Skip to contentSkip to conversation

block: AioContext management, part 2

PATCH Applied Kevin Wolf 3 participants 22 messages May 23, 2019 – Jun 3, 2019 Block layer

lore
Share & tools
lore

Kevin Wolf

author
#0 lore raw

cover block: AioContext management, part 2

Recently, a few bugs were reported that resulted from an inconsistent state regarding AioContexts. Block nodes can end up in different contexts than their users expect - the AioContext of a node can even change under the feet of a device with no way for the device to forbid this. We recently added a few basic checks to scsi-disk and virtio-blk, but they are by far not enough.

Part 1 solved the first half of the problem and made sure that AioContext changes are propagated through the graph as necessary, so that a bdrv_set_aio_context() correctly pulls everything that uses the node into the right context.

This is part 2 that fixes the second half: Attaching new child nodes where the parent and child are already in different AioContexts. This operation may only succeed if we can move one of the node into the context of the other node.

After applying this series, unchecked bdrv_set_aio_context() doesn't exist any more and all users call functions that make sure that the AioContext assignments across the graph stays consistent.

Kevin Wolf (15):
test-block-iothread: Check filter node in test_propagate_mirror
block: Add Error to blk_set_aio_context()
block: Add BlockBackend.ctx
block: Add qdev_prop_drive_iothread property type
scsi-disk: Use qdev_prop_drive_iothread
block: Adjust AioContexts when attaching nodes
test-block-iothread: Test adding parent to iothread node
test-block-iothread: BlockBackend AioContext across root node change
block: Move node without parents to main AioContext
blockdev: Use bdrv_try_set_aio_context() for monitor commands
block: Remove wrong bdrv_set_aio_context() calls
virtio-scsi-test: Test attaching new overlay with iothreads
iotests: Attach new devices to node in non-default iothread
test-bdrv-drain: Use bdrv_try_set_aio_context()
block: Remove bdrv_set_aio_context()

 docs/devel/multiple-iothreads.txt |   4 +-
 include/block/block.h             |   9 ---
 include/block/block_int.h         |   1 +
 include/hw/block/block.h          |   7 +-
 include/hw/qdev-properties.h      |   3 +
 include/hw/scsi/scsi.h            |   1 +
 include/sysemu/block-backend.h    |   5 +-
 tests/libqtest.h                  |  11 ++++
 block.c                           |  70 ++++++++++++++------
 block/backup.c                    |   3 +-
 block/block-backend.c             |  47 +++++++++-----
 block/commit.c                    |  13 ++--
 block/crypto.c                    |   3 +-
 block/mirror.c                    |   4 +-
 block/parallels.c                 |   3 +-
 block/qcow.c                      |   3 +-
 block/qcow2.c                     |   6 +-
 block/qed.c                       |   3 +-
 block/sheepdog.c                  |   3 +-
 block/vdi.c                       |   3 +-
 block/vhdx.c                      |   3 +-
 block/vmdk.c                      |   3 +-
 block/vpc.c                       |   3 +-
 blockdev.c                        |  48 ++++++++------
 blockjob.c                        |  12 +++-
 hmp.c                             |   3 +-
 hw/block/dataplane/virtio-blk.c   |  12 +++-
 hw/block/dataplane/xen-block.c    |   6 +-
 hw/block/fdc.c                    |   2 +-
 hw/block/xen-block.c              |   2 +-
 hw/core/qdev-properties-system.c  |  41 +++++++++++-
 hw/ide/qdev.c                     |   2 +-
 hw/scsi/scsi-disk.c               |  24 ++++---
 hw/scsi/virtio-scsi.c             |  25 ++++---
 migration/block.c                 |   3 +-
 nbd/server.c                      |   5 +-
 tests/libqtest.c                  |  19 ++++++
 tests/test-bdrv-drain.c           |  50 +++++++-------
 tests/test-bdrv-graph-mod.c       |   5 +-
 tests/test-block-backend.c        |   4 +-
 tests/test-block-iothread.c       | 104 +++++++++++++++++++++++++-----
 tests/test-blockjob.c             |   2 +-
 tests/test-throttle.c             |   6 +-
 tests/virtio-scsi-test.c          |  62 ++++++++++++++++++
 tests/qemu-iotests/051            |  24 +++++++
 tests/qemu-iotests/051.out        |   3 +
 tests/qemu-iotests/051.pc.out     |  27 ++++++++
 tests/qemu-iotests/240.out        |   2 +-
 48 files changed, 531 insertions(+), 173 deletions(-)
Signature · 1 line
2.20.1
author
#1 lore raw

PATCH 1/15 test-block-iothread: Check filter node in test_propagate_mirror

Just make the test cover the AioContext of the filter node as well.

---
 tests/test-block-iothread.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
tests/test-block-iothread.c +6 -1
index 59f692892e..e424d360c8 100644
@@ -593,7 +593,7 @@ static void test_propagate_mirror(void)
     IOThread *iothread = iothread_new();
     AioContext *ctx = iothread_get_aio_context(iothread);
     AioContext *main_ctx = qemu_get_aio_context();
-    BlockDriverState *src, *target;
+    BlockDriverState *src, *target, *filter;
     BlockBackend *blk;
     Job *job;
     Error *local_err = NULL;
@@ -610,11 +610,13 @@ static void test_propagate_mirror(void)
                  false, "filter_node", MIRROR_COPY_MODE_BACKGROUND,
                  &error_abort);
     job = job_get("job0");
+    filter = bdrv_find_node("filter_node");
 
     /* Change the AioContext of src */
     bdrv_try_set_aio_context(src, ctx, &error_abort);
     g_assert(bdrv_get_aio_context(src) == ctx);
     g_assert(bdrv_get_aio_context(target) == ctx);
+    g_assert(bdrv_get_aio_context(filter) == ctx);
     g_assert(job->aio_context == ctx);
 
     /* Change the AioContext of target */
@@ -623,6 +625,7 @@ static void test_propagate_mirror(void)
     aio_context_release(ctx);
     g_assert(bdrv_get_aio_context(src) == main_ctx);
     g_assert(bdrv_get_aio_context(target) == main_ctx);
+    g_assert(bdrv_get_aio_context(filter) == main_ctx);
 
     /* With a BlockBackend on src, changing target must fail */
     blk = blk_new(0, BLK_PERM_ALL);
@@ -635,6 +638,7 @@ static void test_propagate_mirror(void)
     g_assert(blk_get_aio_context(blk) == main_ctx);
     g_assert(bdrv_get_aio_context(src) == main_ctx);
     g_assert(bdrv_get_aio_context(target) == main_ctx);
+    g_assert(bdrv_get_aio_context(filter) == main_ctx);
 
     /* ...unless we explicitly allow it */
     aio_context_acquire(ctx);
@@ -645,6 +649,7 @@ static void test_propagate_mirror(void)
     g_assert(blk_get_aio_context(blk) == ctx);
     g_assert(bdrv_get_aio_context(src) == ctx);
     g_assert(bdrv_get_aio_context(target) == ctx);
+    g_assert(bdrv_get_aio_context(filter) == ctx);
 
     job_cancel_sync_all();
 
Signature · 1 line
2.20.1
Reviewed-by on v1 1/15
#2 lore raw

Re: [Qemu-devel] [PATCH 01/15] test-block-iothread: Check filter node in test_propagate_mirror

Quoted 7 lines · Kevin Wolf

Just make the test cover the AioContext of the filter node as well.

---
 tests/test-block-iothread.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
Signature · 3 lines
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org
author
#3 lore raw

PATCH 2/15 block: Add Error to blk_set_aio_context()

In reply to Kevin Wolf

Add an Error parameter to blk_set_aio_context() and use bdrv_child_try_set_aio_context() internally to check whether all involved nodes can actually support the AioContext switch.

---
 include/sysemu/block-backend.h  |  3 ++-
 block/block-backend.c           | 26 ++++++++++++++++----------
 hw/block/dataplane/virtio-blk.c | 12 +++++++++---
 hw/block/dataplane/xen-block.c  |  6 ++++--
 hw/scsi/virtio-scsi.c           | 10 +++++++---
 tests/test-bdrv-drain.c         |  8 ++++----
 tests/test-block-iothread.c     | 22 +++++++++++-----------
 7 files changed, 53 insertions(+), 34 deletions(-)
include/sysemu/block-backend.h +2 -1
index 938de34fe9..228fb3fb83 100644
@@ -208,7 +208,8 @@ void blk_op_unblock(BlockBackend *blk, BlockOpType op, Error *reason);
 void blk_op_block_all(BlockBackend *blk, Error *reason);
 void blk_op_unblock_all(BlockBackend *blk, Error *reason);
 AioContext *blk_get_aio_context(BlockBackend *blk);
-void blk_set_aio_context(BlockBackend *blk, AioContext *new_context);
+int blk_set_aio_context(BlockBackend *blk, AioContext *new_context,
+                        Error **errp);
 void blk_add_aio_context_notifier(BlockBackend *blk,
         void (*attached_aio_context)(AioContext *new_context, void *opaque),
         void (*detach_aio_context)(void *opaque), void *opaque);
block/block-backend.c +16 -10
index 4c0a8ef88d..dfb6bf017a 100644
@@ -1866,30 +1866,36 @@ static AioContext *blk_aiocb_get_aio_context(BlockAIOCB *acb)
     return blk_get_aio_context(blk_acb->blk);
 }
 
-static void blk_do_set_aio_context(BlockBackend *blk, AioContext *new_context,
-                                   bool update_root_node)
+static int blk_do_set_aio_context(BlockBackend *blk, AioContext *new_context,
+                                  bool update_root_node, Error **errp)
 {
     BlockDriverState *bs = blk_bs(blk);
     ThrottleGroupMember *tgm = &blk->public.throttle_group_member;
+    int ret;
 
     if (bs) {
+        if (update_root_node) {
+            ret = bdrv_child_try_set_aio_context(bs, new_context, blk->root,
+                                                 errp);
+            if (ret < 0) {
+                return ret;
+            }
+        }
         if (tgm->throttle_state) {
             bdrv_drained_begin(bs);
             throttle_group_detach_aio_context(tgm);
             throttle_group_attach_aio_context(tgm, new_context);
             bdrv_drained_end(bs);
         }
-        if (update_root_node) {
-            GSList *ignore = g_slist_prepend(NULL, blk->root);
-            bdrv_set_aio_context_ignore(bs, new_context, &ignore);
-            g_slist_free(ignore);
-        }
     }
+
+    return 0;
 }
 
-void blk_set_aio_context(BlockBackend *blk, AioContext *new_context)
+int blk_set_aio_context(BlockBackend *blk, AioContext *new_context,
+                        Error **errp)
 {
-    blk_do_set_aio_context(blk, new_context, true);
+    return blk_do_set_aio_context(blk, new_context, true, errp);
 }
 
 static bool blk_root_can_set_aio_ctx(BdrvChild *child, AioContext *ctx,
@@ -1916,7 +1922,7 @@ static void blk_root_set_aio_ctx(BdrvChild *child, AioContext *ctx,
                                  GSList **ignore)
 {
     BlockBackend *blk = child->opaque;
-    blk_do_set_aio_context(blk, ctx, false);
+    blk_do_set_aio_context(blk, ctx, false, &error_abort);
 }
 
 void blk_add_aio_context_notifier(BlockBackend *blk,
hw/block/dataplane/virtio-blk.c +9 -3
index 8c37bd314a..158c78f852 100644
@@ -173,6 +173,7 @@ int virtio_blk_data_plane_start(VirtIODevice *vdev)
     VirtioBusClass *k = VIRTIO_BUS_GET_CLASS(qbus);
     unsigned i;
     unsigned nvqs = s->conf->num_queues;
+    Error *local_err = NULL;
     int r;
 
     if (vblk->dataplane_started || s->starting) {
@@ -212,7 +213,11 @@ int virtio_blk_data_plane_start(VirtIODevice *vdev)
     vblk->dataplane_started = true;
     trace_virtio_blk_data_plane_start(s);
 
-    blk_set_aio_context(s->conf->conf.blk, s->ctx);
+    r = blk_set_aio_context(s->conf->conf.blk, s->ctx, &local_err);
+    if (r < 0) {
+        error_report_err(local_err);
+        goto fail_guest_notifiers;
+    }
 
     /* Kick right away to begin processing requests already in vring */
     for (i = 0; i < nvqs; i++) {
@@ -281,8 +286,9 @@ void virtio_blk_data_plane_stop(VirtIODevice *vdev)
     aio_context_acquire(s->ctx);
     aio_wait_bh_oneshot(s->ctx, virtio_blk_data_plane_stop_bh, s);
 
-    /* Drain and switch bs back to the QEMU main loop */
-    blk_set_aio_context(s->conf->conf.blk, qemu_get_aio_context());
+    /* Drain and try to switch bs back to the QEMU main loop. If other users
+     * keep the BlockBackend in the iothread, that's ok */
+    blk_set_aio_context(s->conf->conf.blk, qemu_get_aio_context(), NULL);
 
     aio_context_release(s->ctx);
 
hw/block/dataplane/xen-block.c +4 -2
index bb8f1186e4..f7ad452bbd 100644
@@ -682,7 +682,8 @@ void xen_block_dataplane_stop(XenBlockDataPlane *dataplane)
     }
 
     aio_context_acquire(dataplane->ctx);
-    blk_set_aio_context(dataplane->blk, qemu_get_aio_context());
+    /* Xen doesn't have multiple users for nodes, so this can't fail */
+    blk_set_aio_context(dataplane->blk, qemu_get_aio_context(), &error_abort);
     aio_context_release(dataplane->ctx);
 
     xendev = dataplane->xendev;
@@ -811,7 +812,8 @@ void xen_block_dataplane_start(XenBlockDataPlane *dataplane,
     }
 
     aio_context_acquire(dataplane->ctx);
-    blk_set_aio_context(dataplane->blk, dataplane->ctx);
+    /* If other users keep the BlockBackend in the iothread, that's ok */
+    blk_set_aio_context(dataplane->blk, dataplane->ctx, NULL);
     aio_context_release(dataplane->ctx);
     return;
 
hw/scsi/virtio-scsi.c +7 -3
index 839f120256..01c2b85f90 100644
@@ -795,6 +795,7 @@ static void virtio_scsi_hotplug(HotplugHandler *hotplug_dev, DeviceState *dev,
     VirtIODevice *vdev = VIRTIO_DEVICE(hotplug_dev);
     VirtIOSCSI *s = VIRTIO_SCSI(vdev);
     SCSIDevice *sd = SCSI_DEVICE(dev);
+    int ret;
 
     if (s->ctx && !s->dataplane_fenced) {
         AioContext *ctx;
@@ -808,9 +809,11 @@ static void virtio_scsi_hotplug(HotplugHandler *hotplug_dev, DeviceState *dev,
             return;
         }
         virtio_scsi_acquire(s);
-        blk_set_aio_context(sd->conf.blk, s->ctx);
+        ret = blk_set_aio_context(sd->conf.blk, s->ctx, errp);
         virtio_scsi_release(s);
-
+        if (ret < 0) {
+            return;
+        }
     }
 
     if (virtio_vdev_has_feature(vdev, VIRTIO_SCSI_F_HOTPLUG)) {
@@ -839,7 +842,8 @@ static void virtio_scsi_hotunplug(HotplugHandler *hotplug_dev, DeviceState *dev,
 
     if (s->ctx) {
         virtio_scsi_acquire(s);
-        blk_set_aio_context(sd->conf.blk, qemu_get_aio_context());
+        /* If other users keep the BlockBackend in the iothread, that's ok */
+        blk_set_aio_context(sd->conf.blk, qemu_get_aio_context(), NULL);
         virtio_scsi_release(s);
     }
 
tests/test-bdrv-drain.c +4 -4
index eda90750eb..6d5565e8e7 100644
@@ -678,7 +678,7 @@ static void test_iothread_common(enum drain_type drain_type, int drain_thread)
     s = bs->opaque;
     blk_insert_bs(blk, bs, &error_abort);
 
