[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[PATCH v4 18/20] 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@xxxxxxxxxx>
---
 include/hw/scsi/scsi.h          | 14 ++++++++++++
 hw/scsi/scsi-bus.c              | 40 +++++++++++++++++++++++++++++++++
 hw/scsi/scsi-disk.c             | 27 +++++++++++++++++-----
 hw/scsi/virtio-scsi-dataplane.c | 22 ++++++++++--------
 hw/scsi/virtio-scsi.c           | 38 +++++++++++++++++++++++++++++++
 hw/scsi/trace-events            |  2 ++
 6 files changed, 129 insertions(+), 14 deletions(-)

diff --git a/include/hw/scsi/scsi.h b/include/hw/scsi/scsi.h
index 6f23a7a73e..e2bb1a2fbf 100644
--- a/include/hw/scsi/scsi.h
+++ b/include/hw/scsi/scsi.h
@@ -133,6 +133,16 @@ struct SCSIBusInfo {
     void (*save_request)(QEMUFile *f, SCSIRequest *req);
     void *(*load_request)(QEMUFile *f, SCSIRequest *req);
     void (*free_request)(SCSIBus *bus, void *priv);
+
+    /*
+     * Temporarily stop submitting new requests between drained_begin() and
+     * drained_end(). Called from the main loop thread with the BQL held.
+     *
+     * Implement these callbacks if request processing is triggered by a file
+     * descriptor like an EventNotifier. Otherwise set them to NULL.
+     */
+    void (*drained_begin)(SCSIBus *bus);
+    void (*drained_end)(SCSIBus *bus);
 };
 
 #define TYPE_SCSI_BUS "SCSI"
@@ -144,6 +154,8 @@ struct SCSIBus {
 
     SCSISense unit_attention;
     const SCSIBusInfo *info;
+
+    int drain_count; /* protected by BQL */
 };
 
 /**
@@ -213,6 +225,8 @@ void scsi_req_cancel_complete(SCSIRequest *req);
 void scsi_req_cancel(SCSIRequest *req);
 void scsi_req_cancel_async(SCSIRequest *req, Notifier *notifier);
 void scsi_req_retry(SCSIRequest *req);
+void scsi_device_drained_begin(SCSIDevice *sdev);
+void scsi_device_drained_end(SCSIDevice *sdev);
 void scsi_device_purge_requests(SCSIDevice *sdev, SCSISense sense);
 void scsi_device_set_ua(SCSIDevice *sdev, SCSISense sense);
 void scsi_device_report_change(SCSIDevice *dev, SCSISense sense);
diff --git a/hw/scsi/scsi-bus.c b/hw/scsi/scsi-bus.c
index 64d7311757..b571fdf895 100644
--- a/hw/scsi/scsi-bus.c
+++ b/hw/scsi/scsi-bus.c
@@ -1668,6 +1668,46 @@ 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 e01bd84541..2249087d6a 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/virtio-scsi-dataplane.c b/hw/scsi/virtio-scsi-dataplane.c
index 81643445ed..1060038e13 100644
--- a/hw/scsi/virtio-scsi-dataplane.c
+++ b/hw/scsi/virtio-scsi-dataplane.c
@@ -153,14 +153,16 @@ 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:
@@ -206,9 +208,11 @@ void virtio_scsi_dataplane_stop(VirtIODevice *vdev)
     }
     s->dataplane_stopping = true;
 
-    aio_context_acquire(s->ctx);
-    aio_wait_bh_oneshot(s->ctx, virtio_scsi_dataplane_stop_bh, s);
-    aio_context_release(s->ctx);
+    if (s->bus.drain_count == 0) {
+        aio_context_acquire(s->ctx);
+        aio_wait_bh_oneshot(s->ctx, virtio_scsi_dataplane_stop_bh, s);
+        aio_context_release(s->ctx);
+    }
 
     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 a02f9233ec..eba1e84dac 100644
--- a/hw/scsi/virtio-scsi.c
+++ b/hw/scsi/virtio-scsi.c
@@ -1081,6 +1081,42 @@ 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,
@@ -1095,6 +1131,8 @@ 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,
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"
-- 
2.39.2




 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.