From 013989ed30ff32eedbb1a6767f0c38c85150f951 Mon Sep 17 00:00:00 2001
From: Kevin Wolf <kwolf@redhat.com>
Date: Wed, 10 Oct 2018 20:21:39 +0100
Subject: [PATCH 13/49] block: Drain recursively with a single
 BDRV_POLL_WHILE()

RH-Author: Kevin Wolf <kwolf@redhat.com>
Message-id: <20181010202213.7372-1-kwolf@redhat.com>
Patchwork-id: 82590
O-Subject: [RHEL-8 qemu-kvm PATCH 10/44] block: Drain recursively with a single BDRV_POLL_WHILE()
Bugzilla: 1637976
RH-Acked-by: Max Reitz <mreitz@redhat.com>
RH-Acked-by: John Snow <jsnow@redhat.com>
RH-Acked-by: Thomas Huth <thuth@redhat.com>

Anything can happen inside BDRV_POLL_WHILE(), including graph
changes that may interfere with its callers (e.g. child list iteration
in recursive callers of bdrv_do_drained_begin).

Switch to a single BDRV_POLL_WHILE() call for the whole subtree at the
end of bdrv_do_drained_begin() to avoid such effects. The recursion
happens now inside the loop condition. As the graph can only change
between bdrv_drain_poll() calls, but not inside of it, doing the
recursion here is safe.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
(cherry picked from commit fe4f0614ef9e361dae12012d3c400657444836cf)
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Danilo C. L. de Paula <ddepaula@redhat.com>
---
 block.c               |  2 +-
 block/io.c            | 63 ++++++++++++++++++++++++++++++++++++---------------
 include/block/block.h |  9 +++++---
 3 files changed, 52 insertions(+), 22 deletions(-)

diff --git a/block.c b/block.c
index 0d9698a..9afaf26 100644
--- a/block.c
+++ b/block.c
@@ -824,7 +824,7 @@ static void bdrv_child_cb_drained_begin(BdrvChild *child)
 static bool bdrv_child_cb_drained_poll(BdrvChild *child)
 {
     BlockDriverState *bs = child->opaque;
-    return bdrv_drain_poll(bs, NULL);
+    return bdrv_drain_poll(bs, false, NULL);
 }
 
 static void bdrv_child_cb_drained_end(BdrvChild *child)
diff --git a/block/io.c b/block/io.c
index a0e3461..442ded1 100644
--- a/block/io.c
+++ b/block/io.c
@@ -164,6 +164,7 @@ typedef struct {
     bool done;
     bool begin;
     bool recursive;
+    bool poll;
     BdrvChild *parent;
 } BdrvCoDrainData;
 
@@ -199,27 +200,42 @@ static void bdrv_drain_invoke(BlockDriverState *bs, bool begin)
 }
 
 /* Returns true if BDRV_POLL_WHILE() should go into a blocking aio_poll() */
-bool bdrv_drain_poll(BlockDriverState *bs, BdrvChild *ignore_parent)
+bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
+                     BdrvChild *ignore_parent)
 {
+    BdrvChild *child, *next;
+
     if (bdrv_parent_drained_poll(bs, ignore_parent)) {
         return true;
     }
 
-    return atomic_read(&bs->in_flight);
+    if (atomic_read(&bs->in_flight)) {
+        return true;
+    }
+
+    if (recursive) {
+        QLIST_FOREACH_SAFE(child, &bs->children, next, next) {
+            if (bdrv_drain_poll(child->bs, recursive, child)) {
+                return true;
+            }
+        }
+    }
+
+    return false;
 }
 
-static bool bdrv_drain_poll_top_level(BlockDriverState *bs,
+static bool bdrv_drain_poll_top_level(BlockDriverState *bs, bool recursive,
                                       BdrvChild *ignore_parent)
 {
     /* Execute pending BHs first and check everything else only after the BHs
      * have executed. */
     while (aio_poll(bs->aio_context, false));
 
-    return bdrv_drain_poll(bs, ignore_parent);
+    return bdrv_drain_poll(bs, recursive, ignore_parent);
 }
 
 static void bdrv_do_drained_begin(BlockDriverState *bs, bool recursive,
-                                  BdrvChild *parent);
+                                  BdrvChild *parent, bool poll);
 static void bdrv_do_drained_end(BlockDriverState *bs, bool recursive,
                                 BdrvChild *parent);
 
@@ -231,7 +247,7 @@ static void bdrv_co_drain_bh_cb(void *opaque)
 
     bdrv_dec_in_flight(bs);
     if (data->begin) {
-        bdrv_do_drained_begin(bs, data->recursive, data->parent);
+        bdrv_do_drained_begin(bs, data->recursive, data->parent, data->poll);
     } else {
         bdrv_do_drained_end(bs, data->recursive, data->parent);
     }
@@ -242,7 +258,7 @@ static void bdrv_co_drain_bh_cb(void *opaque)
 
 static void coroutine_fn bdrv_co_yield_to_drain(BlockDriverState *bs,
                                                 bool begin, bool recursive,
-                                                BdrvChild *parent)
+                                                BdrvChild *parent, bool poll)
 {
     BdrvCoDrainData data;
 