-    blk_set_aio_context(blk, ctx_a);
+    blk_set_aio_context(blk, ctx_a, &error_abort);
     aio_context_acquire(ctx_a);
 
     s->bh_indirection_ctx = ctx_b;
@@ -742,7 +742,7 @@ static void test_iothread_common(enum drain_type drain_type, int drain_thread)
     }
 
     aio_context_acquire(ctx_a);
-    blk_set_aio_context(blk, qemu_get_aio_context());
+    blk_set_aio_context(blk, qemu_get_aio_context(), &error_abort);
     aio_context_release(ctx_a);
 
     bdrv_unref(bs);
@@ -903,7 +903,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type,
     if (use_iothread) {
         iothread = iothread_new();
         ctx = iothread_get_aio_context(iothread);
-        blk_set_aio_context(blk_src, ctx);
+        blk_set_aio_context(blk_src, ctx, &error_abort);
     } else {
         ctx = qemu_get_aio_context();
     }
@@ -1001,7 +1001,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type,
     g_assert_cmpint(ret, ==, (result == TEST_JOB_SUCCESS ? 0 : -EIO));
 
     if (use_iothread) {
-        blk_set_aio_context(blk_src, qemu_get_aio_context());
+        blk_set_aio_context(blk_src, qemu_get_aio_context(), &error_abort);
     }
     aio_context_release(ctx);
 
tests/test-block-iothread.c +11 -11
index e424d360c8..1d47ea9895 100644
@@ -342,14 +342,14 @@ static void test_sync_op(const void *opaque)
     blk_insert_bs(blk, bs, &error_abort);
     c = QLIST_FIRST(&bs->parents);
 
-    blk_set_aio_context(blk, ctx);
+    blk_set_aio_context(blk, ctx, &error_abort);
     aio_context_acquire(ctx);
     t->fn(c);
     if (t->blkfn) {
         t->blkfn(blk);
     }
     aio_context_release(ctx);
-    blk_set_aio_context(blk, qemu_get_aio_context());
+    blk_set_aio_context(blk, qemu_get_aio_context(), &error_abort);
 
     bdrv_unref(bs);
     blk_unref(blk);
@@ -428,7 +428,7 @@ static void test_attach_blockjob(void)
         aio_poll(qemu_get_aio_context(), false);
     }
 
-    blk_set_aio_context(blk, ctx);
+    blk_set_aio_context(blk, ctx, &error_abort);
 
     tjob->n = 0;
     while (tjob->n == 0) {
@@ -436,7 +436,7 @@ static void test_attach_blockjob(void)
     }
 
     aio_context_acquire(ctx);
-    blk_set_aio_context(blk, qemu_get_aio_context());
+    blk_set_aio_context(blk, qemu_get_aio_context(), &error_abort);
     aio_context_release(ctx);
 
     tjob->n = 0;
@@ -444,7 +444,7 @@ static void test_attach_blockjob(void)
         aio_poll(qemu_get_aio_context(), false);
     }
 
-    blk_set_aio_context(blk, ctx);
+    blk_set_aio_context(blk, ctx, &error_abort);
 
     tjob->n = 0;
     while (tjob->n == 0) {
@@ -453,7 +453,7 @@ static void test_attach_blockjob(void)
 
     aio_context_acquire(ctx);
     job_complete_sync(&tjob->common.job, &error_abort);
-    blk_set_aio_context(blk, qemu_get_aio_context());
+    blk_set_aio_context(blk, qemu_get_aio_context(), &error_abort);
     aio_context_release(ctx);
 
     bdrv_unref(bs);
@@ -497,7 +497,7 @@ static void test_propagate_basic(void)
     bs_verify = bdrv_open(NULL, NULL, options, BDRV_O_RDWR, &error_abort);
 
     /* Switch the AioContext */
-    blk_set_aio_context(blk, ctx);
+    blk_set_aio_context(blk, ctx, &error_abort);
     g_assert(blk_get_aio_context(blk) == ctx);
     g_assert(bdrv_get_aio_context(bs_a) == ctx);
     g_assert(bdrv_get_aio_context(bs_verify) == ctx);
@@ -505,7 +505,7 @@ static void test_propagate_basic(void)
 
     /* Switch the AioContext back */
     ctx = qemu_get_aio_context();
-    blk_set_aio_context(blk, ctx);
+    blk_set_aio_context(blk, ctx, &error_abort);
     g_assert(blk_get_aio_context(blk) == ctx);
     g_assert(bdrv_get_aio_context(bs_a) == ctx);
     g_assert(bdrv_get_aio_context(bs_verify) == ctx);
@@ -565,7 +565,7 @@ static void test_propagate_diamond(void)
     blk_insert_bs(blk, bs_verify, &error_abort);
 
     /* Switch the AioContext */
-    blk_set_aio_context(blk, ctx);
+    blk_set_aio_context(blk, ctx, &error_abort);
     g_assert(blk_get_aio_context(blk) == ctx);
     g_assert(bdrv_get_aio_context(bs_verify) == ctx);
     g_assert(bdrv_get_aio_context(bs_a) == ctx);
@@ -574,7 +574,7 @@ static void test_propagate_diamond(void)
 
     /* Switch the AioContext back */
     ctx = qemu_get_aio_context();
-    blk_set_aio_context(blk, ctx);
+    blk_set_aio_context(blk, ctx, &error_abort);
     g_assert(blk_get_aio_context(blk) == ctx);
     g_assert(bdrv_get_aio_context(bs_verify) == ctx);
     g_assert(bdrv_get_aio_context(bs_a) == ctx);
@@ -654,7 +654,7 @@ static void test_propagate_mirror(void)
     job_cancel_sync_all();
 
     aio_context_acquire(ctx);
-    blk_set_aio_context(blk, main_ctx);
+    blk_set_aio_context(blk, main_ctx, &error_abort);
     bdrv_try_set_aio_context(target, main_ctx, &error_abort);
     aio_context_release(ctx);
 
Signature · 1 line
2.20.1
author
#4 lore raw

PATCH 3/15 block: Add BlockBackend.ctx

In reply to Kevin Wolf

This adds a new parameter to blk_new() which requires its callers to declare from which AioContext this BlockBackend is going to be used (or the locks of which AioContext need to be taken anyway).

The given context is only stored and kept up to date when changing AioContexts. Actually applying the stored AioContext to the root node is saved for another commit.

---
 include/sysemu/block-backend.h   |  2 +-
 block.c                          |  2 +-
 block/backup.c                   |  3 ++-
 block/block-backend.c            | 18 +++++++++++++++---
 block/commit.c                   | 11 +++++++----
 block/crypto.c                   |  3 ++-
 block/mirror.c                   |  3 ++-
 block/parallels.c                |  3 ++-
 block/qcow.c                     |  3 ++-
 block/qcow2.c                    |  6 ++++--
 block/qed.c                      |  3 ++-
 block/sheepdog.c                 |  3 ++-
 block/vdi.c                      |  3 ++-
 block/vhdx.c                     |  3 ++-
 block/vmdk.c                     |  3 ++-
 block/vpc.c                      |  3 ++-
 blockdev.c                       |  4 ++--
 blockjob.c                       |  2 +-
 hmp.c                            |  3 ++-
 hw/block/fdc.c                   |  2 +-
 hw/block/xen-block.c             |  2 +-
 hw/core/qdev-properties-system.c |  4 +++-
 hw/ide/qdev.c                    |  2 +-
 hw/scsi/scsi-disk.c              |  2 +-
 migration/block.c                |  3 ++-
 nbd/server.c                     |  5 +++--
 tests/test-bdrv-drain.c          | 30 +++++++++++++++---------------
 tests/test-bdrv-graph-mod.c      |  5 +++--
 tests/test-block-backend.c       |  4 ++--
 tests/test-block-iothread.c      | 10 +++++-----
 tests/test-blockjob.c            |  2 +-
 tests/test-throttle.c            |  6 +++---
 32 files changed, 96 insertions(+), 62 deletions(-)
include/sysemu/block-backend.h +1 -1
index 228fb3fb83..733c4957eb 100644
@@ -76,7 +76,7 @@ typedef struct BlockBackendPublic {
     ThrottleGroupMember throttle_group_member;
 } BlockBackendPublic;
 
-BlockBackend *blk_new(uint64_t perm, uint64_t shared_perm);
+BlockBackend *blk_new(AioContext *ctx, uint64_t perm, uint64_t shared_perm);
 BlockBackend *blk_new_open(const char *filename, const char *reference,
                            QDict *options, int flags, Error **errp);
 int blk_get_refcnt(BlockBackend *blk);
block.c +1 -1
index 75f370dbba..301135b985 100644
@@ -2878,7 +2878,7 @@ static BlockDriverState *bdrv_open_inherit(const char *filename,
             /* Not requesting BLK_PERM_CONSISTENT_READ because we're only
              * looking at the header to guess the image format. This works even
              * in cases where a guest would not see a consistent state. */
-            file = blk_new(0, BLK_PERM_ALL);
+            file = blk_new(bdrv_get_aio_context(file_bs), 0, BLK_PERM_ALL);
             blk_insert_bs(file, file_bs, &local_err);
             bdrv_unref(file_bs);
             if (local_err) {
block/backup.c +2 -1
index 916817d8b1..d76888d09d 100644
@@ -640,7 +640,8 @@ BlockJob *backup_job_create(const char *job_id, BlockDriverState *bs,
     }
 
     /* The target must match the source in size, so no resize here either */
-    job->target = blk_new(BLK_PERM_WRITE,
+    job->target = blk_new(job->common.job.aio_context,
+                          BLK_PERM_WRITE,
                           BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE |
                           BLK_PERM_WRITE_UNCHANGED | BLK_PERM_GRAPH_MOD);
     ret = blk_insert_bs(job->target, target, errp);
block/block-backend.c +15 -3
index dfb6bf017a..8eca3261e3 100644
@@ -42,6 +42,7 @@ struct BlockBackend {
     char *name;
     int refcnt;
     BdrvChild *root;
+    AioContext *ctx;
     DriveInfo *legacy_dinfo;    /* null unless created by drive_new() */
     QTAILQ_ENTRY(BlockBackend) link;         /* for block_backends */
     QTAILQ_ENTRY(BlockBackend) monitor_link; /* for monitor_block_backends */
@@ -322,12 +323,13 @@ static const BdrvChildRole child_root = {
  *
  * Return the new BlockBackend on success, null on failure.
  */
-BlockBackend *blk_new(uint64_t perm, uint64_t shared_perm)
+BlockBackend *blk_new(AioContext *ctx, uint64_t perm, uint64_t shared_perm)
 {
     BlockBackend *blk;
 
     blk = g_new0(BlockBackend, 1);
     blk->refcnt = 1;
+    blk->ctx = ctx;
     blk->perm = perm;
     blk->shared_perm = shared_perm;
     blk_set_enable_write_cache(blk, true);
@@ -347,6 +349,7 @@ BlockBackend *blk_new(uint64_t perm, uint64_t shared_perm)
 
 /*
  * Creates a new BlockBackend, opens a new BlockDriverState, and connects both.
+ * The new BlockBackend is in the main AioContext.
  *
  * Just as with bdrv_open(), after having called this function the reference to
  * @options belongs to the block layer (even on failure).
@@ -382,7 +385,7 @@ BlockBackend *blk_new_open(const char *filename, const char *reference,
         perm |= BLK_PERM_RESIZE;
     }
 
-    blk = blk_new(perm, BLK_PERM_ALL);
+    blk = blk_new(qemu_get_aio_context(), perm, BLK_PERM_ALL);
     bs = bdrv_open(filename, reference, options, flags, errp);
     if (!bs) {
         blk_unref(blk);
@@ -1857,7 +1860,15 @@ void blk_op_unblock_all(BlockBackend *blk, Error *reason)
 
 AioContext *blk_get_aio_context(BlockBackend *blk)
 {
-    return bdrv_get_aio_context(blk_bs(blk));
+    BlockDriverState *bs = blk_bs(blk);
+
+    /* FIXME The AioContext of bs and blk can be inconsistent. For the moment,
+     * we prefer the one of bs for compatibility. */
+    if (bs) {
+        return bdrv_get_aio_context(blk_bs(blk));
+    }
+
+    return blk->ctx;
 }
 
 static AioContext *blk_aiocb_get_aio_context(BlockAIOCB *acb)
@@ -1889,6 +1900,7 @@ static int blk_do_set_aio_context(BlockBackend *blk, AioContext *new_context,
         }
     }
 
+    blk->ctx = new_context;
     return 0;
 }
 
block/commit.c +7 -4
index 14e5bb394c..4d519506d6 100644
@@ -338,7 +338,8 @@ void commit_start(const char *job_id, BlockDriverState *bs,
         goto fail;
     }
 
