virtio-scsi: implement BlockDevOps->drained_begin()

The virtio-scsi Host Bus Adapter provides access to devices on a SCSI
bus. Those SCSI devices typically have a BlockBackend. When the
BlockBackend enters a drained section, the SCSI device must temporarily
stop submitting new I/O requests.

Implement this behavior by temporarily stopping virtio-scsi virtqueue
processing when one of the SCSI devices enters a drained section. The
new scsi_device_drained_begin() API allows scsi-disk to message the
virtio-scsi HBA.

scsi_device_drained_begin() uses a drain counter so that multiple SCSI
devices can have overlapping drained sections. The HBA only sees one
pair of .drained_begin/end() calls.

After this commit, virtio-scsi no longer depends on hw/virtio's
ioeventfd aio_set_event_notifier(is_external=true). This commit is a
step towards removing the aio_disable_external() API.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20230516190238.8401-19-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

5 files changed, 113 insertions, 12 deletions

diff --git a/hw/scsi/scsi-bus.c b/hw/scsi/scsi-bus.c
index 64013c8a24..f80f4cb4fc 100644
--- a/hw/scsi/scsi-bus.c
+++ b/hw/scsi/scsi-bus.c
@@ -1669,6 +1669,46 @@ void scsi_device_purge_requests(SCSIDevice *sdev, SCSISense sense)
     scsi_device_set_ua(sdev, sense);
 }
 
+void scsi_device_drained_begin(SCSIDevice *sdev)
+{
+    SCSIBus *bus = DO_UPCAST(SCSIBus, qbus, sdev->qdev.parent_bus);
+    if (!bus) {
+        return;
+    }
+
+    assert(qemu_get_current_aio_context() == qemu_get_aio_context());
+    assert(bus->drain_count < INT_MAX);
+
+    /*
+     * Multiple BlockBackends can be on a SCSIBus and each may begin/end
+     * draining at any time. Keep a counter so HBAs only see begin/end once.
+     */
+    if (bus->drain_count++ == 0) {
+        trace_scsi_bus_drained_begin(bus, sdev);
+        if (bus->info->drained_begin) {
+            bus->info->drained_begin(bus);
+        }
+    }
+}
+
+void scsi_device_drained_end(SCSIDevice *sdev)
+{
+    SCSIBus *bus = DO_UPCAST(SCSIBus, qbus, sdev->qdev.parent_bus);
+    if (!bus) {
+        return;
+    }
+
+    assert(qemu_get_current_aio_context() == qemu_get_aio_context());
+    assert(bus->drain_count > 0);
+
+    if (bus->drain_count-- == 1) {
+        trace_scsi_bus_drained_end(bus, sdev);
+        if (bus->info->drained_end) {
+            bus->info->drained_end(bus);
+        }
+    }
+}
+
 static char *scsibus_get_dev_path(DeviceState *dev)
 {
     SCSIDevice *d = SCSI_DEVICE(dev);
diff --git a/hw/scsi/scsi-disk.c b/hw/scsi/scsi-disk.c
index 97c9b1c8cd..e0d79c7966 100644
--- a/hw/scsi/scsi-disk.c
+++ b/hw/scsi/scsi-disk.c
@@ -2360,6 +2360,20 @@ static void scsi_disk_reset(DeviceState *dev)
     s->qdev.scsi_version = s->qdev.default_scsi_version;
 }
 
+static void scsi_disk_drained_begin(void *opaque)
+{
+    SCSIDiskState *s = opaque;
+
+    scsi_device_drained_begin(&s->qdev);
+}
+
+static void scsi_disk_drained_end(void *opaque)
+{
+    SCSIDiskState *s = opaque;
+
+    scsi_device_drained_end(&s->qdev);
+}
+
 static void scsi_disk_resize_cb(void *opaque)
 {
     SCSIDiskState *s = opaque;
@@ -2414,16 +2428,19 @@ static bool scsi_cd_is_medium_locked(void *opaque)
 }
 
 static const BlockDevOps scsi_disk_removable_block_ops = {
-    .change_media_cb = scsi_cd_change_media_cb,
+    .change_media_cb  = scsi_cd_change_media_cb,
+    .drained_begin    = scsi_disk_drained_begin,
+    .drained_end      = scsi_disk_drained_end,
     .eject_request_cb = scsi_cd_eject_request_cb,
-    .is_tray_open = scsi_cd_is_tray_open,
     .is_medium_locked = scsi_cd_is_medium_locked,
-
-    .resize_cb = scsi_disk_resize_cb,
+    .is_tray_open     = scsi_cd_is_tray_open,
+    .resize_cb        = scsi_disk_resize_cb,
 };
 
 static const BlockDevOps scsi_disk_block_ops = {
-    .resize_cb = scsi_disk_resize_cb,
+    .drained_begin = scsi_disk_drained_begin,
+    .drained_end   = scsi_disk_drained_end,
+    .resize_cb     = scsi_disk_resize_cb,
 };
 
