[PATCH for 8.1 v2 0/6] Move ASID test to vhost-vdpa net initialization

[PATCH for 8.1 v2 0/6] Move ASID test to vhost-vdpa net initialization

[Eugenio Pérez]

QEMU v8.0.0-rc0 is able to switch dynamically between vhost-vdpa passthrough
and SVQ mode as long as the net device does not have CVQ.  The net device
state followed (and migrated) by CVQ requires special care.

A pre-requisite to add CVQ to that framework is to determine if devices with
CVQ are migratable or not at initialization time.  The solution to it is to
always shadow only CVQ, and vq groups and ASID are used for that.

However, current qemu version only checks ASID at device start (as "driver set
DRIVER_OK status bit"), not at device initialization.  A check at
initialization time is required.  Otherwise, the guest would be able to set
and remove migration blockers at will [1].

This series is a requisite for migration of vhost-vdpa net devices with CVQ.
However it already makes sense by its own, as it reduces the number of ioctls
at migration time, decreasing the error paths there.

[1] https://lore.kernel.org/qemu-devel/2616f0cd-f9e8-d183-ea78-db1be4825d9c@redhat.com/
---
v2:
* Take out the reset of the device from vhost_vdpa_cvq_is_isolated
  (reported by Lei Yang).
* Expand patch messages by Stefano G. questions.

Eugenio Pérez (6):
  vdpa: Remove status in reset tracing
  vdpa: add vhost_vdpa_reset_status_fd
  vdpa: add vhost_vdpa_set_dev_features_fd
  vdpa: return errno in vhost_vdpa_get_vring_group error
  vdpa: move CVQ isolation check to net_init_vhost_vdpa
  vdpa: Cache cvq group in VhostVDPAState

 include/hw/virtio/vhost-vdpa.h |   2 +
 hw/virtio/vhost-vdpa.c         |  78 ++++++++-----
 net/vhost-vdpa.c               | 199 +++++++++++++++++++++++++++------
 hw/virtio/trace-events         |   2 +-
 4 files changed, 221 insertions(+), 60 deletions(-)

-- 
2.31.1

[PATCH for 8.1 v2 1/6] vdpa: Remove status in reset tracing

[Eugenio Pérez]

It is always 0 and it is not useful to route call through file
descriptor.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
---
 hw/virtio/vhost-vdpa.c | 2 +-
 hw/virtio/trace-events | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c
index bc6bad23d5..bbabea18f3 100644
--- a/hw/virtio/vhost-vdpa.c
+++ b/hw/virtio/vhost-vdpa.c
@@ -716,7 +716,7 @@ static int vhost_vdpa_reset_device(struct vhost_dev *dev)
     uint8_t status = 0;
 
     ret = vhost_vdpa_call(dev, VHOST_VDPA_SET_STATUS, &status);
-    trace_vhost_vdpa_reset_device(dev, status);
+    trace_vhost_vdpa_reset_device(dev);
     v->suspended = false;
     return ret;
 }
diff --git a/hw/virtio/trace-events b/hw/virtio/trace-events
index 8f8d05cf9b..6265231683 100644
--- a/hw/virtio/trace-events
+++ b/hw/virtio/trace-events
@@ -44,7 +44,7 @@ vhost_vdpa_set_mem_table(void *dev, uint32_t nregions, uint32_t padding) "dev: %
 vhost_vdpa_dump_regions(void *dev, int i, uint64_t guest_phys_addr, uint64_t memory_size, uint64_t userspace_addr, uint64_t flags_padding) "dev: %p %d: guest_phys_addr: 0x%"PRIx64" memory_size: 0x%"PRIx64" userspace_addr: 0x%"PRIx64" flags_padding: 0x%"PRIx64
 vhost_vdpa_set_features(void *dev, uint64_t features) "dev: %p features: 0x%"PRIx64
 vhost_vdpa_get_device_id(void *dev, uint32_t device_id) "dev: %p device_id %"PRIu32
-vhost_vdpa_reset_device(void *dev, uint8_t status) "dev: %p status: 0x%"PRIx8
+vhost_vdpa_reset_device(void *dev) "dev: %p"
 vhost_vdpa_get_vq_index(void *dev, int idx, int vq_idx) "dev: %p idx: %d vq idx: %d"
 vhost_vdpa_set_vring_ready(void *dev) "dev: %p"
 vhost_vdpa_dump_config(void *dev, const char *line) "dev: %p %s"
-- 
2.31.1

[PATCH for 8.1 v2 2/6] vdpa: add vhost_vdpa_reset_status_fd

[Eugenio Pérez]

This allows to reset a vhost-vdpa device from external subsystems like
vhost-net, since it does not have any struct vhost_dev by the time we
need to use it.

