diff mbox series

[01/12] virtio-scsi: replace AioContext lock with tmf_bh_lock

Message ID 20231129195553.942921-2-stefanha@redhat.com (mailing list archive)
State New, archived
Headers show
Series aio: remove AioContext lock | expand

Commit Message

Stefan Hajnoczi Nov. 29, 2023, 7:55 p.m. UTC
Protect the Task Management Function BH state with a lock. The TMF BH
runs in the main loop thread. An IOThread might process a TMF at the
same time as the TMF BH is running. Therefore tmf_bh_list and tmf_bh
must be protected by a lock.

Run TMF request completion in the IOThread using aio_wait_bh_oneshot().
This avoids more locking to protect the virtqueue and SCSI layer state.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 include/hw/virtio/virtio-scsi.h |  3 +-
 hw/scsi/virtio-scsi.c           | 62 ++++++++++++++++++++++-----------
 2 files changed, 43 insertions(+), 22 deletions(-)

Comments

Eric Blake Nov. 30, 2023, 3:25 p.m. UTC | #1
On Wed, Nov 29, 2023 at 02:55:42PM -0500, Stefan Hajnoczi wrote:
> Protect the Task Management Function BH state with a lock. The TMF BH
> runs in the main loop thread. An IOThread might process a TMF at the
> same time as the TMF BH is running. Therefore tmf_bh_list and tmf_bh
> must be protected by a lock.
> 
> Run TMF request completion in the IOThread using aio_wait_bh_oneshot().
> This avoids more locking to protect the virtqueue and SCSI layer state.

Are we trying to get this into 8.2?

> 
> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
> ---
>  include/hw/virtio/virtio-scsi.h |  3 +-
>  hw/scsi/virtio-scsi.c           | 62 ++++++++++++++++++++++-----------
>  2 files changed, 43 insertions(+), 22 deletions(-)
>

Reviewed-by: Eric Blake <eblake@redhat.com>
Kevin Wolf Dec. 4, 2023, 12:35 p.m. UTC | #2
Am 29.11.2023 um 20:55 hat Stefan Hajnoczi geschrieben:
> Protect the Task Management Function BH state with a lock. The TMF BH
> runs in the main loop thread. An IOThread might process a TMF at the
> same time as the TMF BH is running. Therefore tmf_bh_list and tmf_bh
> must be protected by a lock.
> 
> Run TMF request completion in the IOThread using aio_wait_bh_oneshot().
> This avoids more locking to protect the virtqueue and SCSI layer state.
> 
> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Kevin Wolf Dec. 4, 2023, 12:46 p.m. UTC | #3
Am 29.11.2023 um 20:55 hat Stefan Hajnoczi geschrieben:
> Protect the Task Management Function BH state with a lock. The TMF BH
> runs in the main loop thread. An IOThread might process a TMF at the
> same time as the TMF BH is running. Therefore tmf_bh_list and tmf_bh
> must be protected by a lock.
> 
> Run TMF request completion in the IOThread using aio_wait_bh_oneshot().
> This avoids more locking to protect the virtqueue and SCSI layer state.
> 
> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

The second part reminds me that the implicit protection of the virtqueue
and SCSI data structures by having all accesses in a single thread is
hard to review and I think we wanted to put some assertions there to
check that we're really running in the right thread. I don't think we
have done that so far, so I suppose after this patch would be the place
in the series to add them, before we remove the protection by the
AioContext lock?

Kevin
Stefan Hajnoczi Dec. 4, 2023, 2:49 p.m. UTC | #4
On Thu, Nov 30, 2023 at 09:25:52AM -0600, Eric Blake wrote:
> On Wed, Nov 29, 2023 at 02:55:42PM -0500, Stefan Hajnoczi wrote:
> > Protect the Task Management Function BH state with a lock. The TMF BH
> > runs in the main loop thread. An IOThread might process a TMF at the
> > same time as the TMF BH is running. Therefore tmf_bh_list and tmf_bh
> > must be protected by a lock.
> > 
> > Run TMF request completion in the IOThread using aio_wait_bh_oneshot().
> > This avoids more locking to protect the virtqueue and SCSI layer state.
> 
> Are we trying to get this into 8.2?

