From: Kevin Wolf <kwolf@redhat.com>
To: qemu-block@nongnu.org
Cc: kwolf@redhat.com, qemu-devel@nongnu.org
Subject: [PULL 09/24] block: move drain outside of bdrv_try_change_aio_context()
Date: Wed, 4 Jun 2025 19:55:58 +0200 [thread overview]
Message-ID: <20250604175613.344113-10-kwolf@redhat.com> (raw)
In-Reply-To: <20250604175613.344113-1-kwolf@redhat.com>
From: Fiona Ebner <f.ebner@proxmox.com>
This is part of resolving the deadlock mentioned in commit "block:
move draining out of bdrv_change_aio_context() and mark GRAPH_RDLOCK".
Convert the function to a _locked() version that has to be called with
the graph lock held and add a convenience wrapper that has to be
called with the graph unlocked, which drains and takes the lock
itself. Since bdrv_try_change_aio_context() is global state code, the
wrapper is too.
Callers are adapted to use the appropriate variant, depending on
whether the caller already holds the lock. In the
test_set_aio_context() unit test, prior drains can be removed, because
draining already happens inside the new wrapper.
Note that bdrv_attach_child_common_abort(), bdrv_attach_child_common()
and bdrv_root_unref_child() hold the graph lock and are not actually
allowed to drain either. This will be addressed in the following
commits.
Functions like qmp_blockdev_mirror() query the nodes to act on before
draining and locking. In theory, draining could invalidate those nodes.
This kind of issue is not addressed by these commits.
Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-ID: <20250530151125.955508-10-f.ebner@proxmox.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
---
include/block/block-global-state.h | 8 +++--
block.c | 58 ++++++++++++++++++++++--------
blockdev.c | 15 +++++---
tests/unit/test-bdrv-drain.c | 4 ---
4 files changed, 59 insertions(+), 26 deletions(-)
diff --git a/include/block/block-global-state.h b/include/block/block-global-state.h
index aad160956a..91f249b5ad 100644
--- a/include/block/block-global-state.h
+++ b/include/block/block-global-state.h
@@ -278,8 +278,12 @@ bool GRAPH_RDLOCK
bdrv_child_change_aio_context(BdrvChild *c, AioContext *ctx,
GHashTable *visited, Transaction *tran,
Error **errp);
-int bdrv_try_change_aio_context(BlockDriverState *bs, AioContext *ctx,
- BdrvChild *ignore_child, Error **errp);
+int GRAPH_UNLOCKED
+bdrv_try_change_aio_context(BlockDriverState *bs, AioContext *ctx,
+ BdrvChild *ignore_child, Error **errp);
+int GRAPH_RDLOCK
+bdrv_try_change_aio_context_locked(BlockDriverState *bs, AioContext *ctx,
+ BdrvChild *ignore_child, Error **errp);
int GRAPH_RDLOCK bdrv_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz);
int bdrv_probe_geometry(BlockDriverState *bs, HDGeometry *geo);
diff --git a/block.c b/block.c
index 6f42c0f1ab..3aaacabf7f 100644
--- a/block.c
+++ b/block.c
@@ -3028,7 +3028,10 @@ static void GRAPH_WRLOCK bdrv_attach_child_common_abort(void *opaque)
bdrv_replace_child_noperm(s->child, NULL);
if (bdrv_get_aio_context(bs) != s->old_child_ctx) {
- bdrv_try_change_aio_context(bs, s->old_child_ctx, NULL, &error_abort);
+ bdrv_drain_all_begin();
+ bdrv_try_change_aio_context_locked(bs, s->old_child_ctx, NULL,
+ &error_abort);
+ bdrv_drain_all_end();
}
if (bdrv_child_get_parent_aio_context(s->child) != s->old_parent_ctx) {
@@ -3115,8 +3118,10 @@ bdrv_attach_child_common(BlockDriverState *child_bs,
parent_ctx = bdrv_child_get_parent_aio_context(new_child);
if (child_ctx != parent_ctx) {
Error *local_err = NULL;
- int ret = bdrv_try_change_aio_context(child_bs, parent_ctx, NULL,
- &local_err);
+ bdrv_drain_all_begin();
+ int ret = bdrv_try_change_aio_context_locked(child_bs, parent_ctx, NULL,
+ &local_err);
+ bdrv_drain_all_end();
if (ret < 0 && child_class->change_aio_ctx) {
Transaction *aio_ctx_tran = tran_new();
@@ -3319,8 +3324,10 @@ void bdrv_root_unref_child(BdrvChild *child)
* When the parent requiring a non-default AioContext is removed, the
* node moves back to the main AioContext
*/
- bdrv_try_change_aio_context(child_bs, qemu_get_aio_context(), NULL,
- NULL);
+ bdrv_drain_all_begin();
+ bdrv_try_change_aio_context_locked(child_bs, qemu_get_aio_context(),
+ NULL, NULL);
+ bdrv_drain_all_end();
}
bdrv_schedule_unref(child_bs);
@@ -7719,9 +7726,13 @@ bdrv_change_aio_context(BlockDriverState *bs, AioContext *ctx,
*
* If ignore_child is not NULL, that child (and its subgraph) will not
* be touched.