It is used in subsequent patches to negotiate features
and probe for CVQ ASID isolation.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
---
 include/hw/virtio/vhost-vdpa.h |  1 +
 hw/virtio/vhost-vdpa.c         | 58 +++++++++++++++++++++++-----------
 2 files changed, 41 insertions(+), 18 deletions(-)

diff --git a/include/hw/virtio/vhost-vdpa.h b/include/hw/virtio/vhost-vdpa.h
index c278a2a8de..28de7da91e 100644
--- a/include/hw/virtio/vhost-vdpa.h
+++ b/include/hw/virtio/vhost-vdpa.h
@@ -54,6 +54,7 @@ typedef struct vhost_vdpa {
     VhostVDPAHostNotifier notifier[VIRTIO_QUEUE_MAX];
 } VhostVDPA;
 
+void vhost_vdpa_reset_status_fd(int fd);
 int vhost_vdpa_get_iova_range(int fd, struct vhost_vdpa_iova_range *iova_range);
 
 int vhost_vdpa_dma_map(struct vhost_vdpa *v, uint32_t asid, hwaddr iova,
diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c
index bbabea18f3..7a2053b8d9 100644
--- a/hw/virtio/vhost-vdpa.c
+++ b/hw/virtio/vhost-vdpa.c
@@ -335,38 +335,45 @@ static const MemoryListener vhost_vdpa_memory_listener = {
     .region_del = vhost_vdpa_listener_region_del,
 };
 
-static int vhost_vdpa_call(struct vhost_dev *dev, unsigned long int request,
-                             void *arg)
+static int vhost_vdpa_dev_fd(const struct vhost_dev *dev)
 {
     struct vhost_vdpa *v = dev->opaque;
-    int fd = v->device_fd;
-    int ret;
 
     assert(dev->vhost_ops->backend_type == VHOST_BACKEND_TYPE_VDPA);
+    return v->device_fd;
+}
+
+static int vhost_vdpa_call_fd(int fd, unsigned long int request, void *arg)
+{
+    int ret = ioctl(fd, request, arg);
 
-    ret = ioctl(fd, request, arg);
     return ret < 0 ? -errno : ret;
 }
 
-static int vhost_vdpa_add_status(struct vhost_dev *dev, uint8_t status)
+static int vhost_vdpa_call(struct vhost_dev *dev, unsigned long int request,
+                           void *arg)
+{
+    return vhost_vdpa_call_fd(vhost_vdpa_dev_fd(dev), request, arg);
+}
+
+static int vhost_vdpa_add_status_fd(int fd, uint8_t status)
 {
     uint8_t s;
     int ret;
 
-    trace_vhost_vdpa_add_status(dev, status);
-    ret = vhost_vdpa_call(dev, VHOST_VDPA_GET_STATUS, &s);
+    ret = vhost_vdpa_call_fd(fd, VHOST_VDPA_GET_STATUS, &s);
     if (ret < 0) {
         return ret;
     }
 
     s |= status;
 
-    ret = vhost_vdpa_call(dev, VHOST_VDPA_SET_STATUS, &s);
+    ret = vhost_vdpa_call_fd(fd, VHOST_VDPA_SET_STATUS, &s);
     if (ret < 0) {
         return ret;
     }
 
-    ret = vhost_vdpa_call(dev, VHOST_VDPA_GET_STATUS, &s);
+    ret = vhost_vdpa_call_fd(fd, VHOST_VDPA_GET_STATUS, &s);
     if (ret < 0) {
         return ret;
     }
@@ -378,6 +385,12 @@ static int vhost_vdpa_add_status(struct vhost_dev *dev, uint8_t status)
     return 0;
 }
 
+static int vhost_vdpa_add_status(struct vhost_dev *dev, uint8_t status)
+{
+    trace_vhost_vdpa_add_status(dev, status);
+    return vhost_vdpa_add_status_fd(vhost_vdpa_dev_fd(dev), status);
+}
+
 int vhost_vdpa_get_iova_range(int fd, struct vhost_vdpa_iova_range *iova_range)
 {
     int ret = ioctl(fd, VHOST_VDPA_GET_IOVA_RANGE, iova_range);
@@ -709,16 +722,20 @@ static int vhost_vdpa_get_device_id(struct vhost_dev *dev,
     return ret;
 }
 