No. 8.2 still has the AioContext lock is therefore safe without this
patch.

Stefan

> 
> > 
> > Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
> > ---
> >  include/hw/virtio/virtio-scsi.h |  3 +-
> >  hw/scsi/virtio-scsi.c           | 62 ++++++++++++++++++++++-----------
> >  2 files changed, 43 insertions(+), 22 deletions(-)
> >
> 
> Reviewed-by: Eric Blake <eblake@redhat.com>
> 
> -- 
> Eric Blake, Principal Software Engineer
> Red Hat, Inc.
> Virtualization:  qemu.org | libguestfs.org
>
Stefan Hajnoczi Dec. 4, 2023, 2:51 p.m. UTC | #5
On Mon, Dec 04, 2023 at 01:46:13PM +0100, Kevin Wolf wrote:
> Am 29.11.2023 um 20:55 hat Stefan Hajnoczi geschrieben:
> > Protect the Task Management Function BH state with a lock. The TMF BH
> > runs in the main loop thread. An IOThread might process a TMF at the
> > same time as the TMF BH is running. Therefore tmf_bh_list and tmf_bh
> > must be protected by a lock.
> > 
> > Run TMF request completion in the IOThread using aio_wait_bh_oneshot().
> > This avoids more locking to protect the virtqueue and SCSI layer state.
> > 
> > Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
> 
> The second part reminds me that the implicit protection of the virtqueue
> and SCSI data structures by having all accesses in a single thread is
> hard to review and I think we wanted to put some assertions there to
> check that we're really running in the right thread. I don't think we
> have done that so far, so I suppose after this patch would be the place
> in the series to add them, before we remove the protection by the
> AioContext lock?

Thanks for reminding me. I will add assertions in the next revision of
this series.

Stefan
diff mbox series

Patch

diff --git a/include/hw/virtio/virtio-scsi.h b/include/hw/virtio/virtio-scsi.h
index 779568ab5d..da8cb928d9 100644
--- a/include/hw/virtio/virtio-scsi.h
+++ b/include/hw/virtio/virtio-scsi.h
@@ -85,8 +85,9 @@  struct VirtIOSCSI {
 
     /*
      * TMFs deferred to main loop BH. These fields are protected by
-     * virtio_scsi_acquire().
+     * tmf_bh_lock.
      */
+    QemuMutex tmf_bh_lock;
     QEMUBH *tmf_bh;
     QTAILQ_HEAD(, VirtIOSCSIReq) tmf_bh_list;
 
diff --git a/hw/scsi/virtio-scsi.c b/hw/scsi/virtio-scsi.c
index 9c751bf296..4f8d35facc 100644
--- a/hw/scsi/virtio-scsi.c
+++ b/hw/scsi/virtio-scsi.c
@@ -123,6 +123,30 @@  static void virtio_scsi_complete_req(VirtIOSCSIReq *req)
     virtio_scsi_free_req(req);
 }
 
+static void virtio_scsi_complete_req_bh(void *opaque)
+{
+    VirtIOSCSIReq *req = opaque;
+
+    virtio_scsi_complete_req(req);
+}
+
+/*
+ * Called from virtio_scsi_do_one_tmf_bh() in main loop thread. The main loop
+ * thread cannot touch the virtqueue since that could race with an IOThread.
+ */
+static void virtio_scsi_complete_req_from_main_loop(VirtIOSCSIReq *req)
+{
+    VirtIOSCSI *s = req->dev;
+
+    if (!s->ctx || s->ctx == qemu_get_aio_context()) {
+        /* No need to schedule a BH when there is no IOThread */
+        virtio_scsi_complete_req(req);
+    } else {
+        /* Run request completion in the IOThread */
+        aio_wait_bh_oneshot(s->ctx, virtio_scsi_complete_req_bh, req);
+    }
+}
+
 static void virtio_scsi_bad_req(VirtIOSCSIReq *req)
 {
     virtio_error(VIRTIO_DEVICE(req->dev), "wrong size for virtio-scsi headers");
@@ -338,10 +362,7 @@  static void virtio_scsi_do_one_tmf_bh(VirtIOSCSIReq *req)
 
 out:
     object_unref(OBJECT(d));
-
-    virtio_scsi_acquire(s);
-    virtio_scsi_complete_req(req);
-    virtio_scsi_release(s);
+    virtio_scsi_complete_req_from_main_loop(req);
 }
 
