prw(BlockBackend *blk, int64_t offset, uint8_t *buf,
 int blk_pread_unthrottled(BlockBackend *blk, int64_t offset, uint8_t *buf,
                           int count)
 {
-    BlockDriverState *bs = blk_bs(blk);
     int ret;
 
     ret = blk_check_byte_request(blk, offset, count);
@@ -803,9 +806,9 @@ int blk_pread_unthrottled(BlockBackend *blk, int64_t offset, uint8_t *buf,
         return ret;
     }
 
-    bdrv_no_throttling_begin(bs);
+    bdrv_no_throttling_begin(blk_bs(blk));
     ret = blk_pread(blk, offset, buf, count);
-    bdrv_no_throttling_end(bs);
+    bdrv_no_throttling_end(blk_bs(blk));
     return ret;
 }
 
@@ -1524,7 +1527,7 @@ void blk_update_root_state(BlockBackend *blk)
         g_free(blk->root_state.throttle_group);
         throttle_group_unref(blk->root_state.throttle_state);
     }
-    if (blk->root->bs->throttle_state) {
+    if (blk->public.throttle_state) {
         const char *name = throttle_group_get_name(blk);
         blk->root_state.throttle_group = g_strdup(name);
         blk->root_state.throttle_state = throttle_group_incref(name);
diff --git a/block/io.c b/block/io.c
index f6fb86883a..bdbaa1c0da 100644
--- a/block/io.c
+++ b/block/io.c
@@ -55,20 +55,31 @@ void bdrv_set_io_limits(BlockDriverState *bs,
 
 void bdrv_no_throttling_begin(BlockDriverState *bs)
 {
-    if (bs->io_limits_disabled++ == 0) {
-        throttle_group_restart_bs(bs);
+    if (!bs->blk) {
+        return;
+    }
+
+    if (blk_get_public(bs->blk)->io_limits_disabled++ == 0) {
+        throttle_group_restart_blk(bs->blk);
     }
 }
 
 void bdrv_no_throttling_end(BlockDriverState *bs)
 {
-    assert(bs->io_limits_disabled);
-    --bs->io_limits_disabled;
+    BlockBackendPublic *blkp;
+
+    if (!bs->blk) {
+        return;
+    }
+
+    blkp = blk_get_public(bs->blk);
+    assert(blkp->io_limits_disabled);
+    --blkp->io_limits_disabled;
 }
 
 void bdrv_io_limits_disable(BlockDriverState *bs)
 {
-    assert(bs->throttle_state);
+    assert(blk_get_public(bs->blk)->throttle_state);
     bdrv_no_throttling_begin(bs);
     throttle_group_unregister_blk(bs->blk);
     bdrv_no_throttling_end(bs);
@@ -77,14 +88,16 @@ void bdrv_io_limits_disable(BlockDriverState *bs)
 /* should be called before bdrv_set_io_limits if a limit is set */
 void bdrv_io_limits_enable(BlockDriverState *bs, const char *group)
 {
-    assert(!bs->throttle_state);
+    BlockBackendPublic *blkp = blk_get_public(bs->blk);
+
+    assert(!blkp->throttle_state);
     throttle_group_register_blk(bs->blk, group);
 }
 
 void bdrv_io_limits_update_group(BlockDriverState *bs, const char *group)
 {
     /* this bs is not part of any group */
-    if (!bs->throttle_state) {
+    if (!blk_get_public(bs->blk)->throttle_state) {
         return;
     }
 
@@ -178,14 +191,15 @@ void bdrv_disable_copy_on_read(BlockDriverState *bs)
 bool bdrv_requests_pending(BlockDriverState *bs)
 {
     BdrvChild *child;
+    BlockBackendPublic *blkp = bs->blk ? blk_get_public(bs->blk) : NULL;
 
     if (!QLIST_EMPTY(&bs->tracked_requests)) {
         return true;
     }
-    if (!qemu_co_queue_empty(&bs->throttled_reqs[0])) {
+    if (blkp && !qemu_co_queue_empty(&blkp->throttled_reqs[0])) {
         return true;
     }
-    if (!qemu_co_queue_empty(&bs->throttled_reqs[1])) {
+    if (blkp && !qemu_co_queue_empty(&blkp->throttled_reqs[1])) {
         return true;
     }
 