+static int vhost_vdpa_reset_device_fd(int fd)
+{
+    uint8_t status = 0;
+
+    return vhost_vdpa_call_fd(fd, VHOST_VDPA_SET_STATUS, &status);
+}
+
 static int vhost_vdpa_reset_device(struct vhost_dev *dev)
 {
     struct vhost_vdpa *v = dev->opaque;
-    int ret;
-    uint8_t status = 0;
 
-    ret = vhost_vdpa_call(dev, VHOST_VDPA_SET_STATUS, &status);
-    trace_vhost_vdpa_reset_device(dev);
     v->suspended = false;
-    return ret;
+    trace_vhost_vdpa_reset_device(dev);
+    return vhost_vdpa_reset_device_fd(vhost_vdpa_dev_fd(dev));
 }
 
 static int vhost_vdpa_get_vq_index(struct vhost_dev *dev, int idx)
@@ -1170,6 +1187,13 @@ static int vhost_vdpa_dev_start(struct vhost_dev *dev, bool started)
     return 0;
 }
 
+void vhost_vdpa_reset_status_fd(int fd)
+{
+    vhost_vdpa_reset_device_fd(fd);
+    vhost_vdpa_add_status_fd(fd, VIRTIO_CONFIG_S_ACKNOWLEDGE |
+                                 VIRTIO_CONFIG_S_DRIVER);
+}
+
 static void vhost_vdpa_reset_status(struct vhost_dev *dev)
 {
     struct vhost_vdpa *v = dev->opaque;
@@ -1178,9 +1202,7 @@ static void vhost_vdpa_reset_status(struct vhost_dev *dev)
         return;
     }
 
-    vhost_vdpa_reset_device(dev);
-    vhost_vdpa_add_status(dev, VIRTIO_CONFIG_S_ACKNOWLEDGE |
-                               VIRTIO_CONFIG_S_DRIVER);
+    vhost_vdpa_reset_status_fd(vhost_vdpa_dev_fd(dev));
     memory_listener_unregister(&v->listener);
 }
 
-- 
2.31.1

[PATCH for 8.1 v2 3/6] vdpa: add vhost_vdpa_set_dev_features_fd

[Eugenio Pérez]

This allows to set the features of a vhost-vdpa device from external
subsystems like vhost-net.  It is used in subsequent patches to
negotiate features and probe for CVQ ASID isolation.

Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
---
 include/hw/virtio/vhost-vdpa.h |  1 +
 hw/virtio/vhost-vdpa.c         | 20 +++++++++++++-------
 2 files changed, 14 insertions(+), 7 deletions(-)

diff --git a/include/hw/virtio/vhost-vdpa.h b/include/hw/virtio/vhost-vdpa.h
index 28de7da91e..a9cb6f3a32 100644
--- a/include/hw/virtio/vhost-vdpa.h
+++ b/include/hw/virtio/vhost-vdpa.h
@@ -55,6 +55,7 @@ typedef struct vhost_vdpa {
 } VhostVDPA;
 
 void vhost_vdpa_reset_status_fd(int fd);
+int vhost_vdpa_set_dev_features_fd(int fd, uint64_t features);
 int vhost_vdpa_get_iova_range(int fd, struct vhost_vdpa_iova_range *iova_range);
 
 int vhost_vdpa_dma_map(struct vhost_vdpa *v, uint32_t asid, hwaddr iova,
diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c
index 7a2053b8d9..acd5be46a9 100644
--- a/hw/virtio/vhost-vdpa.c
+++ b/hw/virtio/vhost-vdpa.c
@@ -651,11 +651,22 @@ static int vhost_vdpa_set_mem_table(struct vhost_dev *dev,
     return 0;
 }
 