-    s->base = blk_new(BLK_PERM_CONSISTENT_READ
+    s->base = blk_new(s->common.job.aio_context,
+                      BLK_PERM_CONSISTENT_READ
                       | BLK_PERM_WRITE
                       | BLK_PERM_RESIZE,
                       BLK_PERM_CONSISTENT_READ
@@ -351,7 +352,7 @@ void commit_start(const char *job_id, BlockDriverState *bs,
     s->base_bs = base;
 
     /* Required permissions are already taken with block_job_add_bdrv() */
-    s->top = blk_new(0, BLK_PERM_ALL);
+    s->top = blk_new(s->common.job.aio_context, 0, BLK_PERM_ALL);
     ret = blk_insert_bs(s->top, top, errp);
     if (ret < 0) {
         goto fail;
@@ -395,6 +396,7 @@ int bdrv_commit(BlockDriverState *bs)
     BlockDriverState *backing_file_bs = NULL;
     BlockDriverState *commit_top_bs = NULL;
     BlockDriver *drv = bs->drv;
+    AioContext *ctx;
     int64_t offset, length, backing_length;
     int ro;
     int64_t n;
@@ -422,8 +424,9 @@ int bdrv_commit(BlockDriverState *bs)
         }
     }
 
-    src = blk_new(BLK_PERM_CONSISTENT_READ, BLK_PERM_ALL);
-    backing = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
+    ctx = bdrv_get_aio_context(bs);
+    src = blk_new(ctx, BLK_PERM_CONSISTENT_READ, BLK_PERM_ALL);
+    backing = blk_new(ctx, BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
 
     ret = blk_insert_bs(src, bs, &local_err);
     if (ret < 0) {
block/crypto.c +2 -1
index 3af46b805f..7351fd479d 100644
@@ -257,7 +257,8 @@ static int block_crypto_co_create_generic(BlockDriverState *bs,
     QCryptoBlock *crypto = NULL;
     struct BlockCryptoCreateData data;
 
-    blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
+    blk = blk_new(bdrv_get_aio_context(bs),
+                  BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
 
     ret = blk_insert_bs(blk, bs, errp);
     if (ret < 0) {
block/mirror.c +2 -1
index ec4bd9f404..eb96b52de9 100644
@@ -1584,7 +1584,8 @@ static void mirror_start_job(const char *job_id, BlockDriverState *bs,
      * We can allow anything except resize there.*/
     target_is_backing = bdrv_chain_contains(bs, target);
     target_graph_mod = (backing_mode != MIRROR_LEAVE_BACKING_CHAIN);
-    s->target = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE |
+    s->target = blk_new(s->common.job.aio_context,
+                        BLK_PERM_WRITE | BLK_PERM_RESIZE |
                         (target_graph_mod ? BLK_PERM_GRAPH_MOD : 0),
                         BLK_PERM_WRITE_UNCHANGED |
                         (target_is_backing ? BLK_PERM_CONSISTENT_READ |
block/parallels.c +2 -1
index 2747400577..00fae125d1 100644
@@ -554,7 +554,8 @@ static int coroutine_fn parallels_co_create(BlockdevCreateOptions* opts,
         return -EIO;
     }
 
-    blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
+    blk = blk_new(bdrv_get_aio_context(bs),
+                  BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
     ret = blk_insert_bs(blk, bs, errp);
     if (ret < 0) {
         goto out;
block/qcow.c +2 -1
index 1bb8fd05e2..6dee5bb792 100644
@@ -844,7 +844,8 @@ static int coroutine_fn qcow_co_create(BlockdevCreateOptions *opts,
         return -EIO;
     }
 
-    qcow_blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
+    qcow_blk = blk_new(bdrv_get_aio_context(bs),
+                       BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
     ret = blk_insert_bs(qcow_blk, bs, errp);
     if (ret < 0) {
         goto exit;
block/qcow2.c +4 -2
index eb93ea05a4..4f27b1aabe 100644
@@ -3103,7 +3103,8 @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp)
     }
 
     /* Create BlockBackend to write to the image */
-    blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
+    blk = blk_new(bdrv_get_aio_context(bs),
+                  BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
     ret = blk_insert_bs(blk, bs, errp);
     if (ret < 0) {
         goto out;
@@ -5100,7 +5101,8 @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts,
     }
 
     if (new_size) {
-        BlockBackend *blk = blk_new(BLK_PERM_RESIZE, BLK_PERM_ALL);
+        BlockBackend *blk = blk_new(bdrv_get_aio_context(bs),
+                                    BLK_PERM_RESIZE, BLK_PERM_ALL);
         ret = blk_insert_bs(blk, bs, errp);
         if (ret < 0) {
             blk_unref(blk);
block/qed.c +2 -1
index dcdcd62b4a..bb4f5c9863 100644
@@ -649,7 +649,8 @@ static int coroutine_fn bdrv_qed_co_create(BlockdevCreateOptions *opts,
         return -EIO;
     }
 
-    blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
+    blk = blk_new(bdrv_get_aio_context(bs),
+                  BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
     ret = blk_insert_bs(blk, bs, errp);
     if (ret < 0) {
         goto out;
block/sheepdog.c +2 -1
index cbdfe9ab6e..f76d6ddbbc 100644
@@ -1800,7 +1800,8 @@ static int sd_prealloc(BlockDriverState *bs, int64_t old_size, int64_t new_size,
     void *buf = NULL;
     int ret;
 
-    blk = blk_new(BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE | BLK_PERM_RESIZE,
+    blk = blk_new(bdrv_get_aio_context(bs),
+                  BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE | BLK_PERM_RESIZE,
                   BLK_PERM_ALL);
 
     ret = blk_insert_bs(blk, bs, errp);
block/vdi.c +2 -1
index d7ef6628e7..b9845a4cbd 100644
@@ -803,7 +803,8 @@ static int coroutine_fn vdi_co_do_create(BlockdevCreateOptions *create_options,
         goto exit;
     }
 
-    blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
+    blk = blk_new(bdrv_get_aio_context(bs_file),
+                  BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
     ret = blk_insert_bs(blk, bs_file, errp);
     if (ret < 0) {
         goto exit;
block/vhdx.c +2 -1
index a143a57657..d6070b6fa8 100644
@@ -1900,7 +1900,8 @@ static int coroutine_fn vhdx_co_create(BlockdevCreateOptions *opts,
         return -EIO;
     }
 
-    blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
+    blk = blk_new(bdrv_get_aio_context(bs),
+                  BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
     ret = blk_insert_bs(blk, bs, errp);
     if (ret < 0) {
         goto delete_and_exit;
block/vmdk.c +2 -1
index de8cb859f8..51067c774f 100644
@@ -2356,7 +2356,8 @@ static BlockBackend *vmdk_co_create_cb(int64_t size, int idx,
     if (!bs) {
         return NULL;
     }
-    blk = blk_new(BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE | BLK_PERM_RESIZE,
+    blk = blk_new(bdrv_get_aio_context(bs),
+                  BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE | BLK_PERM_RESIZE,
                   BLK_PERM_ALL);
     if (blk_insert_bs(blk, bs, errp)) {
         bdrv_unref(bs);
block/vpc.c +2 -1
index 0c279b87c8..d4776ee8a5 100644
@@ -1011,7 +1011,8 @@ static int coroutine_fn vpc_co_create(BlockdevCreateOptions *opts,
         return -EIO;
     }
 
-    blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
+    blk = blk_new(bdrv_get_aio_context(bs),
+                  BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL);
     ret = blk_insert_bs(blk, bs, errp);
     if (ret < 0) {
         goto out;
blockdev.c +2 -2
index e856ca4be9..04abbc61c7 100644
@@ -574,7 +574,7 @@ static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
     if ((!file || !*file) && !qdict_size(bs_opts)) {
         BlockBackendRootState *blk_rs;
 
-        blk = blk_new(0, BLK_PERM_ALL);
+        blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
         blk_rs = blk_get_root_state(blk);
         blk_rs->open_flags    = bdrv_flags;
         blk_rs->read_only     = read_only;
@@ -3138,7 +3138,7 @@ void qmp_block_resize(bool has_device, const char *device,
         goto out;
     }
 
-    blk = blk_new(BLK_PERM_RESIZE, BLK_PERM_ALL);
+    blk = blk_new(qemu_get_aio_context(), BLK_PERM_RESIZE, BLK_PERM_ALL);
     ret = blk_insert_bs(blk, bs, errp);
     if (ret < 0) {
         goto out;
blockjob.c +1 -1
index 9ca942ba01..0700481652 100644
@@ -392,7 +392,7 @@ void *block_job_create(const char *job_id, const BlockJobDriver *driver,
         job_id = bdrv_get_device_name(bs);
     }
 
-    blk = blk_new(perm, shared_perm);
+    blk = blk_new(bdrv_get_aio_context(bs), perm, shared_perm);
     ret = blk_insert_bs(blk, bs, errp);
     if (ret < 0) {
         blk_unref(blk);
hmp.c +2 -1
index 56a3ed7375..be5e345c6f 100644
@@ -2560,7 +2560,8 @@ void hmp_qemu_io(Monitor *mon, const QDict *qdict)
     if (!blk) {
         BlockDriverState *bs = bdrv_lookup_bs(NULL, device, &err);
         if (bs) {
-            blk = local_blk = blk_new(0, BLK_PERM_ALL);
+            blk = local_blk = blk_new(bdrv_get_aio_context(bs),
+                                      0, BLK_PERM_ALL);
             ret = blk_insert_bs(blk, bs, &err);
             if (ret < 0) {
                 goto fail;
hw/block/fdc.c +1 -1
index 6f19f127a5..37ccedc9f7 100644
@@ -538,7 +538,7 @@ static void floppy_drive_realize(DeviceState *qdev, Error **errp)
 
     if (!dev->conf.blk) {
         /* Anonymous BlockBackend for an empty drive */
-        dev->conf.blk = blk_new(0, BLK_PERM_ALL);
+        dev->conf.blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
         ret = blk_attach_dev(dev->conf.blk, qdev);
         assert(ret == 0);
     }
hw/block/xen-block.c +1 -1
index ef635be4c2..31b0f5ccc8 100644
@@ -609,7 +609,7 @@ static void xen_cdrom_realize(XenBlockDevice *blockdev, Error **errp)
         int rc;
 
         /* Set up an empty drive */
-        conf->blk = blk_new(0, BLK_PERM_ALL);
+        conf->blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
 
         rc = blk_attach_dev(conf->blk, DEVICE(blockdev));
         if (!rc) {
hw/core/qdev-properties-system.c +3 -1
index b45a7ef54b..42e048f190 100644
@@ -80,7 +80,9 @@ static void parse_drive(DeviceState *dev, const char *str, void **ptr,
     if (!blk) {
         BlockDriverState *bs = bdrv_lookup_bs(NULL, str, NULL);
         if (bs) {
-            blk = blk_new(0, BLK_PERM_ALL);
+            /* BlockBackends of devices start in the main context and are only
+             * later moved into another context if the device supports that. */
+            blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
             blk_created = true;
 
             ret = blk_insert_bs(blk, bs, errp);
hw/ide/qdev.c +1 -1
index 573b022e1e..360cd20bd8 100644
@@ -168,7 +168,7 @@ static void ide_dev_initfn(IDEDevice *dev, IDEDriveKind kind, Error **errp)
             return;
         } else {
             /* Anonymous BlockBackend for an empty drive */
-            dev->conf.blk = blk_new(0, BLK_PERM_ALL);
+            dev->conf.blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
             ret = blk_attach_dev(dev->conf.blk, &dev->qdev);
             assert(ret == 0);
         }
hw/scsi/scsi-disk.c +1 -1
index e7e865ab3b..91c5a8b1ac 100644
@@ -2417,7 +2417,7 @@ static void scsi_cd_realize(SCSIDevice *dev, Error **errp)
     if (!dev->conf.blk) {
         /* Anonymous BlockBackend for an empty drive. As we put it into
          * dev->conf, qdev takes care of detaching on unplug. */
-        dev->conf.blk = blk_new(0, BLK_PERM_ALL);
+        dev->conf.blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
         ret = blk_attach_dev(dev->conf.blk, &dev->qdev);
         assert(ret == 0);
     }
migration/block.c +2 -1
index 83c633fb3f..91f98ef44a 100644
@@ -417,7 +417,8 @@ static int init_blk_migration(QEMUFile *f)
         }
 
         bmds = g_new0(BlkMigDevState, 1);
-        bmds->blk = blk_new(BLK_PERM_CONSISTENT_READ, BLK_PERM_ALL);
+        bmds->blk = blk_new(qemu_get_aio_context(),
+                            BLK_PERM_CONSISTENT_READ, BLK_PERM_ALL);
         bmds->blk_name = g_strdup(bdrv_get_device_name(bs));
         bmds->bulk_completed = 0;
         bmds->total_sectors = sectors;
nbd/server.c +3 -2
index e21bd501dc..1e2c7f4270 100644
@@ -1484,8 +1484,9 @@ NBDExport *nbd_export_new(BlockDriverState *bs, uint64_t dev_offset,
     if ((nbdflags & NBD_FLAG_READ_ONLY) == 0) {
         perm |= BLK_PERM_WRITE;
     }
-    blk = blk_new(perm, BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE_UNCHANGED |
-                        BLK_PERM_WRITE | BLK_PERM_GRAPH_MOD);
+    blk = blk_new(bdrv_get_aio_context(bs), perm,
+                  BLK_PERM_CONSISTENT_READ | BLK_PERM_WRITE_UNCHANGED |
+                  BLK_PERM_WRITE | BLK_PERM_GRAPH_MOD);
     ret = blk_insert_bs(blk, bs, errp);
     if (ret < 0) {
         goto fail;
tests/test-bdrv-drain.c +15 -15
index 6d5565e8e7..2b907fae8b 100644
@@ -206,7 +206,7 @@ static void test_drv_cb_common(enum drain_type drain_type, bool recursive)
 
     QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, NULL, 0);
 
-    blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     bs = bdrv_new_open_driver(&bdrv_test, "test-node", BDRV_O_RDWR,
                               &error_abort);
     s = bs->opaque;
@@ -290,7 +290,7 @@ static void test_quiesce_common(enum drain_type drain_type, bool recursive)
     BlockBackend *blk;
     BlockDriverState *bs, *backing;
 
-    blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     bs = bdrv_new_open_driver(&bdrv_test, "test-node", BDRV_O_RDWR,
                               &error_abort);
     blk_insert_bs(blk, bs, &error_abort);
@@ -353,7 +353,7 @@ static void test_nested(void)
     BDRVTestState *s, *backing_s;
     enum drain_type outer, inner;
 