+ *
+ * Called with the graph lock held.
+ *
+ * Called while all bs are drained.
*/
-int bdrv_try_change_aio_context(BlockDriverState *bs, AioContext *ctx,
- BdrvChild *ignore_child, Error **errp)
+int bdrv_try_change_aio_context_locked(BlockDriverState *bs, AioContext *ctx,
+ BdrvChild *ignore_child, Error **errp)
{
Transaction *tran;
GHashTable *visited;
@@ -7730,17 +7741,15 @@ int bdrv_try_change_aio_context(BlockDriverState *bs, AioContext *ctx,
/*
* Recursion phase: go through all nodes of the graph.
- * Take care of checking that all nodes support changing AioContext
- * and drain them, building a linear list of callbacks to run if everything
- * is successful (the transaction itself).
+ * Take care of checking that all nodes support changing AioContext,
+ * building a linear list of callbacks to run if everything is successful
+ * (the transaction itself).
*/
tran = tran_new();
visited = g_hash_table_new(NULL, NULL);
if (ignore_child) {
g_hash_table_add(visited, ignore_child);
}
- bdrv_drain_all_begin();
- bdrv_graph_rdlock_main_loop();
ret = bdrv_change_aio_context(bs, ctx, visited, tran, errp);
g_hash_table_destroy(visited);
@@ -7754,15 +7763,34 @@ int bdrv_try_change_aio_context(BlockDriverState *bs, AioContext *ctx,
if (!ret) {
/* Just run clean() callbacks. No AioContext changed. */
tran_abort(tran);
- bdrv_graph_rdunlock_main_loop();
- bdrv_drain_all_end();
return -EPERM;
}
tran_commit(tran);
+ return 0;
+}
+
+/*
+ * Change bs's and recursively all of its parents' and children's AioContext
+ * to the given new context, returning an error if that isn't possible.
+ *
+ * If ignore_child is not NULL, that child (and its subgraph) will not
+ * be touched.
+ */
+int bdrv_try_change_aio_context(BlockDriverState *bs, AioContext *ctx,
+ BdrvChild *ignore_child, Error **errp)
+{
+ int ret;
+
+ GLOBAL_STATE_CODE();
+
+ bdrv_drain_all_begin();
+ bdrv_graph_rdlock_main_loop();
+ ret = bdrv_try_change_aio_context_locked(bs, ctx, ignore_child, errp);
bdrv_graph_rdunlock_main_loop();
bdrv_drain_all_end();
- return 0;
+
+ return ret;
}
void bdrv_add_aio_context_notifier(BlockDriverState *bs,
diff --git a/blockdev.c b/blockdev.c
index 3982f9776b..750beba41f 100644
--- a/blockdev.c
+++ b/blockdev.c
@@ -3601,12 +3601,13 @@ void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread,
AioContext *new_context;
BlockDriverState *bs;
- GRAPH_RDLOCK_GUARD_MAINLOOP();
+ bdrv_drain_all_begin();
+ bdrv_graph_rdlock_main_loop();
bs = bdrv_find_node(node_name);
if (!bs) {
error_setg(errp, "Failed to find node with node-name='%s'", node_name);
- return;
+ goto out;
}
/* Protects against accidents. */
@@ -3614,14 +3615,14 @@ void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread,
error_setg(errp, "Node %s is associated with a BlockBackend and could "
"be in use (use force=true to override this check)",
node_name);
- return;
+ goto out;
}
if (iothread->type == QTYPE_QSTRING) {
IOThread *obj = iothread_by_id(iothread->u.s);
if (!obj) {
error_setg(errp, "Cannot find iothread %s", iothread->u.s);
- return;
+ goto out;
}
new_context = iothread_get_aio_context(obj);
@@ -3629,7 +3630,11 @@ void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread,
new_context = qemu_get_aio_context();
}
- bdrv_try_change_aio_context(bs, new_context, NULL, errp);
+ bdrv_try_change_aio_context_locked(bs, new_context, NULL, errp);
+
+out:
+ bdrv_graph_rdunlock_main_loop();
+ bdrv_drain_all_end();
}
QemuOptsList qemu_common_drive_opts = {
diff --git a/tests/unit/test-bdrv-drain.c b/tests/unit/test-bdrv-drain.c