+int vhost_vdpa_set_dev_features_fd(int fd, uint64_t features)
+{
+    int ret;
+
+    ret = vhost_vdpa_call_fd(fd, VHOST_SET_FEATURES, &features);
+    if (ret) {
+        return ret;
+    }
+
+    return vhost_vdpa_add_status_fd(fd, VIRTIO_CONFIG_S_FEATURES_OK);
+}
+
 static int vhost_vdpa_set_features(struct vhost_dev *dev,
                                    uint64_t features)
 {
     struct vhost_vdpa *v = dev->opaque;
-    int ret;
 
     if (!vhost_vdpa_first_dev(dev)) {
         return 0;
@@ -678,12 +689,7 @@ static int vhost_vdpa_set_features(struct vhost_dev *dev,
     }
 
     trace_vhost_vdpa_set_features(dev, features);
-    ret = vhost_vdpa_call(dev, VHOST_SET_FEATURES, &features);
-    if (ret) {
-        return ret;
-    }
-
-    return vhost_vdpa_add_status(dev, VIRTIO_CONFIG_S_FEATURES_OK);
+    return vhost_vdpa_set_dev_features_fd(vhost_vdpa_dev_fd(dev), features);
 }
 
 static int vhost_vdpa_set_backend_cap(struct vhost_dev *dev)
-- 
2.31.1

[PATCH for 8.1 v2 4/6] vdpa: return errno in vhost_vdpa_get_vring_group error

[Eugenio Pérez]

We need to tell in the caller, as some errors are expected in a normal
workflow.  In particular, parent drivers in recent kernels with
VHOST_BACKEND_F_IOTLB_ASID may not support vring groups.  In that case,
-ENOTSUP is returned.

This is the case of vp_vdpa in Linux 6.2.

Next patches in this series will use that information to know if it must
abort or not.  Also, next patches return properly an errp instead of
printing with error_report.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
---
 net/vhost-vdpa.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c
index 99904a0da7..4397c0d4b3 100644
--- a/net/vhost-vdpa.c
+++ b/net/vhost-vdpa.c
@@ -361,6 +361,14 @@ static NetClientInfo net_vhost_vdpa_info = {
         .check_peer_type = vhost_vdpa_check_peer_type,
 };
 
+/**
+ * Get vring virtqueue group
+ *
+ * @device_fd  vdpa device fd
+ * @vq_index   Virtqueue index
+ *
+ * Return -errno in case of error, or vq group if success.
+ */
 static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index)
 {
     struct vhost_vring_state state = {
@@ -369,6 +377,7 @@ static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index)
     int r = ioctl(device_fd, VHOST_VDPA_GET_VRING_GROUP, &state);
 
     if (unlikely(r < 0)) {
+        r = -errno;
         error_report("Cannot get VQ %u group: %s", vq_index,
                      g_strerror(errno));
         return r;
-- 
2.31.1

[PATCH for 8.1 v2 5/6] vdpa: move CVQ isolation check to net_init_vhost_vdpa

[Eugenio Pérez]

Evaluating it at start time instead of initialization time may make the
guest capable of dynamically adding or removing migration blockers.

Also, moving to initialization reduces the number of ioctls in the
migration, reducing failure possibilities.

As a drawback we need to check for CVQ isolation twice: one time with no
MQ negotiated and another one acking it, as long as the device supports
it.  This is because Vring ASID / group management is based on vq
indexes, but we don't know the index of CVQ before negotiating MQ.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
---
v2: Take out the reset of the device from vhost_vdpa_cvq_is_isolated
---
 net/vhost-vdpa.c | 194 ++++++++++++++++++++++++++++++++++++-----------
 1 file changed, 151 insertions(+), 43 deletions(-)

diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c
index 4397c0d4b3..db2c9afcb3 100644
--- a/net/vhost-vdpa.c
+++ b/net/vhost-vdpa.c
@@ -43,6 +43,13 @@ typedef struct VhostVDPAState {
 
     /* The device always have SVQ enabled */
     bool always_svq;
+
+    /* The device can isolate CVQ in its own ASID if MQ is negotiated */
+    bool cvq_isolated_mq;
+
+    /* The device can isolate CVQ in its own ASID if MQ is not negotiated */
+    bool cvq_isolated;
+
     bool started;
 } VhostVDPAState;
 
@@ -361,15 +368,8 @@ static NetClientInfo net_vhost_vdpa_info = {
         .check_peer_type = vhost_vdpa_check_peer_type,
 };
 
-/**
- * Get vring virtqueue group
- *
- * @device_fd  vdpa device fd
- * @vq_index   Virtqueue index
- *
- * Return -errno in case of error, or vq group if success.
- */
-static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index)
+static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index,
+                                          Error **errp)
 {
     struct vhost_vring_state state = {
         .index = vq_index,
@@ -378,8 +378,7 @@ static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index)
 
     if (unlikely(r < 0)) {
         r = -errno;
-        error_report("Cannot get VQ %u group: %s", vq_index,
-                     g_strerror(errno));
+        error_setg_errno(errp, errno, "Cannot get VQ %u group", vq_index);
         return r;
     }
 