-    blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     bs = bdrv_new_open_driver(&bdrv_test, "test-node", BDRV_O_RDWR,
                               &error_abort);
     s = bs->opaque;
@@ -402,13 +402,13 @@ static void test_multiparent(void)
     BlockDriverState *bs_a, *bs_b, *backing;
     BDRVTestState *a_s, *b_s, *backing_s;
 
-    blk_a = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk_a = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     bs_a = bdrv_new_open_driver(&bdrv_test, "test-node-a", BDRV_O_RDWR,
                                 &error_abort);
     a_s = bs_a->opaque;
     blk_insert_bs(blk_a, bs_a, &error_abort);
 
-    blk_b = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk_b = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     bs_b = bdrv_new_open_driver(&bdrv_test, "test-node-b", BDRV_O_RDWR,
                                 &error_abort);
     b_s = bs_b->opaque;
@@ -475,13 +475,13 @@ static void test_graph_change_drain_subtree(void)
     BlockDriverState *bs_a, *bs_b, *backing;
     BDRVTestState *a_s, *b_s, *backing_s;
 
-    blk_a = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk_a = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     bs_a = bdrv_new_open_driver(&bdrv_test, "test-node-a", BDRV_O_RDWR,
                                 &error_abort);
     a_s = bs_a->opaque;
     blk_insert_bs(blk_a, bs_a, &error_abort);
 
-    blk_b = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk_b = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     bs_b = bdrv_new_open_driver(&bdrv_test, "test-node-b", BDRV_O_RDWR,
                                 &error_abort);
     b_s = bs_b->opaque;
@@ -555,7 +555,7 @@ static void test_graph_change_drain_all(void)
     BDRVTestState *a_s, *b_s;
 
     /* Create node A with a BlockBackend */
-    blk_a = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk_a = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     bs_a = bdrv_new_open_driver(&bdrv_test, "test-node-a", BDRV_O_RDWR,
                                 &error_abort);
     a_s = bs_a->opaque;
@@ -571,7 +571,7 @@ static void test_graph_change_drain_all(void)
     g_assert_cmpint(a_s->drain_count, ==, 1);
 
     /* Create node B with a BlockBackend */
-    blk_b = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk_b = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     bs_b = bdrv_new_open_driver(&bdrv_test, "test-node-b", BDRV_O_RDWR,
                                 &error_abort);
     b_s = bs_b->opaque;
@@ -672,7 +672,7 @@ static void test_iothread_common(enum drain_type drain_type, int drain_thread)
         goto out;
     }
 
-    blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     bs = bdrv_new_open_driver(&bdrv_test, "test-node", BDRV_O_RDWR,
                               &error_abort);
     s = bs->opaque;
@@ -883,7 +883,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type,
     bdrv_set_backing_hd(src, src_backing, &error_abort);
     bdrv_unref(src_backing);
 
-    blk_src = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk_src = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     blk_insert_bs(blk_src, src_overlay, &error_abort);
 
     switch (drain_node) {
@@ -910,7 +910,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type,
 
     target = bdrv_new_open_driver(&bdrv_test, "target", BDRV_O_RDWR,
                                   &error_abort);
-    blk_target = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk_target = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     blk_insert_bs(blk_target, target, &error_abort);
 
     aio_context_acquire(ctx);
@@ -1205,7 +1205,7 @@ static void do_test_delete_by_drain(bool detach_instead_of_delete,
                         &error_abort);
     bdrv_attach_child(bs, null_bs, "null-child", &child_file, &error_abort);
 
-    blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     blk_insert_bs(blk, bs, &error_abort);
 
     /* Referenced by blk now */
@@ -1368,7 +1368,7 @@ static void test_detach_indirect(bool by_parent_cb)
     c = bdrv_new_open_driver(&bdrv_test, "c", BDRV_O_RDWR, &error_abort);
 
     /* blk is a BB for parent-a */
-    blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     blk_insert_bs(blk, parent_a, &error_abort);
     bdrv_unref(parent_a);
 
@@ -1466,7 +1466,7 @@ static void test_append_to_drained(void)
     BlockDriverState *base, *overlay;
     BDRVTestState *base_s, *overlay_s;
 
-    blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     base = bdrv_new_open_driver(&bdrv_test, "base", BDRV_O_RDWR, &error_abort);
     base_s = base->opaque;
     blk_insert_bs(blk, base, &error_abort);
tests/test-bdrv-graph-mod.c +3 -2
index 283dc84869..21401be454 100644
@@ -102,7 +102,8 @@ static void test_update_perm_tree(void)
 {
     Error *local_err = NULL;
 
-    BlockBackend *root = blk_new(BLK_PERM_WRITE | BLK_PERM_CONSISTENT_READ,
+    BlockBackend *root = blk_new(qemu_get_aio_context(),
+                                 BLK_PERM_WRITE | BLK_PERM_CONSISTENT_READ,
                                  BLK_PERM_ALL & ~BLK_PERM_WRITE);
     BlockDriverState *bs = no_perm_node("node");
     BlockDriverState *filter = pass_through_node("filter");
@@ -166,7 +167,7 @@ static void test_update_perm_tree(void)
  */
 static void test_should_update_child(void)
 {
-    BlockBackend *root = blk_new(0, BLK_PERM_ALL);
+    BlockBackend *root = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
     BlockDriverState *bs = no_perm_node("node");
     BlockDriverState *filter = no_perm_node("filter");
     BlockDriverState *target = no_perm_node("target");
tests/test-block-backend.c +2 -2
index fd59f02bd0..dee3c92103 100644
@@ -37,7 +37,7 @@ static void test_drain_aio_error_flush_cb(void *opaque, int ret)
 
 static void test_drain_aio_error(void)
 {
-    BlockBackend *blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    BlockBackend *blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     BlockAIOCB *acb;
     bool completed = false;
 
@@ -53,7 +53,7 @@ static void test_drain_aio_error(void)
 
 static void test_drain_all_aio_error(void)
 {
-    BlockBackend *blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    BlockBackend *blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     BlockAIOCB *acb;
     bool completed = false;
 
tests/test-block-iothread.c +5 -5
index 1d47ea9895..2200487d76 100644
@@ -336,7 +336,7 @@ static void test_sync_op(const void *opaque)
     BlockDriverState *bs;
     BdrvChild *c;
 
-    blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     bs = bdrv_new_open_driver(&bdrv_test, "base", BDRV_O_RDWR, &error_abort);
     bs->total_sectors = 65536 / BDRV_SECTOR_SIZE;
     blk_insert_bs(blk, bs, &error_abort);
@@ -415,7 +415,7 @@ static void test_attach_blockjob(void)
     BlockDriverState *bs;
     TestBlockJob *tjob;
 
-    blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     bs = bdrv_new_open_driver(&bdrv_test, "base", BDRV_O_RDWR, &error_abort);
     blk_insert_bs(blk, bs, &error_abort);
 
@@ -481,7 +481,7 @@ static void test_propagate_basic(void)
     QDict *options;
 
     /* Create bs_a and its BlockBackend */
-    blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     bs_a = bdrv_new_open_driver(&bdrv_test, "bs_a", BDRV_O_RDWR, &error_abort);
     blk_insert_bs(blk, bs_a, &error_abort);
 
@@ -561,7 +561,7 @@ static void test_propagate_diamond(void)
     qdict_put_str(options, "raw", "bs_c");
 
     bs_verify = bdrv_open(NULL, NULL, options, BDRV_O_RDWR, &error_abort);
-    blk = blk_new(BLK_PERM_ALL, BLK_PERM_ALL);
+    blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     blk_insert_bs(blk, bs_verify, &error_abort);
 
     /* Switch the AioContext */
@@ -628,7 +628,7 @@ static void test_propagate_mirror(void)
     g_assert(bdrv_get_aio_context(filter) == main_ctx);
 
     /* With a BlockBackend on src, changing target must fail */
-    blk = blk_new(0, BLK_PERM_ALL);
+    blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
     blk_insert_bs(blk, src, &error_abort);
 
     bdrv_try_set_aio_context(target, ctx, &local_err);
tests/test-blockjob.c +1 -1
index 652d1e8359..8c91980c70 100644
@@ -68,7 +68,7 @@ static BlockJob *do_test_id(BlockBackend *blk, const char *id,
 static BlockBackend *create_blk(const char *name)
 {
     /* No I/O is performed on this device */
-    BlockBackend *blk = blk_new(0, BLK_PERM_ALL);
+    BlockBackend *blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
     BlockDriverState *bs;
 
     bs = bdrv_open("null-co://", NULL, NULL, 0, &error_abort);
tests/test-throttle.c +3 -3
index 948a42c991..5644cf95ca 100644
@@ -675,9 +675,9 @@ static void test_groups(void)
     ThrottleGroupMember *tgm1, *tgm2, *tgm3;
 
     /* No actual I/O is performed on these devices */
-    blk1 = blk_new(0, BLK_PERM_ALL);
-    blk2 = blk_new(0, BLK_PERM_ALL);
-    blk3 = blk_new(0, BLK_PERM_ALL);
+    blk1 = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
+    blk2 = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
+    blk3 = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
 
     blkp1 = blk_get_public(blk1);
     blkp2 = blk_get_public(blk2);
Signature · 1 line
2.20.1
author
#5 lore raw

PATCH 4/15 block: Add qdev_prop_drive_iothread property type

In reply to Kevin Wolf

Some qdev block devices have support for iothreads and take care of the AioContext they are running in, but most devices don't know about any of this. For the latter category, the qdev drive property must make sure that their BlockBackend is in the main AioContext.

Unfortunately, while the current code just does the same thing for devices that do support iothreads, this is not correct and it would show as soon as we actually try to keep a consistent AioContext assignment across all nodes and users of a block graph subtree: If a node is already in a non-default AioContext because of one of its users, attaching a new device should still be possible if that device can work in the same AioContext. Switching the node back to the main context first and only then into the device AioContext causes failure (because the existing user wouldn't allow the switch to the main context).

So devices that support iothreads need a different kind of drive property that leaves the node in its current AioContext, but by using this type, the device promises to check later that it can work with this context.

This patch adds the qdev infrastructure that allows devices to signal that they handle iothreads and qdev should leave the AioContext alone.

---
 include/hw/block/block.h         |  7 ++++--
 include/hw/qdev-properties.h     |  3 +++
 hw/core/qdev-properties-system.c | 43 ++++++++++++++++++++++++++++----
 3 files changed, 46 insertions(+), 7 deletions(-)
include/hw/block/block.h +5 -2
index d06f25aa0f..607539057a 100644
@@ -45,8 +45,7 @@ static inline unsigned int get_physical_block_exp(BlockConf *conf)
     return exp;
 }
 
-#define DEFINE_BLOCK_PROPERTIES(_state, _conf)                          \
-    DEFINE_PROP_DRIVE("drive", _state, _conf.blk),                      \
+#define DEFINE_BLOCK_PROPERTIES_BASE(_state, _conf)                     \
     DEFINE_PROP_BLOCKSIZE("logical_block_size", _state,                 \
                           _conf.logical_block_size),                    \
     DEFINE_PROP_BLOCKSIZE("physical_block_size", _state,                \
@@ -59,6 +58,10 @@ static inline unsigned int get_physical_block_exp(BlockConf *conf)
                             ON_OFF_AUTO_AUTO), \
     DEFINE_PROP_BOOL("share-rw", _state, _conf.share_rw, false)
 
+#define DEFINE_BLOCK_PROPERTIES(_state, _conf)                          \
+    DEFINE_PROP_DRIVE("drive", _state, _conf.blk),                      \
+    DEFINE_BLOCK_PROPERTIES_BASE(_state, _conf)
+
 #define DEFINE_BLOCK_CHS_PROPERTIES(_state, _conf)      \
     DEFINE_PROP_UINT32("cyls", _state, _conf.cyls, 0),  \
     DEFINE_PROP_UINT32("heads", _state, _conf.heads, 0), \
include/hw/qdev-properties.h +3 -0
index b6758c852e..1eae5ab056 100644
@@ -28,6 +28,7 @@ extern const PropertyInfo qdev_prop_blockdev_on_error;
 extern const PropertyInfo qdev_prop_bios_chs_trans;
 extern const PropertyInfo qdev_prop_fdc_drive_type;
 extern const PropertyInfo qdev_prop_drive;
+extern const PropertyInfo qdev_prop_drive_iothread;
 extern const PropertyInfo qdev_prop_netdev;
 extern const PropertyInfo qdev_prop_pci_devfn;
 extern const PropertyInfo qdev_prop_blocksize;
@@ -198,6 +199,8 @@ extern const PropertyInfo qdev_prop_pcie_link_width;
     DEFINE_PROP(_n, _s, _f, qdev_prop_netdev, NICPeers)
 #define DEFINE_PROP_DRIVE(_n, _s, _f) \
     DEFINE_PROP(_n, _s, _f, qdev_prop_drive, BlockBackend *)
+#define DEFINE_PROP_DRIVE_IOTHREAD(_n, _s, _f) \
+    DEFINE_PROP(_n, _s, _f, qdev_prop_drive_iothread, BlockBackend *)
 #define DEFINE_PROP_MACADDR(_n, _s, _f)         \
     DEFINE_PROP(_n, _s, _f, qdev_prop_macaddr, MACAddr)
 #define DEFINE_PROP_ON_OFF_AUTO(_n, _s, _f, _d) \
hw/core/qdev-properties-system.c +38 -5

73 lines not on this page · raw message

Signature · 1 line
2.20.1
author
#6 lore raw

PATCH 5/15 scsi-disk: Use qdev_prop_drive_iothread

In reply to Kevin Wolf

This makes use of qdev_prop_drive_iothread for scsi-disk so that the disk can be attached to a node that is already in the target AioContext. We need to check that the HBA actually supports iothreads, otherwise scsi-disk must make sure that the node is already in the main AioContext.

This changes the error message for conflicting iothread settings. Previously, virtio-scsi produced the error message, now it comes from blk_set_aio_context(). Update a test case accordingly.