@@ -1070,7 +1084,7 @@ int coroutine_fn bdrv_co_preadv(BlockDriverState *bs,
     }
 
     /* throttling disk I/O */
-    if (bs->throttle_state) {
+    if (bs->blk && blk_get_public(bs->blk)->throttle_state) {
         throttle_group_co_io_limits_intercept(bs, bytes, false);
     }
 
@@ -1431,7 +1445,7 @@ int coroutine_fn bdrv_co_pwritev(BlockDriverState *bs,
     }
 
     /* throttling disk I/O */
-    if (bs->throttle_state) {
+    if (bs->blk && blk_get_public(bs->blk)->throttle_state) {
         throttle_group_co_io_limits_intercept(bs, bytes, true);
     }
 
diff --git a/block/qapi.c b/block/qapi.c
index a3e514d89b..1e4bb8a0a5 100644
--- a/block/qapi.c
+++ b/block/qapi.c
@@ -67,7 +67,7 @@ BlockDeviceInfo *bdrv_block_device_info(BlockBackend *blk,
     info->backing_file_depth = bdrv_get_backing_file_depth(bs);
     info->detect_zeroes = bs->detect_zeroes;
 
-    if (bs->throttle_state) {
+    if (bs->blk && blk_get_public(bs->blk)->throttle_state) {
         ThrottleConfig cfg;
 
         throttle_group_get_config(bs, &cfg);
diff --git a/block/throttle-groups.c b/block/throttle-groups.c
index e50ccaaf7e..56dc311867 100644
--- a/block/throttle-groups.c
+++ b/block/throttle-groups.c
@@ -30,7 +30,7 @@
 #include "sysemu/qtest.h"
 
 /* The ThrottleGroup structure (with its ThrottleState) is shared
- * among different BlockDriverState and it's independent from
+ * among different BlockBackends and it's independent from
  * AioContext, so in order to use it from different threads it needs
  * its own locking.
  *
@@ -40,18 +40,18 @@
  * The whole ThrottleGroup structure is private and invisible to
  * outside users, that only use it through its ThrottleState.
  *
- * In addition to the ThrottleGroup structure, BlockDriverState has
+ * In addition to the ThrottleGroup structure, BlockBackendPublic has
  * fields that need to be accessed by other members of the group and
- * therefore also need to be protected by this lock. Once a BDS is
- * registered in a group those fields can be accessed by other threads
- * any time.
+ * therefore also need to be protected by this lock. Once a
+ * BlockBackend is registered in a group those fields can be accessed
+ * by other threads any time.
  *
  * Again, all this is handled internally and is mostly transparent to
  * the outside. The 'throttle_timers' field however has an additional
  * constraint because it may be temporarily invalid (see for example
  * bdrv_set_aio_context()). Therefore in this file a thread will
- * access some other BDS's timers only after verifying that that BDS
- * has throttled requests in the queue.
+ * access some other BlockBackend's timers only after verifying that
+ * that BlockBackend has throttled requests in the queue.
  */
 typedef struct ThrottleGroup {
     char *name; /* This is constant during the lifetime of the group */
@@ -141,8 +141,8 @@ void throttle_group_unref(ThrottleState *ts)
  */
 const char *throttle_group_get_name(BlockBackend *blk)
 {
-    ThrottleGroup *tg = container_of(blk_bs(blk)->throttle_state,
-                                     ThrottleGroup, ts);
+    BlockBackendPublic *blkp = blk_get_public(blk);
+    ThrottleGroup *tg = container_of(blkp->throttle_state, ThrottleGroup, ts);
     return tg->name;
 }
 
@@ -156,10 +156,10 @@ const char *throttle_group_get_name(BlockBackend *blk)
  */
 static BlockBackend *throttle_group_next_blk(BlockBackend *blk)
 {
-    BlockDriverState *bs = blk_bs(blk);
-    ThrottleState *ts = bs->throttle_state;
+    BlockBackendPublic *blkp = blk_get_public(blk);
+    ThrottleState *ts = blkp->throttle_state;
     ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
-    BlockBackendPublic *next = QLIST_NEXT(blk_get_public(blk), round_robin);
+    BlockBackendPublic *next = QLIST_NEXT(blkp, round_robin);
 
     if (!next) {
         next = QLIST_FIRST(&tg->head);
@@ -180,15 +180,15 @@ static BlockBackend *throttle_group_next_blk(BlockBackend *blk)
  */
 static BlockBackend *next_throttle_token(BlockBackend *blk, bool is_write)
 {
-    ThrottleGroup *tg = container_of(blk_bs(blk)->throttle_state,
-                                     ThrottleGroup, ts);
+    BlockBackendPublic *blkp = blk_get_public(blk);
+    ThrottleGroup *tg = container_of(blkp->throttle_state, ThrottleGroup, ts);
     BlockBackend *token, *start;
 
     start = token = tg->tokens[is_write];
 