@@ -479,9 +478,9 @@ static int vhost_vdpa_net_cvq_start(NetClientState *nc)
 {
     VhostVDPAState *s, *s0;
     struct vhost_vdpa *v;
-    uint64_t backend_features;
     int64_t cvq_group;
-    int cvq_index, r;
+    int r;
+    Error *err = NULL;
 
     assert(nc->info->type == NET_CLIENT_DRIVER_VHOST_VDPA);
 
@@ -501,42 +500,29 @@ static int vhost_vdpa_net_cvq_start(NetClientState *nc)
     /*
      * If we early return in these cases SVQ will not be enabled. The migration
      * will be blocked as long as vhost-vdpa backends will not offer _F_LOG.
-     *
-     * Calling VHOST_GET_BACKEND_FEATURES as they are not available in v->dev
-     * yet.
      */
-    r = ioctl(v->device_fd, VHOST_GET_BACKEND_FEATURES, &backend_features);
-    if (unlikely(r < 0)) {
-        error_report("Cannot get vdpa backend_features: %s(%d)",
-            g_strerror(errno), errno);
-        return -1;
-    }
-    if (!(backend_features & BIT_ULL(VHOST_BACKEND_F_IOTLB_ASID)) ||
-        !vhost_vdpa_net_valid_svq_features(v->dev->features, NULL)) {
+    if (!vhost_vdpa_net_valid_svq_features(v->dev->features, NULL)) {
         return 0;
     }
 
-    /*
-     * Check if all the virtqueues of the virtio device are in a different vq
-     * than the last vq. VQ group of last group passed in cvq_group.
-     */
-    cvq_index = v->dev->vq_index_end - 1;
-    cvq_group = vhost_vdpa_get_vring_group(v->device_fd, cvq_index);
-    if (unlikely(cvq_group < 0)) {
-        return cvq_group;
-    }
-    for (int i = 0; i < cvq_index; ++i) {
-        int64_t group = vhost_vdpa_get_vring_group(v->device_fd, i);
-
-        if (unlikely(group < 0)) {
-            return group;
+    if (v->dev->features & BIT_ULL(VIRTIO_NET_F_MQ)) {
+        if (!s->cvq_isolated_mq) {
+            return 0;
         }
-
-        if (group == cvq_group) {
+    } else {
+        if (!s->cvq_isolated) {
             return 0;
         }
     }
 
+    cvq_group = vhost_vdpa_get_vring_group(v->device_fd,
+                                           v->dev->vq_index_end - 1,
+                                           &err);
+    if (unlikely(cvq_group < 0)) {
+        error_report_err(err);
+        return cvq_group;
+    }
+
     r = vhost_vdpa_set_address_space_id(v, cvq_group, VHOST_VDPA_NET_CVQ_ASID);
     if (unlikely(r < 0)) {
         return r;
@@ -798,6 +784,116 @@ static const VhostShadowVirtqueueOps vhost_vdpa_net_svq_ops = {
     .avail_handler = vhost_vdpa_net_handle_ctrl_avail,
 };
 