---
 include/hw/scsi/scsi.h     |  1 +
 hw/scsi/scsi-disk.c        | 22 +++++++++++++++-------
 hw/scsi/virtio-scsi.c      | 15 ++++++++-------
 tests/qemu-iotests/240.out |  2 +-
 4 files changed, 25 insertions(+), 15 deletions(-)
include/hw/scsi/scsi.h +1 -0
index acef25faa4..426566a5c6 100644
@@ -88,6 +88,7 @@ struct SCSIDevice
     int scsi_version;
     int default_scsi_version;
     bool needs_vpd_bl_emulation;
+    bool hba_supports_iothread;
 };
 
 extern const VMStateDescription vmstate_scsi_device;
hw/scsi/scsi-disk.c +15 -7
index 91c5a8b1ac..7b89ac798b 100644
@@ -2336,6 +2336,13 @@ static void scsi_realize(SCSIDevice *dev, Error **errp)
         return;
     }
 
+    if (blk_get_aio_context(s->qdev.conf.blk) != qemu_get_aio_context() &&
+        !s->qdev.hba_supports_iothread)
+    {
+        error_setg(errp, "HBA does not support iothreads");
+        return;
+    }
+
     if (dev->type == TYPE_DISK) {
         if (!blkconf_geometry(&dev->conf, NULL, 65535, 255, 255, errp)) {
             return;
@@ -2929,13 +2936,14 @@ static const TypeInfo scsi_disk_base_info = {
     .abstract      = true,
 };
 
-#define DEFINE_SCSI_DISK_PROPERTIES()                                \
-    DEFINE_BLOCK_PROPERTIES(SCSIDiskState, qdev.conf),               \
-    DEFINE_BLOCK_ERROR_PROPERTIES(SCSIDiskState, qdev.conf),         \
-    DEFINE_PROP_STRING("ver", SCSIDiskState, version),               \
-    DEFINE_PROP_STRING("serial", SCSIDiskState, serial),             \
-    DEFINE_PROP_STRING("vendor", SCSIDiskState, vendor),             \
-    DEFINE_PROP_STRING("product", SCSIDiskState, product),           \
+#define DEFINE_SCSI_DISK_PROPERTIES()                                   \
+    DEFINE_PROP_DRIVE_IOTHREAD("drive", SCSIDiskState, qdev.conf.blk),  \
+    DEFINE_BLOCK_PROPERTIES_BASE(SCSIDiskState, qdev.conf),             \
+    DEFINE_BLOCK_ERROR_PROPERTIES(SCSIDiskState, qdev.conf),            \
+    DEFINE_PROP_STRING("ver", SCSIDiskState, version),                  \
+    DEFINE_PROP_STRING("serial", SCSIDiskState, serial),                \
+    DEFINE_PROP_STRING("vendor", SCSIDiskState, vendor),                \
+    DEFINE_PROP_STRING("product", SCSIDiskState, product),              \
     DEFINE_PROP_STRING("device_id", SCSIDiskState, device_id)
 
 
hw/scsi/virtio-scsi.c +8 -7
index 01c2b85f90..2994f0738f 100644
@@ -789,6 +789,13 @@ static void virtio_scsi_change(SCSIBus *bus, SCSIDevice *dev, SCSISense sense)
     }
 }
 
+static void virtio_scsi_pre_hotplug(HotplugHandler *hotplug_dev,
+                                    DeviceState *dev, Error **errp)
+{
+    SCSIDevice *sd = SCSI_DEVICE(dev);
+    sd->hba_supports_iothread = true;
+}
+
 static void virtio_scsi_hotplug(HotplugHandler *hotplug_dev, DeviceState *dev,
                                 Error **errp)
 {
@@ -798,16 +805,9 @@ static void virtio_scsi_hotplug(HotplugHandler *hotplug_dev, DeviceState *dev,
     int ret;
 
     if (s->ctx && !s->dataplane_fenced) {
-        AioContext *ctx;
         if (blk_op_is_blocked(sd->conf.blk, BLOCK_OP_TYPE_DATAPLANE, errp)) {
             return;
         }
-        ctx = blk_get_aio_context(sd->conf.blk);
-        if (ctx != s->ctx && ctx != qemu_get_aio_context()) {
-            error_setg(errp, "Cannot attach a blockdev that is using "
-                       "a different iothread");
-            return;
-        }
         virtio_scsi_acquire(s);
         ret = blk_set_aio_context(sd->conf.blk, s->ctx, errp);
         virtio_scsi_release(s);
@@ -990,6 +990,7 @@ static void virtio_scsi_class_init(ObjectClass *klass, void *data)
     vdc->reset = virtio_scsi_reset;
     vdc->start_ioeventfd = virtio_scsi_dataplane_start;
     vdc->stop_ioeventfd = virtio_scsi_dataplane_stop;
+    hc->pre_plug = virtio_scsi_pre_hotplug;
     hc->plug = virtio_scsi_hotplug;
     hc->unplug = virtio_scsi_hotunplug;
 }
tests/qemu-iotests/240.out +1 -1
index d76392966c..b2288ead70 100644
@@ -43,7 +43,7 @@ QMP_VERSION
 {"return": {}}
 {"return": {}}
 {"return": {}}
-{"error": {"class": "GenericError", "desc": "Cannot attach a blockdev that is using a different iothread"}}
+{"error": {"class": "GenericError", "desc": "Cannot change iothread of active block backend"}}
 {"return": {}}
 {"return": {}}
 {"return": {}}
Signature · 1 line
2.20.1
author
#7 lore raw

PATCH 6/15 block: Adjust AioContexts when attaching nodes

In reply to Kevin Wolf

So far, we only made sure that updating the AioContext of a node affected the whole subtree. However, if a node is newly attached to a new parent, we also need to make sure that both the subtree of the node and the parent are in the same AioContext. This tries to move the new child node to the parent AioContext and returns an error if this isn't possible.

BlockBackends now actually apply their AioContext to their root node.

---
 include/block/block_int.h |  1 +
 block.c                   | 33 ++++++++++++++++++++++++++++++++-
 block/block-backend.c     |  9 ++++-----
 blockjob.c                | 10 ++++++++--
 tests/test-bdrv-drain.c   |  6 ++++--
 5 files changed, 49 insertions(+), 10 deletions(-)
include/block/block_int.h +1 -0
index 1eebc7c8f3..06df2bda1b 100644
@@ -1163,6 +1163,7 @@ void hmp_drive_add_node(Monitor *mon, const char *optstr);
 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
                                   const char *child_name,
                                   const BdrvChildRole *child_role,
+                                  AioContext *ctx,
                                   uint64_t perm, uint64_t shared_perm,
                                   void *opaque, Error **errp);
 void bdrv_root_unref_child(BdrvChild *child);
block.c +32 -1
index 301135b985..5219f61279 100644
@@ -2243,13 +2243,17 @@ static void bdrv_replace_child(BdrvChild *child, BlockDriverState *new_bs)
     }
 }
 
+/* The caller must hold the AioContext lock @child_bs, but not that of @ctx
+ * (unless @child_bs is already in @ctx). */
 BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
                                   const char *child_name,
                                   const BdrvChildRole *child_role,
+                                  AioContext *ctx,
                                   uint64_t perm, uint64_t shared_perm,
                                   void *opaque, Error **errp)
 {
     BdrvChild *child;
+    Error *local_err = NULL;
     int ret;
 
     ret = bdrv_check_update_perm(child_bs, NULL, perm, shared_perm, NULL, errp);
@@ -2268,12 +2272,39 @@ BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
         .opaque         = opaque,
     };
 
+    /* If the AioContexts don't match, first try to move the subtree of
+     * child_bs into the AioContext of the new parent. If this doesn't work,
+     * try moving the parent into the AioContext of child_bs instead. */
+    if (bdrv_get_aio_context(child_bs) != ctx) {
+        ret = bdrv_try_set_aio_context(child_bs, ctx, &local_err);
+        if (ret < 0 && child_role->can_set_aio_ctx) {
+            GSList *ignore = g_slist_prepend(NULL, child);;
+            ctx = bdrv_get_aio_context(child_bs);
+            if (child_role->can_set_aio_ctx(child, ctx, &ignore, NULL)) {
+                error_free(local_err);
+                ret = 0;
+                g_slist_free(ignore);
+                ignore = g_slist_prepend(NULL, child);;
+                child_role->set_aio_ctx(child, ctx, &ignore);
+            }
+            g_slist_free(ignore);
+        }
+        if (ret < 0) {
+            error_propagate(errp, local_err);
+            g_free(child);
+            bdrv_abort_perm_update(child_bs);
+            return NULL;
+        }
+    }
+
     /* This performs the matching bdrv_set_perm() for the above check. */
     bdrv_replace_child(child, child_bs);
 
     return child;
 }
 
+/* If @parent_bs and @child_bs are in different AioContexts, the caller must
+ * hold the AioContext lock for @child_bs, but not for @parent_bs. */
 BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
                              BlockDriverState *child_bs,
                              const char *child_name,
@@ -2286,11 +2317,11 @@ BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
     bdrv_get_cumulative_perm(parent_bs, &perm, &shared_perm);
 
     assert(parent_bs->drv);
-    assert(bdrv_get_aio_context(parent_bs) == bdrv_get_aio_context(child_bs));
     bdrv_child_perm(parent_bs, child_bs, NULL, child_role, NULL,
                     perm, shared_perm, &perm, &shared_perm);
 
     child = bdrv_root_attach_child(child_bs, child_name, child_role,
+                                   bdrv_get_aio_context(parent_bs),
                                    perm, shared_perm, parent_bs, errp);
     if (child == NULL) {
         return NULL;
block/block-backend.c +4 -5
index 8eca3261e3..6b0c4ef58c 100644
@@ -392,7 +392,7 @@ BlockBackend *blk_new_open(const char *filename, const char *reference,
         return NULL;
     }
 
-    blk->root = bdrv_root_attach_child(bs, "root", &child_root,
+    blk->root = bdrv_root_attach_child(bs, "root", &child_root, blk->ctx,
                                        perm, BLK_PERM_ALL, blk, errp);
     if (!blk->root) {
         bdrv_unref(bs);
@@ -803,7 +803,7 @@ void blk_remove_bs(BlockBackend *blk)
 int blk_insert_bs(BlockBackend *blk, BlockDriverState *bs, Error **errp)
 {
     ThrottleGroupMember *tgm = &blk->public.throttle_group_member;
-    blk->root = bdrv_root_attach_child(bs, "root", &child_root,
+    blk->root = bdrv_root_attach_child(bs, "root", &child_root, blk->ctx,
                                        blk->perm, blk->shared_perm, blk, errp);
     if (blk->root == NULL) {
         return -EPERM;
@@ -1862,10 +1862,9 @@ AioContext *blk_get_aio_context(BlockBackend *blk)
 {
     BlockDriverState *bs = blk_bs(blk);
 
-    /* FIXME The AioContext of bs and blk can be inconsistent. For the moment,
-     * we prefer the one of bs for compatibility. */
     if (bs) {
-        return bdrv_get_aio_context(blk_bs(blk));
+        AioContext *ctx = bdrv_get_aio_context(blk_bs(blk));
+        assert(ctx == blk->ctx);
     }
 
     return blk->ctx;
blockjob.c +8 -2
index 0700481652..c27be0e07e 100644
@@ -204,8 +204,14 @@ int block_job_add_bdrv(BlockJob *job, const char *name, BlockDriverState *bs,
 {
     BdrvChild *c;
 
-    c = bdrv_root_attach_child(bs, name, &child_job, perm, shared_perm,
-                               job, errp);
+    if (job->job.aio_context != qemu_get_aio_context()) {
+        aio_context_release(job->job.aio_context);
+    }
+    c = bdrv_root_attach_child(bs, name, &child_job, job->job.aio_context,
+                               perm, shared_perm, job, errp);
+    if (job->job.aio_context != qemu_get_aio_context()) {
+        aio_context_acquire(job->job.aio_context);
+    }
     if (c == NULL) {
         return -EPERM;
     }
tests/test-bdrv-drain.c +4 -2
index 2b907fae8b..447f6d12eb 100644
@@ -912,6 +912,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type,
                                   &error_abort);
     blk_target = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);
     blk_insert_bs(blk_target, target, &error_abort);
+    blk_set_allow_aio_context_change(blk_target, true);
 
     aio_context_acquire(ctx);
     tjob = block_job_create("job0", &test_job_driver, NULL, src,
@@ -972,7 +973,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type,
     g_assert_false(job->job.paused);
     g_assert_true(job->job.busy); /* We're in qemu_co_sleep_ns() */
 
-    do_drain_begin(drain_type, target);
+    do_drain_begin_unlocked(drain_type, target);
 
     if (drain_type == BDRV_DRAIN_ALL) {
         /* bdrv_drain_all() drains both src and target */
@@ -983,7 +984,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type,
     g_assert_true(job->job.paused);
     g_assert_false(job->job.busy); /* The job is paused */
 
-    do_drain_end(drain_type, target);
+    do_drain_end_unlocked(drain_type, target);
 
     if (use_iothread) {
         /* paused is reset in the I/O thread, wait for it */
@@ -1002,6 +1003,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type,
 
     if (use_iothread) {
         blk_set_aio_context(blk_src, qemu_get_aio_context(), &error_abort);
+        blk_set_aio_context(blk_target, qemu_get_aio_context(), &error_abort);
     }
     aio_context_release(ctx);
 
Signature · 1 line
2.20.1
author
#8 lore raw

PATCH 7/15 test-block-iothread: Test adding parent to iothread node

In reply to Kevin Wolf

Opening a new parent node for a node that has already been moved into a different AioContext must cause the new parent ot move into the same context.