     /* get next bs round in round robin style */
     token = throttle_group_next_blk(token);
-    while (token != start && !blk_bs(token)->pending_reqs[is_write]) {
+    while (token != start && !blkp->pending_reqs[is_write]) {
         token = throttle_group_next_blk(token);
     }
 
@@ -196,7 +196,7 @@ static BlockBackend *next_throttle_token(BlockBackend *blk, bool is_write)
      * then decide the token is the current bs because chances are
      * the current bs get the current request queued.
      */
-    if (token == start && !blk_bs(token)->pending_reqs[is_write]) {
+    if (token == start && !blkp->pending_reqs[is_write]) {
         token = blk;
     }
 
@@ -215,12 +215,13 @@ static BlockBackend *next_throttle_token(BlockBackend *blk, bool is_write)
  */
 static bool throttle_group_schedule_timer(BlockBackend *blk, bool is_write)
 {
-    ThrottleState *ts = blk_bs(blk)->throttle_state;
-    ThrottleTimers *tt = &blk_bs(blk)->throttle_timers;
+    BlockBackendPublic *blkp = blk_get_public(blk);
+    ThrottleState *ts = blkp->throttle_state;
+    ThrottleTimers *tt = &blkp->throttle_timers;
     ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
     bool must_wait;
 
-    if (blk_bs(blk)->io_limits_disabled) {
+    if (blkp->io_limits_disabled) {
         return false;
     }
 
@@ -249,14 +250,14 @@ static bool throttle_group_schedule_timer(BlockBackend *blk, bool is_write)
  */
 static void schedule_next_request(BlockBackend *blk, bool is_write)
 {
-    BlockDriverState *bs = blk_bs(blk);
-    ThrottleGroup *tg = container_of(bs->throttle_state, ThrottleGroup, ts);
+    BlockBackendPublic *blkp = blk_get_public(blk);
+    ThrottleGroup *tg = container_of(blkp->throttle_state, ThrottleGroup, ts);
     bool must_wait;
     BlockBackend *token;
 
     /* Check if there's any pending request to schedule next */
     token = next_throttle_token(blk, is_write);
-    if (!blk_bs(token)->pending_reqs[is_write]) {
+    if (!blkp->pending_reqs[is_write]) {
         return;
     }
 
@@ -265,12 +266,12 @@ static void schedule_next_request(BlockBackend *blk, bool is_write)
 
     /* If it doesn't have to wait, queue it for immediate execution */
     if (!must_wait) {
-        /* Give preference to requests from the current bs */
+        /* Give preference to requests from the current blk */
         if (qemu_in_coroutine() &&
-            qemu_co_queue_next(&bs->throttled_reqs[is_write])) {
+            qemu_co_queue_next(&blkp->throttled_reqs[is_write])) {
             token = blk;
         } else {
-            ThrottleTimers *tt = &blk_bs(token)->throttle_timers;
+            ThrottleTimers *tt = &blkp->throttle_timers;
             int64_t now = qemu_clock_get_ns(tt->clock_type);
             timer_mod(tt->timers[is_write], now + 1);
             tg->any_timer_armed[is_write] = true;
@@ -294,37 +295,40 @@ void coroutine_fn throttle_group_co_io_limits_intercept(BlockDriverState *bs,
     bool must_wait;
     BlockBackend *token;
 
-    ThrottleGroup *tg = container_of(bs->throttle_state, ThrottleGroup, ts);
+    BlockBackend *blk = bs->blk;
+    BlockBackendPublic *blkp = blk_get_public(blk);
+    ThrottleGroup *tg = container_of(blkp->throttle_state, ThrottleGroup, ts);
     qemu_mutex_lock(&tg->lock);
 