+/**
+ * Probe the device to check control virtqueue is isolated.
+ *
+ * @device_fd vhost-vdpa file descriptor
+ * @features features to negotiate
+ * @cvq_index Control vq index
+ *
+ * Returns -1 in case of error, 0 if false and 1 if true
+ */
+static int vhost_vdpa_cvq_is_isolated(int device_fd, uint64_t features,
+                                      unsigned cvq_index, Error **errp)
+{
+    int64_t cvq_group;
+    int r;
+
+    r = vhost_vdpa_set_dev_features_fd(device_fd, features);
+    if (unlikely(r < 0)) {
+        error_setg_errno(errp, -r, "Cannot set device features");
+        return r;
+    }
+
+    cvq_group = vhost_vdpa_get_vring_group(device_fd, cvq_index, errp);
+    if (unlikely(cvq_group < 0)) {
+        return cvq_group;
+    }
+
+    for (int i = 0; i < cvq_index; ++i) {
+        int64_t group = vhost_vdpa_get_vring_group(device_fd, i, errp);
+
+        if (unlikely(group < 0)) {
+            return group;
+        }
+
+        if (group == (int64_t)cvq_group) {
+            return 0;
+        }
+    }
+
+    return 1;
+}
+
+/**
+ * Probe if CVQ is isolated when the device is MQ and when it is not MQ
+ *
+ * @device_fd         The vdpa device fd
+ * @features          Features offered by the device.
+ * @cvq_index         The control vq index if mq is negotiated. Ignored
+ *                    otherwise.
+ * @cvq_isolated      It'll be set to true if cvq is isolated if mq is not
+ *                    negotiated.
+ * @cvq_isolated_mq   It'll be set to true if cvq is isolated if mq is
+ *                    negotiated.
+ *
+ * Returns -1 in case of failure
+ */
+static int vhost_vdpa_probe_cvq_isolation(int device_fd, uint64_t features,
+                                          int cvq_index, bool *cvq_isolated,
+                                          bool *cvq_isolated_mq, Error **errp)
+{
+    uint64_t backend_features;
+    int r;
+
+    ERRP_GUARD();
+
+    *cvq_isolated = false;
+    *cvq_isolated_mq = false;
+    r = ioctl(device_fd, VHOST_GET_BACKEND_FEATURES, &backend_features);
+    if (unlikely(r < 0)) {
+        error_setg_errno(errp, errno, "Cannot get vdpa backend_features");
+        return r;
+    }
+
+    if (!(backend_features & BIT_ULL(VHOST_BACKEND_F_IOTLB_ASID))) {
+        return 0;
+    }
+
+    r = vhost_vdpa_cvq_is_isolated(device_fd,
+                                   features & ~BIT_ULL(VIRTIO_NET_F_MQ), 2,
+                                   errp);
+    if (unlikely(r < 0)) {
+        if (r == -ENOTSUP) {
+            /*
+             * The kernel report VHOST_BACKEND_F_IOTLB_ASID if the vdpa
+             * frontend support ASID but the parent driver does not.  The CVQ
+             * cannot be isolated in this case.
+             */
+            error_free(*errp);
+            *errp = NULL;
+            return 0;
+        }
+
+        return r;
+    }
+
+    *cvq_isolated = r == 1;
+    vhost_vdpa_reset_status_fd(device_fd);
+    if (!(features & BIT_ULL(VIRTIO_NET_F_MQ))) {
+        return 0;
+    }
+
+    r = vhost_vdpa_cvq_is_isolated(device_fd, features, cvq_index * 2, errp);
+    if (unlikely(r < 0)) {
+        return r;
+    }
+
+    *cvq_isolated_mq = r == 1;
+    vhost_vdpa_reset_status_fd(device_fd);
+    return 0;
+}
+
 static NetClientState *net_vhost_vdpa_init(NetClientState *peer,
                                        const char *device,
                                        const char *name,
@@ -807,16 +903,26 @@ static NetClientState *net_vhost_vdpa_init(NetClientState *peer,
                                        bool is_datapath,
                                        bool svq,
                                        struct vhost_vdpa_iova_range iova_range,
-                                       uint64_t features)
+                                       uint64_t features,
+                                       Error **errp)
 {
     NetClientState *nc = NULL;
     VhostVDPAState *s;
     int ret = 0;
     assert(name);
+    bool cvq_isolated, cvq_isolated_mq;
+
     if (is_datapath) {
         nc = qemu_new_net_client(&net_vhost_vdpa_info, peer, device,
                                  name);
     } else {
+        ret = vhost_vdpa_probe_cvq_isolation(vdpa_device_fd, features,
+                                             queue_pair_index, &cvq_isolated,
+                                             &cvq_isolated_mq, errp);
+        if (unlikely(ret)) {
+            return NULL;
+        }
+
         nc = qemu_new_net_control_client(&net_vhost_vdpa_cvq_info, peer,
                                          device, name);
     }
@@ -843,6 +949,8 @@ static NetClientState *net_vhost_vdpa_init(NetClientState *peer,
 
         s->vhost_vdpa.shadow_vq_ops = &vhost_vdpa_net_svq_ops;
         s->vhost_vdpa.shadow_vq_ops_opaque = s;
+        s->cvq_isolated = cvq_isolated;
+        s->cvq_isolated_mq = cvq_isolated_mq;
 