---
 tests/test-block-iothread.c | 33 +++++++++++++++++++++++++++++++++
 1 file changed, 33 insertions(+)
tests/test-block-iothread.c +33 -0
index 2200487d76..38f59999ab 100644
@@ -663,6 +663,38 @@ static void test_propagate_mirror(void)
     bdrv_unref(target);
 }
 
+static void test_attach_second_node(void)
+{
+    IOThread *iothread = iothread_new();
+    AioContext *ctx = iothread_get_aio_context(iothread);
+    AioContext *main_ctx = qemu_get_aio_context();
+    BlockBackend *blk;
+    BlockDriverState *bs, *filter;
+    QDict *options;
+
+    blk = blk_new(ctx, BLK_PERM_ALL, BLK_PERM_ALL);
+    bs = bdrv_new_open_driver(&bdrv_test, "base", BDRV_O_RDWR, &error_abort);
+    blk_insert_bs(blk, bs, &error_abort);
+
+    options = qdict_new();
+    qdict_put_str(options, "driver", "raw");
+    qdict_put_str(options, "file", "base");
+
+    filter = bdrv_open(NULL, NULL, options, BDRV_O_RDWR, &error_abort);
+    g_assert(blk_get_aio_context(blk) == ctx);
+    g_assert(bdrv_get_aio_context(bs) == ctx);
+    g_assert(bdrv_get_aio_context(filter) == ctx);
+
+    blk_set_aio_context(blk, main_ctx, &error_abort);
+    g_assert(blk_get_aio_context(blk) == main_ctx);
+    g_assert(bdrv_get_aio_context(bs) == main_ctx);
+    g_assert(bdrv_get_aio_context(filter) == main_ctx);
+
+    bdrv_unref(filter);
+    bdrv_unref(bs);
+    blk_unref(blk);
+}
+
 int main(int argc, char **argv)
 {
     int i;
@@ -678,6 +710,7 @@ int main(int argc, char **argv)
     }
 
     g_test_add_func("/attach/blockjob", test_attach_blockjob);
+    g_test_add_func("/attach/second_node", test_attach_second_node);
     g_test_add_func("/propagate/basic", test_propagate_basic);
     g_test_add_func("/propagate/diamond", test_propagate_diamond);
     g_test_add_func("/propagate/mirror", test_propagate_mirror);
Signature · 1 line
2.20.1
#9 lore raw

Re: [Qemu-devel] [PATCH 07/15] test-block-iothread: Test adding parent to iothread node

Quoted 2 lines · Kevin Wolf

Opening a new parent node for a node that has already been moved into a different AioContext must cause the new parent ot move into the same

to

Quoted 7 lines

context.

---
 tests/test-block-iothread.c | 33 +++++++++++++++++++++++++++++++++
 1 file changed, 33 insertions(+)
Signature · 3 lines
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org
author
#10 lore raw

PATCH 8/15 test-block-iothread: BlockBackend AioContext across root node change

In reply to Kevin Wolf

Test that BlockBackends preserve their assigned AioContext even when the root node goes away. Inserting a new root node will move it to the right AioContext.

---
 tests/test-block-iothread.c | 33 +++++++++++++++++++++++++++++++++
 1 file changed, 33 insertions(+)
tests/test-block-iothread.c +33 -0
index 38f59999ab..f41082e3bd 100644
@@ -695,6 +695,38 @@ static void test_attach_second_node(void)
     blk_unref(blk);
 }
 
+static void test_attach_preserve_blk_ctx(void)
+{
+    IOThread *iothread = iothread_new();
+    AioContext *ctx = iothread_get_aio_context(iothread);
+    BlockBackend *blk;
+    BlockDriverState *bs;
+
+    blk = blk_new(ctx, BLK_PERM_ALL, BLK_PERM_ALL);
+    bs = bdrv_new_open_driver(&bdrv_test, "base", BDRV_O_RDWR, &error_abort);
+    bs->total_sectors = 65536 / BDRV_SECTOR_SIZE;
+
+    /* Add node to BlockBackend that has an iothread context assigned */
+    blk_insert_bs(blk, bs, &error_abort);
+    g_assert(blk_get_aio_context(blk) == ctx);
+    g_assert(bdrv_get_aio_context(bs) == ctx);
+
+    /* Remove the node again */
+    blk_remove_bs(blk);
+    /* TODO bs should move back to main context here */
+    g_assert(blk_get_aio_context(blk) == ctx);
+    g_assert(bdrv_get_aio_context(bs) == ctx);
+
+    /* Re-attach the node */
+    blk_insert_bs(blk, bs, &error_abort);
+    g_assert(blk_get_aio_context(blk) == ctx);
+    g_assert(bdrv_get_aio_context(bs) == ctx);
+
+    blk_set_aio_context(blk, qemu_get_aio_context(), &error_abort);
+    bdrv_unref(bs);
+    blk_unref(blk);
+}
+
 int main(int argc, char **argv)
 {
     int i;
@@ -711,6 +743,7 @@ int main(int argc, char **argv)
 
     g_test_add_func("/attach/blockjob", test_attach_blockjob);
     g_test_add_func("/attach/second_node", test_attach_second_node);
+    g_test_add_func("/attach/preserve_blk_ctx", test_attach_preserve_blk_ctx);
     g_test_add_func("/propagate/basic", test_propagate_basic);
     g_test_add_func("/propagate/diamond", test_propagate_diamond);
     g_test_add_func("/propagate/mirror", test_propagate_mirror);
Signature · 1 line
2.20.1
author
#11 lore raw

PATCH 9/15 block: Move node without parents to main AioContext

In reply to Kevin Wolf

A node should only be in a non-default AioContext if a user is attached to it that requires this. When the last parent of a node is gone, it can move back to the main AioContext.

---
 block.c                     | 4 ++++
 tests/test-bdrv-drain.c     | 2 +-
 tests/test-block-iothread.c | 3 +--
 3 files changed, 6 insertions(+), 3 deletions(-)
block.c +4 -0
index 5219f61279..19268f52f8 100644
@@ -2235,6 +2235,10 @@ static void bdrv_replace_child(BdrvChild *child, BlockDriverState *new_bs)
         bdrv_get_cumulative_perm(old_bs, &perm, &shared_perm);
         bdrv_check_perm(old_bs, NULL, perm, shared_perm, NULL, &error_abort);
         bdrv_set_perm(old_bs, perm, shared_perm);
+
+        /* When the parent requiring a non-default AioContext is removed, the
+         * node moves back to the main AioContext */
+        bdrv_try_set_aio_context(old_bs, qemu_get_aio_context(), NULL);
     }
 
     if (new_bs) {
tests/test-bdrv-drain.c +1 -1
index 447f6d12eb..2ff04db07a 100644
@@ -1003,7 +1003,7 @@ static void test_blockjob_common_drain_node(enum drain_type drain_type,
 
     if (use_iothread) {
         blk_set_aio_context(blk_src, qemu_get_aio_context(), &error_abort);
-        blk_set_aio_context(blk_target, qemu_get_aio_context(), &error_abort);
+        assert(blk_get_aio_context(blk_target) == qemu_get_aio_context());
     }
     aio_context_release(ctx);
 
tests/test-block-iothread.c +1 -2
index f41082e3bd..79d9cf8a57 100644
@@ -713,9 +713,8 @@ static void test_attach_preserve_blk_ctx(void)
 
     /* Remove the node again */
     blk_remove_bs(blk);
-    /* TODO bs should move back to main context here */
     g_assert(blk_get_aio_context(blk) == ctx);
-    g_assert(bdrv_get_aio_context(bs) == ctx);
+    g_assert(bdrv_get_aio_context(bs) == qemu_get_aio_context());
 
     /* Re-attach the node */
     blk_insert_bs(blk, bs, &error_abort);
Signature · 1 line
2.20.1
author
#12 lore raw

PATCH 10/15 blockdev: Use bdrv_try_set_aio_context() for monitor commands

In reply to Kevin Wolf

Monitor commands can handle errors, so they can easily be converted to using the safer bdrv_try_set_aio_context() function.

---
 blockdev.c | 44 ++++++++++++++++++++++++++++----------------
 1 file changed, 28 insertions(+), 16 deletions(-)
blockdev.c +28 -16
index 04abbc61c7..624534a05d 100644
@@ -1535,6 +1535,7 @@ static void external_snapshot_prepare(BlkActionState *common,
                              DO_UPCAST(ExternalSnapshotState, common, common);
     TransactionAction *action = common->action;
     AioContext *aio_context;
+    int ret;
 
     /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
      * purpose but a different set of parameters */
@@ -1674,7 +1675,10 @@ static void external_snapshot_prepare(BlkActionState *common,
         goto out;
     }
 
-    bdrv_set_aio_context(state->new_bs, aio_context);
+    ret = bdrv_try_set_aio_context(state->new_bs, aio_context, errp);
+    if (ret < 0) {
+        goto out;
+    }
 
     /* This removes our old bs and adds the new bs. This is an operation that
      * can fail, so we need to do it in .prepare; undoing it for abort is
@@ -3424,6 +3428,7 @@ static BlockJob *do_drive_backup(DriveBackup *backup, JobTxn *txn,
     int flags, job_flags = JOB_DEFAULT;
     int64_t size;
     bool set_backing_hd = false;
+    int ret;
 
     if (!backup->has_speed) {
         backup->speed = 0;
@@ -3520,7 +3525,11 @@ static BlockJob *do_drive_backup(DriveBackup *backup, JobTxn *txn,
         goto out;
     }
 
-    bdrv_set_aio_context(target_bs, aio_context);
+    ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
+    if (ret < 0) {
+        bdrv_unref(target_bs);
+        goto out;
+    }
 
     if (set_backing_hd) {
         bdrv_set_backing_hd(target_bs, source, &local_err);
@@ -3592,6 +3601,7 @@ BlockJob *do_blockdev_backup(BlockdevBackup *backup, JobTxn *txn,
     AioContext *aio_context;
     BlockJob *job = NULL;
     int job_flags = JOB_DEFAULT;
+    int ret;
 
     if (!backup->has_speed) {
         backup->speed = 0;
@@ -3628,16 +3638,9 @@ BlockJob *do_blockdev_backup(BlockdevBackup *backup, JobTxn *txn,
         goto out;
     }
 
-    if (bdrv_get_aio_context(target_bs) != aio_context) {
-        if (!bdrv_has_blk(target_bs)) {
-            /* The target BDS is not attached, we can safely move it to another
-             * AioContext. */
-            bdrv_set_aio_context(target_bs, aio_context);
-        } else {
-            error_setg(errp, "Target is attached to a different thread from "
-                             "source.");
-            goto out;
-        }
+    ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
+    if (ret < 0) {
+        goto out;
     }
 
     if (backup->has_bitmap) {
@@ -3810,6 +3813,7 @@ void qmp_drive_mirror(DriveMirror *arg, Error **errp)
     int flags;
     int64_t size;
     const char *format = arg->format;
+    int ret;
 
     bs = qmp_get_root_bs(arg->device, errp);
     if (!bs) {
@@ -3910,7 +3914,11 @@ void qmp_drive_mirror(DriveMirror *arg, Error **errp)
         goto out;
     }
 
-    bdrv_set_aio_context(target_bs, aio_context);
+    ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
+    if (ret < 0) {
+        bdrv_unref(target_bs);
+        goto out;
+    }
 
     blockdev_mirror_common(arg->has_job_id ? arg->job_id : NULL, bs, target_bs,
                            arg->has_replaces, arg->replaces, arg->sync,
@@ -3954,6 +3962,7 @@ void qmp_blockdev_mirror(bool has_job_id, const char *job_id,
     AioContext *aio_context;
     BlockMirrorBackingMode backing_mode = MIRROR_LEAVE_BACKING_CHAIN;
     Error *local_err = NULL;
+    int ret;
 
     bs = qmp_get_root_bs(device, errp);
     if (!bs) {
@@ -3968,7 +3977,10 @@ void qmp_blockdev_mirror(bool has_job_id, const char *job_id,
     aio_context = bdrv_get_aio_context(bs);
     aio_context_acquire(aio_context);
 
-    bdrv_set_aio_context(target_bs, aio_context);
+    ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
+    if (ret < 0) {
+        goto out;
+    }
 
     blockdev_mirror_common(has_job_id ? job_id : NULL, bs, target_bs,
                            has_replaces, replaces, sync, backing_mode,
@@ -3984,7 +3996,7 @@ void qmp_blockdev_mirror(bool has_job_id, const char *job_id,
                            has_auto_dismiss, auto_dismiss,
                            &local_err);
     error_propagate(errp, local_err);
-
+out:
     aio_context_release(aio_context);
 }
 
@@ -4474,7 +4486,7 @@ void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread,
     old_context = bdrv_get_aio_context(bs);
     aio_context_acquire(old_context);
 
-    bdrv_set_aio_context(bs, new_context);
+    bdrv_try_set_aio_context(bs, new_context, errp);
 
     aio_context_release(old_context);
 }
Signature · 1 line
2.20.1
author
#13 lore raw

PATCH 11/15 block: Remove wrong bdrv_set_aio_context() calls

In reply to Kevin Wolf

The mirror and commit block jobs use bdrv_set_aio_context() to move their filter node into the right AioContext before hooking it up in the graph. Similarly, bdrv_open_backing_file() explicitly moves the backing file node into the right AioContext first.

This isn't necessary any more, they get automatically moved into the right context now when attaching them.

However, in the case of bdrv_open_backing_file() with a node reference, it's actually not only unnecessary, but even wrong: The unchecked bdrv_set_aio_context() changes the AioContext of the child node even if other parents require it to retain the old context. So this is not only a simplification, but a bug fix, too.

---
 block.c        | 1 -
 block/commit.c | 2 --
 block/mirror.c | 1 -
 3 files changed, 4 deletions(-)
block.c +0 -1
index 19268f52f8..e9261c650d 100644
@@ -2539,7 +2539,6 @@ int bdrv_open_backing_file(BlockDriverState *bs, QDict *parent_options,
         ret = -EINVAL;
         goto free_exit;
     }
-    bdrv_set_aio_context(backing_hd, bdrv_get_aio_context(bs));
 
     if (implicit_backing) {
         bdrv_refresh_filename(backing_hd);
block/commit.c +0 -2
index 4d519506d6..c815def89a 100644
@@ -301,7 +301,6 @@ void commit_start(const char *job_id, BlockDriverState *bs,
         commit_top_bs->implicit = true;
     }
     commit_top_bs->total_sectors = top->total_sectors;
-    bdrv_set_aio_context(commit_top_bs, bdrv_get_aio_context(top));
 
     bdrv_append(commit_top_bs, top, &local_err);
     if (local_err) {
@@ -443,7 +442,6 @@ int bdrv_commit(BlockDriverState *bs)
         error_report_err(local_err);
         goto ro_cleanup;
     }
-    bdrv_set_aio_context(commit_top_bs, bdrv_get_aio_context(backing_file_bs));
 
     bdrv_set_backing_hd(commit_top_bs, backing_file_bs, &error_abort);
     bdrv_set_backing_hd(bs, commit_top_bs, &error_abort);
block/mirror.c +0 -1
index eb96b52de9..f8bdb5b21b 100644
@@ -1543,7 +1543,6 @@ static void mirror_start_job(const char *job_id, BlockDriverState *bs,
                                           BDRV_REQ_NO_FALLBACK;
     bs_opaque = g_new0(MirrorBDSOpaque, 1);
     mirror_top_bs->opaque = bs_opaque;
-    bdrv_set_aio_context(mirror_top_bs, bdrv_get_aio_context(bs));
 
     /* bdrv_append takes ownership of the mirror_top_bs reference, need to keep
      * it alive until block_job_create() succeeds even if bs has no parent. */
Signature · 1 line
2.20.1
author
#14 lore raw

PATCH 12/15 virtio-scsi-test: Test attaching new overlay with iothreads

In reply to Kevin Wolf

This tests that blockdev-add can correctly add a qcow2 overlay to an image used by a virtio-scsi disk in an iothread. The interesting point here is whether the newly added node gets correctly moved into the iothread AioContext.