 /* Some TMFs must be processed from the main loop thread */
@@ -354,18 +375,16 @@  static void virtio_scsi_do_tmf_bh(void *opaque)
 
     GLOBAL_STATE_CODE();
 
-    virtio_scsi_acquire(s);
+    WITH_QEMU_LOCK_GUARD(&s->tmf_bh_lock) {
+        QTAILQ_FOREACH_SAFE(req, &s->tmf_bh_list, next, tmp) {
+            QTAILQ_REMOVE(&s->tmf_bh_list, req, next);
+            QTAILQ_INSERT_TAIL(&reqs, req, next);
+        }
 
-    QTAILQ_FOREACH_SAFE(req, &s->tmf_bh_list, next, tmp) {
-        QTAILQ_REMOVE(&s->tmf_bh_list, req, next);
-        QTAILQ_INSERT_TAIL(&reqs, req, next);
+        qemu_bh_delete(s->tmf_bh);
+        s->tmf_bh = NULL;
     }
 
-    qemu_bh_delete(s->tmf_bh);
-    s->tmf_bh = NULL;
-
-    virtio_scsi_release(s);
-
     QTAILQ_FOREACH_SAFE(req, &reqs, next, tmp) {
         QTAILQ_REMOVE(&reqs, req, next);
         virtio_scsi_do_one_tmf_bh(req);
@@ -379,8 +398,7 @@  static void virtio_scsi_reset_tmf_bh(VirtIOSCSI *s)
 
     GLOBAL_STATE_CODE();
 
-    virtio_scsi_acquire(s);
-
+    /* Called after ioeventfd has been stopped, so tmf_bh_lock is not needed */
     if (s->tmf_bh) {
         qemu_bh_delete(s->tmf_bh);
         s->tmf_bh = NULL;
@@ -393,19 +411,19 @@  static void virtio_scsi_reset_tmf_bh(VirtIOSCSI *s)
         req->resp.tmf.response = VIRTIO_SCSI_S_TARGET_FAILURE;
         virtio_scsi_complete_req(req);
     }
-
-    virtio_scsi_release(s);
 }
 
 static void virtio_scsi_defer_tmf_to_bh(VirtIOSCSIReq *req)
 {
     VirtIOSCSI *s = req->dev;
 
-    QTAILQ_INSERT_TAIL(&s->tmf_bh_list, req, next);
+    WITH_QEMU_LOCK_GUARD(&s->tmf_bh_lock) {
+        QTAILQ_INSERT_TAIL(&s->tmf_bh_list, req, next);
 
-    if (!s->tmf_bh) {
-        s->tmf_bh = qemu_bh_new(virtio_scsi_do_tmf_bh, s);
-        qemu_bh_schedule(s->tmf_bh);
+        if (!s->tmf_bh) {
+            s->tmf_bh = qemu_bh_new(virtio_scsi_do_tmf_bh, s);
+            qemu_bh_schedule(s->tmf_bh);
+        }
     }
 }
 
@@ -1235,6 +1253,7 @@  static void virtio_scsi_device_realize(DeviceState *dev, Error **errp)
     Error *err = NULL;
 
     QTAILQ_INIT(&s->tmf_bh_list);
+    qemu_mutex_init(&s->tmf_bh_lock);
 
     virtio_scsi_common_realize(dev,
                                virtio_scsi_handle_ctrl,
@@ -1277,6 +1296,7 @@  static void virtio_scsi_device_unrealize(DeviceState *dev)
 
     qbus_set_hotplug_handler(BUS(&s->bus), NULL);
     virtio_scsi_common_unrealize(dev);
+    qemu_mutex_destroy(&s->tmf_bh_lock);
 }
 
 static Property virtio_scsi_properties[] = {