     /* First we check if this I/O has to be throttled. */
-    token = next_throttle_token(bs->blk, is_write);
+    token = next_throttle_token(blk, is_write);
     must_wait = throttle_group_schedule_timer(token, is_write);
 
     /* Wait if there's a timer set or queued requests of this type */
-    if (must_wait || bs->pending_reqs[is_write]) {
-        bs->pending_reqs[is_write]++;
+    if (must_wait || blkp->pending_reqs[is_write]) {
+        blkp->pending_reqs[is_write]++;
         qemu_mutex_unlock(&tg->lock);
-        qemu_co_queue_wait(&bs->throttled_reqs[is_write]);
+        qemu_co_queue_wait(&blkp->throttled_reqs[is_write]);
         qemu_mutex_lock(&tg->lock);
-        bs->pending_reqs[is_write]--;
+        blkp->pending_reqs[is_write]--;
     }
 
     /* The I/O will be executed, so do the accounting */
-    throttle_account(bs->throttle_state, is_write, bytes);
+    throttle_account(blkp->throttle_state, is_write, bytes);
 
     /* Schedule the next request */
-    schedule_next_request(bs->blk, is_write);
+    schedule_next_request(blk, is_write);
 
     qemu_mutex_unlock(&tg->lock);
 }
 
-void throttle_group_restart_bs(BlockDriverState *bs)
+void throttle_group_restart_blk(BlockBackend *blk)
 {
+    BlockBackendPublic *blkp = blk_get_public(blk);
     int i;
 
     for (i = 0; i < 2; i++) {
-        while (qemu_co_enter_next(&bs->throttled_reqs[i])) {
+        while (qemu_co_enter_next(&blkp->throttled_reqs[i])) {
             ;
         }
     }
@@ -339,8 +343,9 @@ void throttle_group_restart_bs(BlockDriverState *bs)
  */
 void throttle_group_config(BlockDriverState *bs, ThrottleConfig *cfg)
 {
-    ThrottleTimers *tt = &bs->throttle_timers;
-    ThrottleState *ts = bs->throttle_state;
+    BlockBackendPublic *blkp = blk_get_public(bs->blk);
+    ThrottleTimers *tt = &blkp->throttle_timers;
+    ThrottleState *ts = blkp->throttle_state;
     ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
     qemu_mutex_lock(&tg->lock);
     /* throttle_config() cancels the timers */
@@ -353,8 +358,8 @@ void throttle_group_config(BlockDriverState *bs, ThrottleConfig *cfg)
     throttle_config(ts, tt, cfg);
     qemu_mutex_unlock(&tg->lock);
 
-    qemu_co_enter_next(&bs->throttled_reqs[0]);
-    qemu_co_enter_next(&bs->throttled_reqs[1]);
+    qemu_co_enter_next(&blkp->throttled_reqs[0]);
+    qemu_co_enter_next(&blkp->throttled_reqs[1]);
 }
 
 /* Get the throttle configuration from a particular group. Similar to
@@ -366,7 +371,8 @@ void throttle_group_config(BlockDriverState *bs, ThrottleConfig *cfg)
  */
 void throttle_group_get_config(BlockDriverState *bs, ThrottleConfig *cfg)
 {
-    ThrottleState *ts = bs->throttle_state;
+    BlockBackendPublic *blkp = blk_get_public(bs->blk);
+    ThrottleState *ts = blkp->throttle_state;
     ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
     qemu_mutex_lock(&tg->lock);
     throttle_get_config(ts, cfg);
@@ -376,12 +382,13 @@ void throttle_group_get_config(BlockDriverState *bs, ThrottleConfig *cfg)
 /* ThrottleTimers callback. This wakes up a request that was waiting
  * because it had been throttled.
  *
- * @bs:        the BlockDriverState whose request had been throttled
+ * @blk:       the BlockBackend whose request had been throttled
  * @is_write:  the type of operation (read/write)
  */
-static void timer_cb(BlockDriverState *bs, bool is_write)
+static void timer_cb(BlockBackend *blk, bool is_write)
 {
-    ThrottleState *ts = bs->throttle_state;
+    BlockBackendPublic *blkp = blk_get_public(blk);
+    ThrottleState *ts = blkp->throttle_state;
     ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
     bool empty_queue;
 