If it isn't, we get an assertion failure in virtio-scsi while processing the next request:

    virtio_scsi_ctx_check: Assertion `blk_get_aio_context(d->conf.blk) == s->ctx' failed.
---
 tests/libqtest.h         | 11 +++++++
 tests/libqtest.c         | 19 ++++++++++++
 tests/virtio-scsi-test.c | 62 ++++++++++++++++++++++++++++++++++++++++
 3 files changed, 92 insertions(+)
tests/libqtest.h +11 -0
index a98ea15b7d..32d927755d 100644
@@ -619,6 +619,17 @@ static inline void qtest_end(void)
 QDict *qmp(const char *fmt, ...) GCC_FMT_ATTR(1, 2);
 
 /**
+ * qmp_assert_success:
+ * @fmt...: QMP message to send to qemu, formatted like
+ * qobject_from_jsonf_nofail().  See parse_escape() for what's
+ * supported after '%'.
+ *
+ * Sends a QMP message to QEMU and asserts that a 'return' key is present in
+ * the response.
+ */
+void qmp_assert_success(const char *fmt, ...) GCC_FMT_ATTR(1, 2);
+
+/*
  * qmp_eventwait:
  * @s: #event event to wait for.
  *
tests/libqtest.c +19 -0
index 8ac0c02af4..546a875913 100644
@@ -1038,6 +1038,25 @@ QDict *qmp(const char *fmt, ...)
     return response;
 }
 
+void qmp_assert_success(const char *fmt, ...)
+{
+    va_list ap;
+    QDict *response;
+
+    va_start(ap, fmt);
+    response = qtest_vqmp(global_qtest, fmt, ap);
+    va_end(ap);
+
+    g_assert(response);
+    if (!qdict_haskey(response, "return")) {
+        QString *s = qobject_to_json_pretty(QOBJECT(response));
+        g_test_message("%s", qstring_get_str(s));
+        qobject_unref(s);
+    }
+    g_assert(qdict_haskey(response, "return"));
+    qobject_unref(response);
+}
+
 char *hmp(const char *fmt, ...)
 {
     va_list ap;
tests/virtio-scsi-test.c +62 -0

80 lines not on this page · raw message

Signature · 1 line
2.20.1
author
#15 lore raw

PATCH 13/15 iotests: Attach new devices to node in non-default iothread

In reply to Kevin Wolf

This tests that devices refuse to be attached to a node that has already been moved to a different iothread if they can't be or aren't configured to work in the same iothread.

---
 tests/qemu-iotests/051        | 24 ++++++++++++++++++++++++
 tests/qemu-iotests/051.out    |  3 +++
 tests/qemu-iotests/051.pc.out | 27 +++++++++++++++++++++++++++
 3 files changed, 54 insertions(+)
tests/qemu-iotests/051 +24 -0
index a3deb1fcad..200660f977 100755
@@ -191,6 +191,30 @@ case "$QEMU_DEFAULT_MACHINE" in
         ;;
 esac
 
+echo
+echo === Attach to node in non-default iothread ===
+echo
+
+case "$QEMU_DEFAULT_MACHINE" in
+    pc)
+        iothread="-drive file=$TEST_IMG,if=none,node-name=disk -object iothread,id=thread0 -device virtio-scsi,iothread=thread0,id=virtio-scsi0 -device scsi-hd,bus=virtio-scsi0.0,drive=disk,share-rw=on"
+
+        # Can't add a device in the main thread while virtio-scsi0 uses the node
+        run_qemu $iothread -device ide-hd,drive=disk,share-rw=on
+        run_qemu $iothread -device virtio-blk-pci,drive=disk,share-rw=on
+        run_qemu $iothread -device lsi53c895a,id=lsi0 -device scsi-hd,bus=lsi0.0,drive=disk,share-rw=on
+        run_qemu $iothread -device virtio-scsi,id=virtio-scsi1 -device scsi-hd,bus=virtio-scsi1.0,drive=disk,share-rw=on
+
+        # virtio-blk enables the iothread only when the driver initialises the
+        # device, so a second virtio-blk device can't be added even with the
+        # same iothread. virtio-scsi allows this.
+        run_qemu $iothread -device virtio-blk-pci,drive=disk,iohtread=iothread0,share-rw=on
+        run_qemu $iothread -device virtio-scsi,id=virtio-scsi1,iothread=thread0 -device scsi-hd,bus=virtio-scsi1.0,drive=disk,share-rw=on
+        ;;
+     *)
+        ;;
+esac
+
 echo
 echo === Read-only ===
 echo
tests/qemu-iotests/051.out +3 -0
index 9f1cf22608..8993835b94 100644
@@ -137,6 +137,9 @@ QEMU X.Y.Z monitor - type 'help' for more information
 (qemu) QEMU_PROG: -drive if=virtio: Device needs media, but drive is empty
 
 
+=== Attach to node in non-default iothread ===
+
+
 === Read-only ===
 
 Testing: -drive file=TEST_DIR/t.qcow2,if=virtio,readonly=on
tests/qemu-iotests/051.pc.out +27 -0
index c4743cc31c..2d811c166c 100644
@@ -173,6 +173,33 @@ QEMU X.Y.Z monitor - type 'help' for more information
 (qemu) QEMU_PROG: -device scsi-hd,drive=disk: Device needs media, but drive is empty
 
 
+=== Attach to node in non-default iothread ===
+
+Testing: -drive file=TEST_DIR/t.qcow2,if=none,node-name=disk -object iothread,id=thread0 -device virtio-scsi,iothread=thread0,id=virtio-scsi0 -device scsi-hd,bus=virtio-scsi0.0,drive=disk,share-rw=on -device ide-hd,drive=disk,share-rw=on
+QEMU X.Y.Z monitor - type 'help' for more information
+(qemu) QEMU_PROG: -device ide-hd,drive=disk,share-rw=on: Cannot change iothread of active block backend
+
+Testing: -drive file=TEST_DIR/t.qcow2,if=none,node-name=disk -object iothread,id=thread0 -device virtio-scsi,iothread=thread0,id=virtio-scsi0 -device scsi-hd,bus=virtio-scsi0.0,drive=disk,share-rw=on -device virtio-blk-pci,drive=disk,share-rw=on
+QEMU X.Y.Z monitor - type 'help' for more information
+(qemu) QEMU_PROG: -device virtio-blk-pci,drive=disk,share-rw=on: Cannot change iothread of active block backend
+
+Testing: -drive file=TEST_DIR/t.qcow2,if=none,node-name=disk -object iothread,id=thread0 -device virtio-scsi,iothread=thread0,id=virtio-scsi0 -device scsi-hd,bus=virtio-scsi0.0,drive=disk,share-rw=on -device lsi53c895a,id=lsi0 -device scsi-hd,bus=lsi0.0,drive=disk,share-rw=on
+QEMU X.Y.Z monitor - type 'help' for more information
+(qemu) QEMU_PROG: -device scsi-hd,bus=lsi0.0,drive=disk,share-rw=on: HBA does not support iothreads
+
+Testing: -drive file=TEST_DIR/t.qcow2,if=none,node-name=disk -object iothread,id=thread0 -device virtio-scsi,iothread=thread0,id=virtio-scsi0 -device scsi-hd,bus=virtio-scsi0.0,drive=disk,share-rw=on -device virtio-scsi,id=virtio-scsi1 -device scsi-hd,bus=virtio-scsi1.0,drive=disk,share-rw=on
+QEMU X.Y.Z monitor - type 'help' for more information
+(qemu) QEMU_PROG: -device scsi-hd,bus=virtio-scsi1.0,drive=disk,share-rw=on: Cannot change iothread of active block backend
+
+Testing: -drive file=TEST_DIR/t.qcow2,if=none,node-name=disk -object iothread,id=thread0 -device virtio-scsi,iothread=thread0,id=virtio-scsi0 -device scsi-hd,bus=virtio-scsi0.0,drive=disk,share-rw=on -device virtio-blk-pci,drive=disk,iohtread=iothread0,share-rw=on
+QEMU X.Y.Z monitor - type 'help' for more information
+(qemu) QEMU_PROG: -device virtio-blk-pci,drive=disk,iohtread=iothread0,share-rw=on: Cannot change iothread of active block backend
+
+Testing: -drive file=TEST_DIR/t.qcow2,if=none,node-name=disk -object iothread,id=thread0 -device virtio-scsi,iothread=thread0,id=virtio-scsi0 -device scsi-hd,bus=virtio-scsi0.0,drive=disk,share-rw=on -device virtio-scsi,id=virtio-scsi1,iothread=thread0 -device scsi-hd,bus=virtio-scsi1.0,drive=disk,share-rw=on
+QEMU X.Y.Z monitor - type 'help' for more information
+(qemu) quit
+
+
 === Read-only ===
 
 Testing: -drive file=TEST_DIR/t.qcow2,if=floppy,readonly=on
Signature · 1 line
2.20.1
author
#16 lore raw

PATCH 14/15 test-bdrv-drain: Use bdrv_try_set_aio_context()

In reply to Kevin Wolf

No reason to use the unchecked version in tests, even more so when these are the last callers of bdrv_set_aio_context() outside of block.c.

---
 tests/test-bdrv-drain.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
tests/test-bdrv-drain.c +3 -3
index 2ff04db07a..fb7866efbd 100644
@@ -1515,16 +1515,16 @@ static void test_set_aio_context(void)
                               &error_abort);
 
     bdrv_drained_begin(bs);
-    bdrv_set_aio_context(bs, ctx_a);
+    bdrv_try_set_aio_context(bs, ctx_a, &error_abort);
 
     aio_context_acquire(ctx_a);
     bdrv_drained_end(bs);
 
     bdrv_drained_begin(bs);
-    bdrv_set_aio_context(bs, ctx_b);
+    bdrv_try_set_aio_context(bs, ctx_b, &error_abort);
     aio_context_release(ctx_a);
     aio_context_acquire(ctx_b);
-    bdrv_set_aio_context(bs, qemu_get_aio_context());
+    bdrv_try_set_aio_context(bs, qemu_get_aio_context(), &error_abort);
     aio_context_release(ctx_b);
     bdrv_drained_end(bs);
 
Signature · 1 line
2.20.1
author
#17 lore raw

PATCH 15/15 block: Remove bdrv_set_aio_context()

In reply to Kevin Wolf

All callers of bdrv_set_aio_context() are eliminated now, they have moved to bdrv_try_set_aio_context() and related safe functions. Remove bdrv_set_aio_context().

With this, we can now know that the .set_aio_ctx callback must be present in bdrv_set_aio_context_ignore() because bdrv_can_set_aio_context() would have returned false previously, so instead of checking the condition, we can assert it.

---
 docs/devel/multiple-iothreads.txt |  4 ++--
 include/block/block.h             |  9 ---------
 block.c                           | 30 ++++++++++++++----------------
 3 files changed, 16 insertions(+), 27 deletions(-)
docs/devel/multiple-iothreads.txt +2 -2
index 4f9012d154..aeb997bed5 100644
@@ -109,7 +109,7 @@ The AioContext originates from the QEMU block layer, even though nowadays
 AioContext is a generic event loop that can be used by any QEMU subsystem.
 
 The block layer has support for AioContext integrated.  Each BlockDriverState
-is associated with an AioContext using bdrv_set_aio_context() and
+is associated with an AioContext using bdrv_try_set_aio_context() and
 bdrv_get_aio_context().  This allows block layer code to process I/O inside the
 right AioContext.  Other subsystems may wish to follow a similar approach.
 
@@ -134,5 +134,5 @@ Long-running jobs (usually in the form of coroutines) are best scheduled in
 the BlockDriverState's AioContext to avoid the need to acquire/release around
 each bdrv_*() call.  The functions bdrv_add/remove_aio_context_notifier,
 or alternatively blk_add/remove_aio_context_notifier if you use BlockBackends,
-can be used to get a notification whenever bdrv_set_aio_context() moves a
+can be used to get a notification whenever bdrv_try_set_aio_context() moves a
 BlockDriverState to a different AioContext.
include/block/block.h +0 -9
index 531cf595cf..13ea050a5b 100644
@@ -583,15 +583,6 @@ AioContext *bdrv_get_aio_context(BlockDriverState *bs);
  */
 void bdrv_coroutine_enter(BlockDriverState *bs, Coroutine *co);
 
-/**
- * bdrv_set_aio_context:
- *
- * Changes the #AioContext used for fd handlers, timers, and BHs by this
- * BlockDriverState and all its children.
- *
- * This function must be called with iothread lock held.
- */
-void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context);
 void bdrv_set_aio_context_ignore(BlockDriverState *bs,
                                  AioContext *new_context, GSList **ignore);
 int bdrv_try_set_aio_context(BlockDriverState *bs, AioContext *ctx,
block.c +14 -16
index e9261c650d..628f65c41a 100644
@@ -5791,8 +5791,17 @@ static void bdrv_attach_aio_context(BlockDriverState *bs,
     bs->walking_aio_notifiers = false;
 }
 
-/* @ignore will accumulate all visited BdrvChild object. The caller is
- * responsible for freeing the list afterwards. */
+/*
+ * Changes the AioContext used for fd handlers, timers, and BHs by this
+ * BlockDriverState and all its children and parents.
+ *
+ * The caller must own the AioContext lock for the old AioContext of bs, but it
+ * must not own the AioContext lock for new_context (unless new_context is the
+ * same as the current context of bs).
+ *
+ * @ignore will accumulate all visited BdrvChild object. The caller is
+ * responsible for freeing the list afterwards.
+ */
 void bdrv_set_aio_context_ignore(BlockDriverState *bs,
                                  AioContext *new_context, GSList **ignore)
 {
@@ -5815,10 +5824,9 @@ void bdrv_set_aio_context_ignore(BlockDriverState *bs,
         if (g_slist_find(*ignore, child)) {
             continue;
         }
-        if (child->role->set_aio_ctx) {
-            *ignore = g_slist_prepend(*ignore, child);
-            child->role->set_aio_ctx(child, new_context, ignore);
-        }
+        assert(child->role->set_aio_ctx);
+        *ignore = g_slist_prepend(*ignore, child);
+        child->role->set_aio_ctx(child, new_context, ignore);
     }
 
     bdrv_detach_aio_context(bs);
@@ -5832,16 +5840,6 @@ void bdrv_set_aio_context_ignore(BlockDriverState *bs,
     aio_context_release(new_context);
 }
 