         /*
          * TODO: We cannot migrate devices with CVQ as there is no way to set
@@ -971,7 +1079,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name,
     for (i = 0; i < queue_pairs; i++) {
         ncs[i] = net_vhost_vdpa_init(peer, TYPE_VHOST_VDPA, name,
                                      vdpa_device_fd, i, 2, true, opts->x_svq,
-                                     iova_range, features);
+                                     iova_range, features, errp);
         if (!ncs[i])
             goto err;
     }
@@ -979,7 +1087,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name,
     if (has_cvq) {
         nc = net_vhost_vdpa_init(peer, TYPE_VHOST_VDPA, name,
                                  vdpa_device_fd, i, 1, false,
-                                 opts->x_svq, iova_range, features);
+                                 opts->x_svq, iova_range, features, errp);
         if (!nc)
             goto err;
     }
-- 
2.31.1

[PATCH for 8.1 v2 6/6] vdpa: Cache cvq group in VhostVDPAState

[Eugenio Pérez]

Continue the move of code that interacts with the device from control
virtqueue start to control virtqueue init.

As with previous patches, it reduces the number of ioctls in the
migration, reducing failure possibilities.

Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
---
 net/vhost-vdpa.c | 40 +++++++++++++++++++++++++++-------------
 1 file changed, 27 insertions(+), 13 deletions(-)

diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c
index db2c9afcb3..6a60e8cc2b 100644
--- a/net/vhost-vdpa.c
+++ b/net/vhost-vdpa.c
@@ -41,6 +41,12 @@ typedef struct VhostVDPAState {
     void *cvq_cmd_out_buffer;
     virtio_net_ctrl_ack *status;
 
+    /* CVQ group if cvq_isolated_mq */
+    uint32_t cvq_group_mq;
+
+    /* CVQ group if cvq_isolated */
+    uint32_t cvq_group;
+
     /* The device always have SVQ enabled */
     bool always_svq;
 
@@ -480,7 +486,6 @@ static int vhost_vdpa_net_cvq_start(NetClientState *nc)
     struct vhost_vdpa *v;
     int64_t cvq_group;
     int r;
-    Error *err = NULL;
 
     assert(nc->info->type == NET_CLIENT_DRIVER_VHOST_VDPA);
 
@@ -509,18 +514,14 @@ static int vhost_vdpa_net_cvq_start(NetClientState *nc)
         if (!s->cvq_isolated_mq) {
             return 0;
         }
+
+        cvq_group = s->cvq_group_mq;
     } else {
         if (!s->cvq_isolated) {
             return 0;
         }
-    }
 
-    cvq_group = vhost_vdpa_get_vring_group(v->device_fd,
-                                           v->dev->vq_index_end - 1,
-                                           &err);
-    if (unlikely(cvq_group < 0)) {
-        error_report_err(err);
-        return cvq_group;
+        cvq_group = s->cvq_group;
     }
 
     r = vhost_vdpa_set_address_space_id(v, cvq_group, VHOST_VDPA_NET_CVQ_ASID);
@@ -790,11 +791,13 @@ static const VhostShadowVirtqueueOps vhost_vdpa_net_svq_ops = {
  * @device_fd vhost-vdpa file descriptor
  * @features features to negotiate
  * @cvq_index Control vq index
+ * @pcvq_group: Returns CVQ group if cvq is isolated.
  *
  * Returns -1 in case of error, 0 if false and 1 if true
  */
 static int vhost_vdpa_cvq_is_isolated(int device_fd, uint64_t features,
-                                      unsigned cvq_index, Error **errp)
+                                      unsigned cvq_index, uint32_t *pcvq_group,
+                                      Error **errp)
 {
     int64_t cvq_group;
     int r;
@@ -810,6 +813,7 @@ static int vhost_vdpa_cvq_is_isolated(int device_fd, uint64_t features,
         return cvq_group;
     }
 
+    *pcvq_group = (uint32_t)cvq_group;
     for (int i = 0; i < cvq_index; ++i) {
         int64_t group = vhost_vdpa_get_vring_group(device_fd, i, errp);
 