 static void scsi_disk_unit_attention_reported(SCSIDevice *dev)
diff --git a/hw/scsi/trace-events b/hw/scsi/trace-events
index ab238293f0..bdd4e2c7c7 100644
--- a/hw/scsi/trace-events
+++ b/hw/scsi/trace-events
@@ -6,6 +6,8 @@ scsi_req_cancel(int target, int lun, int tag) "target %d lun %d tag %d"
 scsi_req_data(int target, int lun, int tag, int len) "target %d lun %d tag %d len %d"
 scsi_req_data_canceled(int target, int lun, int tag, int len) "target %d lun %d tag %d len %d"
 scsi_req_dequeue(int target, int lun, int tag) "target %d lun %d tag %d"
+scsi_bus_drained_begin(void *bus, void *sdev) "bus %p sdev %p"
+scsi_bus_drained_end(void *bus, void *sdev) "bus %p sdev %p"
 scsi_req_continue(int target, int lun, int tag) "target %d lun %d tag %d"
 scsi_req_continue_canceled(int target, int lun, int tag) "target %d lun %d tag %d"
 scsi_req_parsed(int target, int lun, int tag, int cmd, int mode, int xfer) "target %d lun %d tag %d command %d dir %d length %d"
diff --git a/hw/scsi/virtio-scsi-dataplane.c b/hw/scsi/virtio-scsi-dataplane.c
index b3a1ed21f7..d55de4c8ca 100644
--- a/hw/scsi/virtio-scsi-dataplane.c
+++ b/hw/scsi/virtio-scsi-dataplane.c
@@ -158,14 +158,16 @@ int virtio_scsi_dataplane_start(VirtIODevice *vdev)
     s->dataplane_starting = false;
     s->dataplane_started = true;
 
-    aio_context_acquire(s->ctx);
-    virtio_queue_aio_attach_host_notifier(vs->ctrl_vq, s->ctx);
-    virtio_queue_aio_attach_host_notifier_no_poll(vs->event_vq, s->ctx);
+    if (s->bus.drain_count == 0) {
+        aio_context_acquire(s->ctx);
+        virtio_queue_aio_attach_host_notifier(vs->ctrl_vq, s->ctx);
+        virtio_queue_aio_attach_host_notifier_no_poll(vs->event_vq, s->ctx);
 
-    for (i = 0; i < vs->conf.num_queues; i++) {
-        virtio_queue_aio_attach_host_notifier(vs->cmd_vqs[i], s->ctx);
+        for (i = 0; i < vs->conf.num_queues; i++) {
+            virtio_queue_aio_attach_host_notifier(vs->cmd_vqs[i], s->ctx);
+        }
+        aio_context_release(s->ctx);
     }
-    aio_context_release(s->ctx);
     return 0;
 
 fail_host_notifiers:
@@ -211,7 +213,9 @@ void virtio_scsi_dataplane_stop(VirtIODevice *vdev)
     }
     s->dataplane_stopping = true;
 
-    aio_wait_bh_oneshot(s->ctx, virtio_scsi_dataplane_stop_bh, s);
+    if (s->bus.drain_count == 0) {
+        aio_wait_bh_oneshot(s->ctx, virtio_scsi_dataplane_stop_bh, s);
+    }
 
     blk_drain_all(); /* ensure there are no in-flight requests */
 
diff --git a/hw/scsi/virtio-scsi.c b/hw/scsi/virtio-scsi.c
index c1a7ea9ae2..4a8849cc7e 100644
--- a/hw/scsi/virtio-scsi.c
+++ b/hw/scsi/virtio-scsi.c
@@ -1117,6 +1117,42 @@ static void virtio_scsi_hotunplug(HotplugHandler *hotplug_dev, DeviceState *dev,
     }
 }
 
+/* Suspend virtqueue ioeventfd processing during drain */
+static void virtio_scsi_drained_begin(SCSIBus *bus)
+{
+    VirtIOSCSI *s = container_of(bus, VirtIOSCSI, bus);
+    VirtIODevice *vdev = VIRTIO_DEVICE(s);
+    uint32_t total_queues = VIRTIO_SCSI_VQ_NUM_FIXED +
+                            s->parent_obj.conf.num_queues;
+
+    if (!s->dataplane_started) {
+        return;
+    }
+
+    for (uint32_t i = 0; i < total_queues; i++) {
+        VirtQueue *vq = virtio_get_queue(vdev, i);
+        virtio_queue_aio_detach_host_notifier(vq, s->ctx);
+    }
+}
+
+/* Resume virtqueue ioeventfd processing after drain */
+static void virtio_scsi_drained_end(SCSIBus *bus)
+{
+    VirtIOSCSI *s = container_of(bus, VirtIOSCSI, bus);
+    VirtIODevice *vdev = VIRTIO_DEVICE(s);
+    uint32_t total_queues = VIRTIO_SCSI_VQ_NUM_FIXED +
+                            s->parent_obj.conf.num_queues;
+
+    if (!s->dataplane_started) {
+        return;
+    }
+
+    for (uint32_t i = 0; i < total_queues; i++) {
+        VirtQueue *vq = virtio_get_queue(vdev, i);
+        virtio_queue_aio_attach_host_notifier(vq, s->ctx);
+    }
+}
+
 static struct SCSIBusInfo virtio_scsi_scsi_info = {
     .tcq = true,
     .max_channel = VIRTIO_SCSI_MAX_CHANNEL,
@@ -1131,6 +1167,8 @@ static struct SCSIBusInfo virtio_scsi_scsi_info = {
     .get_sg_list = virtio_scsi_get_sg_list,
     .save_request = virtio_scsi_save_request,
     .load_request = virtio_scsi_load_request,
+    .drained_begin = virtio_scsi_drained_begin,
+    .drained_end = virtio_scsi_drained_end,
 };
 
 void virtio_scsi_common_realize(DeviceState *dev,