-/* The caller must own the AioContext lock for the old AioContext of bs, but it
- * must not own the AioContext lock for new_context (unless new_context is
- * the same as the current context of bs). */
-void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context)
-{
-    GSList *ignore_list = NULL;
-    bdrv_set_aio_context_ignore(bs, new_context, &ignore_list);
-    g_slist_free(ignore_list);
-}
-
 static bool bdrv_parent_can_set_aio_context(BdrvChild *c, AioContext *ctx,
                                             GSList **ignore, Error **errp)
 {
Signature · 1 line
2.20.1
#18 lore raw
In reply to Kevin Wolf
Patchew URL: https://patchew.org/QEMU/[email protected]/

Patchew URL: https://patchew.org/QEMU/[email protected]/

Hi,

This series seems to have some coding style problems. See output below for more information:

Subject: [Qemu-devel] [PATCH 00/15] block: AioContext management, part 2
Type: series
Message-id: [email protected]

=== TEST SCRIPT BEGIN ===
#!/bin/bash
git rev-parse base > /dev/null || exit 0
git config --local diff.renamelimit 0
git config --local diff.renames True
git config --local diff.algorithm histogram
./scripts/checkpatch.pl --mailback base..
=== TEST SCRIPT END ===

Updating 3c8cf5a9c21ff8782164d1def7f44bd888713384
Switched to a new branch 'test'
6ca27a5 block: Remove bdrv_set_aio_context()
984ba6d test-bdrv-drain: Use bdrv_try_set_aio_context()
b054561 iotests: Attach new devices to node in non-default iothread 98252b3 virtio-scsi-test: Test attaching new overlay with iothreads 7f6cd93 block: Remove wrong bdrv_set_aio_context() calls
0419720 blockdev: Use bdrv_try_set_aio_context() for monitor commands 26a1acf block: Move node without parents to main AioContext cd2b0d7 test-block-iothread: BlockBackend AioContext across root node change 5a4c60c test-block-iothread: Test adding parent to iothread node 1767a79 block: Adjust AioContexts when attaching nodes
16a9345 scsi-disk: Use qdev_prop_drive_iothread
cc6b17c block: Add qdev_prop_drive_iothread property type 2e30704 block: Add BlockBackend.ctx
15a3e30 block: Add Error to blk_set_aio_context()
88715f0 test-block-iothread: Check filter node in test_propagate_mirror

=== OUTPUT BEGIN ===
1/15 Checking commit 88715f0ee897 (test-block-iothread: Check filter node in test_propagate_mirror) 2/15 Checking commit 15a3e308457c (block: Add Error to blk_set_aio_context())
WARNING: Block comments use a leading /* on a separate line
#104: FILE: hw/block/dataplane/virtio-blk.c:289:
+ /* Drain and try to switch bs back to the QEMU main loop. If other users

WARNING: Block comments use a trailing */ on a separate line #105: FILE: hw/block/dataplane/virtio-blk.c:290:
+ * keep the BlockBackend in the iothread, that's ok */

total: 0 errors, 2 warnings, 259 lines checked

Patch 2/15 has style problems, please review. If any of these errors are false positives report them to the maintainer, see CHECKPATCH in MAINTAINERS.
3/15 Checking commit 2e30704b8e57 (block: Add BlockBackend.ctx)
WARNING: Block comments use a leading /* on a separate line #96: FILE: block/block-backend.c:1865:
+ /* FIXME The AioContext of bs and blk can be inconsistent. For the moment,

WARNING: Block comments use a trailing */ on a separate line #97: FILE: block/block-backend.c:1866:
+ * we prefer the one of bs for compatibility. */

WARNING: Block comments use a leading /* on a separate line #405: FILE: hw/core/qdev-properties-system.c:83:
+ /* BlockBackends of devices start in the main context and are only

WARNING: Block comments use a trailing */ on a separate line #406: FILE: hw/core/qdev-properties-system.c:84:
+ * later moved into another context if the device supports that. */

WARNING: line over 80 characters
#647: FILE: tests/test-block-backend.c:40:
+ BlockBackend *blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);

WARNING: line over 80 characters
#656: FILE: tests/test-block-backend.c:56:
+ BlockBackend *blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL);

total: 0 errors, 6 warnings, 533 lines checked

Patch 3/15 has style problems, please review. If any of these errors are false positives report them to the maintainer, see
CHECKPATCH in MAINTAINERS.
4/15 Checking commit cc6b17c2c35e (block: Add qdev_prop_drive_iothread property type)
ERROR: Macros with complex values should be enclosed in parenthesis #133: FILE: include/hw/block/block.h:61:
+#define DEFINE_BLOCK_PROPERTIES(_state, _conf) \
+ DEFINE_PROP_DRIVE("drive", _state, _conf.blk), \
+ DEFINE_BLOCK_PROPERTIES_BASE(_state, _conf)

total: 1 errors, 0 warnings, 107 lines checked

Patch 4/15 has style problems, please review. If any of these errors are false positives report them to the maintainer, see CHECKPATCH in MAINTAINERS.

5/15 Checking commit 16a934533950 (scsi-disk: Use qdev_prop_drive_iothread)
ERROR: Macros with complex values should be enclosed in parenthesis #49: FILE: hw/scsi/scsi-disk.c:2939:

+#define DEFINE_SCSI_DISK_PROPERTIES()                                   \
+    DEFINE_PROP_DRIVE_IOTHREAD("drive", SCSIDiskState, qdev.conf.blk),  \
+    DEFINE_BLOCK_PROPERTIES_BASE(SCSIDiskState, qdev.conf),             \
+    DEFINE_BLOCK_ERROR_PROPERTIES(SCSIDiskState, qdev.conf),            \
+    DEFINE_PROP_STRING("ver", SCSIDiskState, version),                  \
+    DEFINE_PROP_STRING("serial", SCSIDiskState, serial),                \
+    DEFINE_PROP_STRING("vendor", SCSIDiskState, vendor),                \
+    DEFINE_PROP_STRING("product", SCSIDiskState, product),              \

DEFINE_PROP_STRING("device_id", SCSIDiskState, device_id)

total: 1 errors, 0 warnings, 85 lines checked

Patch 5/15 has style problems, please review. If any of these errors are false positives report them to the maintainer, see CHECKPATCH in MAINTAINERS.

6/15 Checking commit 1767a7917e97 (block: Adjust AioContexts when attaching nodes)
WARNING: Block comments use a leading /* on a separate line
#27: FILE: block.c:2246:
+/* The caller must hold the AioContext lock @child_bs, but not that of @ctx

WARNING: Block comments use a trailing */ on a separate line #28: FILE: block.c:2247:
+ * (unless @child_bs is already in @ctx). */

WARNING: Block comments use a leading /* on a separate line #45: FILE: block.c:2275:
+ /* If the AioContexts don't match, first try to move the subtree of

WARNING: Block comments use a trailing */ on a separate line #47: FILE: block.c:2277:
+ * try moving the parent into the AioContext of child_bs instead. */

WARNING: Block comments use a leading /* on a separate line
#76: FILE: block.c:2306:
+/* If @parent_bs and @child_bs are in different AioContexts, the caller must

WARNING: Block comments use a trailing */ on a separate line #77: FILE: block.c:2307:
+ * hold the AioContext lock for @child_bs, but not for @parent_bs. */

total: 0 errors, 6 warnings, 149 lines checked

Patch 6/15 has style problems, please review. If any of these errors
are false positives report them to the maintainer, see
CHECKPATCH in MAINTAINERS.
7/15 Checking commit 5a4c60c72bfc (test-block-iothread: Test adding parent to iothread node) 8/15 Checking commit cd2b0d787e1b (test-block-iothread: BlockBackend AioContext across root node change) 9/15 Checking commit 26a1acffbe16 (block: Move node without parents to main AioContext)
WARNING: Block comments use a leading /* on a separate line
#23: FILE: block.c:2239:
+ /* When the parent requiring a non-default AioContext is removed, the

WARNING: Block comments use a trailing */ on a separate line #24: FILE: block.c:2240:
+ * node moves back to the main AioContext */

total: 0 errors, 2 warnings, 28 lines checked

Patch 9/15 has style problems, please review. If any of these errors
are false positives report them to the maintainer, see
CHECKPATCH in MAINTAINERS.
10/15 Checking commit 0419720e7eb2 (blockdev: Use bdrv_try_set_aio_context() for monitor commands) 11/15 Checking commit 7f6cd93a5e95 (block: Remove wrong bdrv_set_aio_context() calls) 12/15 Checking commit 98252b3fe05c (virtio-scsi-test: Test attaching new overlay with iothreads) 13/15 Checking commit b0545619fcbb (iotests: Attach new devices to node in non-default iothread) 14/15 Checking commit 984ba6d265e1 (test-bdrv-drain: Use bdrv_try_set_aio_context()) 15/15 Checking commit 6ca27a5484fa (block: Remove bdrv_set_aio_context())
=== OUTPUT END ===

Test command exited with code: 1

The full log is available at
http://patchew.org/logs/[email protected]/testing.checkpatch/?type=message.
---
Email generated automatically by Patchew [https://patchew.org/].
Please send your feedback to [email protected]

author
#19 lore raw

PATCH 1/1 nbd-server: Call blk_set_allow_aio_context_change()

In reply to Kevin Wolf

The NBD server uses an AioContext notifier, so it can tolerate that its BlockBackend is switched to a different AioContext. Before we start actually calling bdrv_try_set_aio_context(), which checks for consistency, outside of test cases, we need to make sure that the NBD server actually allows this.

---
 nbd/server.c               |  1 +
 tests/qemu-iotests/240     | 21 +++++++++++++++++++++
 tests/qemu-iotests/240.out | 13 +++++++++++++
 3 files changed, 35 insertions(+)
nbd/server.c +1 -0
index e21bd501dc..d1375350bc 100644
@@ -1491,6 +1491,7 @@ NBDExport *nbd_export_new(BlockDriverState *bs, uint64_t dev_offset,
         goto fail;
     }
     blk_set_enable_write_cache(blk, !writethrough);
+    blk_set_allow_aio_context_change(blk, true);
 
     exp->refcount = 1;
     QTAILQ_INIT(&exp->clients);
tests/qemu-iotests/240 +21 -0
index b4cf95096d..5be6b9c0f7 100755
@@ -27,6 +27,12 @@ echo "QA output created by $seq"
 
 status=1	# failure is the default!
 
+_cleanup()
+{
+    rm -f "$TEST_DIR/nbd"
+}
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
 # get standard environment, filters and checks
 . ./common.rc
 . ./common.filter
@@ -122,6 +128,21 @@ run_qemu <<EOF
 { "execute": "quit"}
 EOF
 
+echo
+echo === Attach a SCSI disks using the same block device as a NBD server ===
+echo
+
+run_qemu <<EOF
+{ "execute": "qmp_capabilities" }
+{ "execute": "blockdev-add", "arguments": {"driver": "null-co", "node-name": "hd0", "read-only": true}}
+{ "execute": "nbd-server-start", "arguments": {"addr":{"type":"unix","data":{"path":"$TEST_DIR/nbd"}}}}
+{ "execute": "nbd-server-add", "arguments": {"device":"hd0"}}
+{ "execute": "object-add", "arguments": {"qom-type": "iothread", "id": "iothread0"}}
+{ "execute": "device_add", "arguments": {"id": "scsi0", "driver": "${virtio_scsi}", "iothread": "iothread0"}}
+{ "execute": "device_add", "arguments": {"id": "scsi-hd0", "driver": "scsi-hd", "drive": "hd0", "bus": "scsi0.0"}}
+{ "execute": "quit"}
+EOF
+
 # success, all done
 echo "*** done"
 rm -f $seq.full
tests/qemu-iotests/240.out +13 -0
index d76392966c..84e0a43ce5 100644
@@ -51,4 +51,17 @@ QMP_VERSION
 {"return": {}}
 {"return": {}}
 {"return": {}}
+
+=== Attach a SCSI disks using the same block device as a NBD server ===
+
+Testing:
+QMP_VERSION
+{"return": {}}
+{"return": {}}
+{"return": {}}
+{"return": {}}
+{"return": {}}
+{"return": {}}
+{"return": {}}
+{"return": {}}
 *** done
Signature · 1 line
2.20.1
Reviewed-by on v1 1/1
#20 lore raw

Re: [Qemu-devel] [PATCH 1.5/15] nbd-server: Call blk_set_allow_aio_context_change()

Quoted 6 lines · Kevin Wolf

The NBD server uses an AioContext notifier, so it can tolerate that its BlockBackend is switched to a different AioContext. Before we start actually calling bdrv_try_set_aio_context(), which checks for consistency, outside of test cases, we need to make sure that the NBD server actually allows this.

Quoted 7 lines
---
 nbd/server.c               |  1 +
 tests/qemu-iotests/240     | 21 +++++++++++++++++++++
 tests/qemu-iotests/240.out | 13 +++++++++++++
 3 files changed, 35 insertions(+)
Signature · 3 lines
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org
author Applied
#21 lore raw
In reply to Kevin Wolf

Am 23.05.2019 um 18:00 hat Kevin Wolf geschrieben:

Quoted 12 lines

Recently, a few bugs were reported that resulted from an inconsistent state regarding AioContexts. Block nodes can end up in different contexts than their users expect - the AioContext of a node can even change under the feet of a device with no way for the device to forbid this. We recently added a few basic checks to scsi-disk and virtio-blk, but they are by far not enough.

Part 1 solved the first half of the problem and made sure that AioContext changes are propagated through the graph as necessary, so that a bdrv_set_aio_context() correctly pulls everything that uses the node into the right context.

Quoted 8 lines

This is part 2 that fixes the second half: Attaching new child nodes where the parent and child are already in different AioContexts. This operation may only succeed if we can move one of the node into the context of the other node.

After applying this series, unchecked bdrv_set_aio_context() doesn't exist any more and all users call functions that make sure that the AioContext assignments across the graph stays consistent.

Applied to the block branch.

Kevin