index 290cd2a70e..3185f3f429 100644
--- a/tests/unit/test-bdrv-drain.c
+++ b/tests/unit/test-bdrv-drain.c
@@ -1396,14 +1396,10 @@ static void test_set_aio_context(void)
bs = bdrv_new_open_driver(&bdrv_test, "test-node", BDRV_O_RDWR,
&error_abort);
- bdrv_drained_begin(bs);
bdrv_try_change_aio_context(bs, ctx_a, NULL, &error_abort);
- bdrv_drained_end(bs);
- bdrv_drained_begin(bs);
bdrv_try_change_aio_context(bs, ctx_b, NULL, &error_abort);
bdrv_try_change_aio_context(bs, qemu_get_aio_context(), NULL, &error_abort);
- bdrv_drained_end(bs);
bdrv_unref(bs);
iothread_join(a);
--
2.49.0
next prev parent reply other threads:[~2025-06-04 17:58 UTC|newest]
Thread overview: 26+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-06-04 17:55 [PULL 00/24] Block layer patches Kevin Wolf
2025-06-04 17:55 ` [PULL 01/24] block: remove outdated comments about AioContext locking Kevin Wolf
2025-06-04 17:55 ` [PULL 02/24] block: move drain outside of read-locked bdrv_reopen_queue_child() Kevin Wolf
2025-06-04 17:55 ` [PULL 03/24] block/snapshot: move drain outside of read-locked bdrv_snapshot_delete() Kevin Wolf
2025-06-04 17:55 ` [PULL 04/24] block: move drain outside of read-locked bdrv_inactivate_recurse() Kevin Wolf
2025-06-04 17:55 ` [PULL 05/24] block: mark bdrv_parent_change_aio_context() GRAPH_RDLOCK Kevin Wolf
2025-06-04 17:55 ` [PULL 06/24] block: mark change_aio_ctx() callback and instances as GRAPH_RDLOCK(_PTR) Kevin Wolf
2025-06-04 17:55 ` [PULL 07/24] block: mark bdrv_child_change_aio_context() GRAPH_RDLOCK Kevin Wolf
2025-06-04 17:55 ` [PULL 08/24] block: move drain outside of bdrv_change_aio_context() and mark GRAPH_RDLOCK Kevin Wolf
2025-06-04 17:55 ` Kevin Wolf [this message]
2025-06-04 17:55 ` [PULL 10/24] block: move drain outside of bdrv_attach_child_common(_abort)() Kevin Wolf
2025-06-04 17:56 ` [PULL 11/24] block: move drain outside of bdrv_set_backing_hd_drained() Kevin Wolf
2025-06-04 17:56 ` [PULL 12/24] block: move drain outside of bdrv_root_attach_child() Kevin Wolf
2025-06-04 17:56 ` [PULL 13/24] block: move drain outside of bdrv_attach_child() Kevin Wolf via
2025-06-04 17:56 ` [PULL 14/24] block: move drain outside of quorum_add_child() Kevin Wolf
2025-06-04 17:56 ` [PULL 15/24] block: move drain outside of bdrv_root_unref_child() Kevin Wolf
2025-06-04 17:56 ` [PULL 16/24] block: move drain outside of quorum_del_child() Kevin Wolf
2025-06-04 17:56 ` [PULL 17/24] blockdev: drain while unlocked in internal_snapshot_action() Kevin Wolf
2025-06-04 17:56 ` [PULL 18/24] blockdev: drain while unlocked in external_snapshot_action() Kevin Wolf
2025-06-04 17:56 ` [PULL 19/24] block: mark bdrv_drained_begin() and friends as GRAPH_UNLOCKED Kevin Wolf
2025-06-04 17:56 ` [PULL 20/24] iotests/graph-changes-while-io: remove image file after test Kevin Wolf
2025-06-04 17:56 ` [PULL 21/24] iotests/graph-changes-while-io: add test case with removal of lower snapshot Kevin Wolf
2025-06-04 17:56 ` [PULL 22/24] block/io: remove duplicate GLOBAL_STATE_CODE() in bdrv_do_drained_end() Kevin Wolf
2025-06-04 17:56 ` [PULL 23/24] iotests: fix 240 Kevin Wolf
2025-06-04 17:56 ` [PULL 24/24] hw/core/qdev-properties-system: Add missing return in set_drive_helper() Kevin Wolf
2025-06-05 19:00 ` [PULL 00/24] Block layer patches Stefan Hajnoczi
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20250604175613.344113-10-kwolf@redhat.com \
--to=kwolf@redhat.com \
--cc=qemu-block@nongnu.org \
--cc=qemu-devel@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).