@@ -257,6 +273,7 @@ static void coroutine_fn bdrv_co_yield_to_drain(BlockDriverState *bs,
         .begin = begin,
         .recursive = recursive,
         .parent = parent,
+        .poll = poll,
     };
     bdrv_inc_in_flight(bs);
     aio_bh_schedule_oneshot(bdrv_get_aio_context(bs),
@@ -269,12 +286,12 @@ static void coroutine_fn bdrv_co_yield_to_drain(BlockDriverState *bs,
 }
 
 void bdrv_do_drained_begin(BlockDriverState *bs, bool recursive,
-                           BdrvChild *parent)
+                           BdrvChild *parent, bool poll)
 {
     BdrvChild *child, *next;
 
     if (qemu_in_coroutine()) {
-        bdrv_co_yield_to_drain(bs, true, recursive, parent);
+        bdrv_co_yield_to_drain(bs, true, recursive, parent, poll);
         return;
     }
 
@@ -286,25 +303,35 @@ void bdrv_do_drained_begin(BlockDriverState *bs, bool recursive,
     bdrv_parent_drained_begin(bs, parent);
     bdrv_drain_invoke(bs, true);
 
-    /* Wait for drained requests to finish */
-    BDRV_POLL_WHILE(bs, bdrv_drain_poll_top_level(bs, parent));
-
     if (recursive) {
         bs->recursive_quiesce_counter++;
         QLIST_FOREACH_SAFE(child, &bs->children, next, next) {
-            bdrv_do_drained_begin(child->bs, true, child);
+            bdrv_do_drained_begin(child->bs, true, child, false);
         }
     }
+
+    /*
+     * Wait for drained requests to finish.
+     *
+     * Calling BDRV_POLL_WHILE() only once for the top-level node is okay: The
+     * call is needed so things in this AioContext can make progress even
+     * though we don't return to the main AioContext loop - this automatically
+     * includes other nodes in the same AioContext and therefore all child
+     * nodes.
+     */
+    if (poll) {
+        BDRV_POLL_WHILE(bs, bdrv_drain_poll_top_level(bs, recursive, parent));
+    }
 }
 
 void bdrv_drained_begin(BlockDriverState *bs)
 {
-    bdrv_do_drained_begin(bs, false, NULL);
+    bdrv_do_drained_begin(bs, false, NULL, true);
 }
 
 void bdrv_subtree_drained_begin(BlockDriverState *bs)
 {
-    bdrv_do_drained_begin(bs, true, NULL);
+    bdrv_do_drained_begin(bs, true, NULL, true);
 }
 
 void bdrv_do_drained_end(BlockDriverState *bs, bool recursive,
@@ -314,7 +341,7 @@ void bdrv_do_drained_end(BlockDriverState *bs, bool recursive,
     int old_quiesce_counter;
 
     if (qemu_in_coroutine()) {
-        bdrv_co_yield_to_drain(bs, false, recursive, parent);
+        bdrv_co_yield_to_drain(bs, false, recursive, parent, false);
         return;
     }
     assert(bs->quiesce_counter > 0);
@@ -350,7 +377,7 @@ void bdrv_apply_subtree_drain(BdrvChild *child, BlockDriverState *new_parent)
     int i;
 
     for (i = 0; i < new_parent->recursive_quiesce_counter; i++) {
-        bdrv_do_drained_begin(child->bs, true, child);
+        bdrv_do_drained_begin(child->bs, true, child, true);
     }
 }
 
@@ -420,7 +447,7 @@ void bdrv_drain_all_begin(void)
         AioContext *aio_context = bdrv_get_aio_context(bs);
 
         aio_context_acquire(aio_context);
-        bdrv_do_drained_begin(bs, true, NULL);
+        bdrv_do_drained_begin(bs, true, NULL, true);
         aio_context_release(aio_context);
     }
 
diff --git a/include/block/block.h b/include/block/block.h
index 8c91d4c..2bbea7c 100644
--- a/include/block/block.h
+++ b/include/block/block.h
@@ -598,10 +598,13 @@ void bdrv_parent_drained_end(BlockDriverState *bs, BdrvChild *ignore);
 /**
  * bdrv_drain_poll:
  *
- * Poll for pending requests in @bs and its parents (except for
- * @ignore_parent). This is part of bdrv_drained_begin.
+ * Poll for pending requests in @bs, its parents (except for @ignore_parent),
+ * and if @recursive is true its children as well.
+ *
+ * This is part of bdrv_drained_begin.
  */
-bool bdrv_drain_poll(BlockDriverState *bs, BdrvChild *ignore_parent);
+bool bdrv_drain_poll(BlockDriverState *bs, bool recursive,
+                     BdrvChild *ignore_parent);
 
 /**
  * bdrv_drained_begin:
-- 
1.8.3.1