@@ -391,13 +398,13 @@ static void timer_cb(BlockDriverState *bs, bool is_write)
     qemu_mutex_unlock(&tg->lock);
 
     /* Run the request that was waiting for this timer */
-    empty_queue = !qemu_co_enter_next(&bs->throttled_reqs[is_write]);
+    empty_queue = !qemu_co_enter_next(&blkp->throttled_reqs[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(bs->blk, is_write);
+        schedule_next_request(blk, is_write);
         qemu_mutex_unlock(&tg->lock);
     }
 }
@@ -422,7 +429,7 @@ static void write_timer_cb(void *opaque)
 void throttle_group_register_blk(BlockBackend *blk, const char *groupname)
 {
     int i;
-    BlockDriverState *bs = blk_bs(blk);
+    BlockBackendPublic *blkp = blk_get_public(blk);
     ThrottleState *ts = throttle_group_incref(groupname);
     ThrottleGroup *tg = container_of(ts, ThrottleGroup, ts);
     int clock_type = QEMU_CLOCK_REALTIME;
@@ -432,7 +439,7 @@ void throttle_group_register_blk(BlockBackend *blk, const char *groupname)
         clock_type = QEMU_CLOCK_VIRTUAL;
     }
 
-    bs->throttle_state = ts;
+    blkp->throttle_state = ts;
 
     qemu_mutex_lock(&tg->lock);
     /* If the ThrottleGroup is new set this BlockBackend as the token */
@@ -442,14 +449,14 @@ void throttle_group_register_blk(BlockBackend *blk, const char *groupname)
         }
     }
 
-    QLIST_INSERT_HEAD(&tg->head, blk_get_public(blk), round_robin);
+    QLIST_INSERT_HEAD(&tg->head, blkp, round_robin);
 
-    throttle_timers_init(&bs->throttle_timers,
-                         bdrv_get_aio_context(bs),
+    throttle_timers_init(&blkp->throttle_timers,
+                         blk_get_aio_context(blk),
                          clock_type,
                          read_timer_cb,
                          write_timer_cb,
-                         bs);
+                         blk);
 
     qemu_mutex_unlock(&tg->lock);
 }
@@ -466,19 +473,19 @@ void throttle_group_register_blk(BlockBackend *blk, const char *groupname)
  */
 void throttle_group_unregister_blk(BlockBackend *blk)
 {
-    BlockDriverState *bs = blk_bs(blk);
-    ThrottleGroup *tg = container_of(bs->throttle_state, ThrottleGroup, ts);
+    BlockBackendPublic *blkp = blk_get_public(blk);
+    ThrottleGroup *tg = container_of(blkp->throttle_state, ThrottleGroup, ts);
     int i;
 
-    assert(bs->pending_reqs[0] == 0 && bs->pending_reqs[1] == 0);
-    assert(qemu_co_queue_empty(&bs->throttled_reqs[0]));
-    assert(qemu_co_queue_empty(&bs->throttled_reqs[1]));
+    assert(blkp->pending_reqs[0] == 0 && blkp->pending_reqs[1] == 0);
+    assert(qemu_co_queue_empty(&blkp->throttled_reqs[0]));
+    assert(qemu_co_queue_empty(&blkp->throttled_reqs[1]));
 
     qemu_mutex_lock(&tg->lock);
     for (i = 0; i < 2; i++) {
         if (tg->tokens[i] == blk) {
             BlockBackend *token = throttle_group_next_blk(blk);
-            /* Take care of the case where this is the last bs in the group */
+            /* Take care of the case where this is the last blk in the group */
             if (token == blk) {
                 token = NULL;
             }
@@ -486,13 +493,13 @@ void throttle_group_unregister_blk(BlockBackend *blk)
         }
     }
 
-    /* remove the current bs from the list */
-    QLIST_REMOVE(blk_get_public(blk), round_robin);
-    throttle_timers_destroy(&bs->throttle_timers);
+    /* remove the current blk from the list */
+    QLIST_REMOVE(blkp, round_robin);
+    throttle_timers_destroy(&blkp->throttle_timers);
     qemu_mutex_unlock(&tg->lock);
 
     throttle_group_unref(&tg->ts);
-    bs->throttle_state = NULL;
+    blkp->throttle_state = NULL;
 }
 
 static void throttle_groups_init(void)