@@ -836,12 +840,15 @@ static int vhost_vdpa_cvq_is_isolated(int device_fd, uint64_t features,
  *                    negotiated.
  * @cvq_isolated_mq   It'll be set to true if cvq is isolated if mq is
  *                    negotiated.
+ * @cvq_group         CVQ group if MQ is not negotiated.
+ * @cvq_group_mq      CVQ group if MQ is negotiated.
  *
  * Returns -1 in case of failure
  */
 static int vhost_vdpa_probe_cvq_isolation(int device_fd, uint64_t features,
                                           int cvq_index, bool *cvq_isolated,
-                                          bool *cvq_isolated_mq, Error **errp)
+                                          bool *cvq_isolated_mq, uint32_t *cvq_group,
+                                          uint32_t *cvq_group_mq, Error **errp)
 {
     uint64_t backend_features;
     int r;
@@ -850,6 +857,8 @@ static int vhost_vdpa_probe_cvq_isolation(int device_fd, uint64_t features,
 
     *cvq_isolated = false;
     *cvq_isolated_mq = false;
+    *cvq_group = 0;
+    *cvq_group_mq = 0;
     r = ioctl(device_fd, VHOST_GET_BACKEND_FEATURES, &backend_features);
     if (unlikely(r < 0)) {
         error_setg_errno(errp, errno, "Cannot get vdpa backend_features");
@@ -862,7 +871,7 @@ static int vhost_vdpa_probe_cvq_isolation(int device_fd, uint64_t features,
 
     r = vhost_vdpa_cvq_is_isolated(device_fd,
                                    features & ~BIT_ULL(VIRTIO_NET_F_MQ), 2,
-                                   errp);
+                                   cvq_group, errp);
     if (unlikely(r < 0)) {
         if (r == -ENOTSUP) {
             /*
@@ -884,7 +893,8 @@ static int vhost_vdpa_probe_cvq_isolation(int device_fd, uint64_t features,
         return 0;
     }
 
-    r = vhost_vdpa_cvq_is_isolated(device_fd, features, cvq_index * 2, errp);
+    r = vhost_vdpa_cvq_is_isolated(device_fd, features, cvq_index * 2,
+                                   cvq_group_mq, errp);
     if (unlikely(r < 0)) {
         return r;
     }
@@ -911,6 +921,7 @@ static NetClientState *net_vhost_vdpa_init(NetClientState *peer,
     int ret = 0;
     assert(name);
     bool cvq_isolated, cvq_isolated_mq;
+    uint32_t cvq_group, cvq_group_mq;
 
     if (is_datapath) {
         nc = qemu_new_net_client(&net_vhost_vdpa_info, peer, device,
@@ -918,7 +929,8 @@ static NetClientState *net_vhost_vdpa_init(NetClientState *peer,
     } else {
         ret = vhost_vdpa_probe_cvq_isolation(vdpa_device_fd, features,
                                              queue_pair_index, &cvq_isolated,
-                                             &cvq_isolated_mq, errp);
+                                             &cvq_isolated_mq, &cvq_group,
+                                             &cvq_group_mq, errp);
         if (unlikely(ret)) {
             return NULL;
         }
@@ -951,6 +963,8 @@ static NetClientState *net_vhost_vdpa_init(NetClientState *peer,
         s->vhost_vdpa.shadow_vq_ops_opaque = s;
         s->cvq_isolated = cvq_isolated;
         s->cvq_isolated_mq = cvq_isolated_mq;
+        s->cvq_group = cvq_group;
+        s->cvq_group_mq = cvq_group_mq;
 
         /*
          * TODO: We cannot migrate devices with CVQ as there is no way to set
-- 
2.31.1

Re: [PATCH for 8.1 v2 1/6] vdpa: Remove status in reset tracing

[Stefano Garzarella]

On Thu, Mar 23, 2023 at 08:53:59PM +0100, Eugenio Pérez wrote:
>It is always 0 and it is not useful to route call through file
>descriptor.
>
>Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
>---
> hw/virtio/vhost-vdpa.c | 2 +-
> hw/virtio/trace-events | 2 +-
> 2 files changed, 2 insertions(+), 2 deletions(-)

Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>

>
>diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c
>index bc6bad23d5..bbabea18f3 100644
>--- a/hw/virtio/vhost-vdpa.c
>+++ b/hw/virtio/vhost-vdpa.c
>@@ -716,7 +716,7 @@ static int vhost_vdpa_reset_device(struct vhost_dev *dev)
>     uint8_t status = 0;
>
>     ret = vhost_vdpa_call(dev, VHOST_VDPA_SET_STATUS, &status);
>-    trace_vhost_vdpa_reset_device(dev, status);
>+    trace_vhost_vdpa_reset_device(dev);
>     v->suspended = false;
>     return ret;
> }
>diff --git a/hw/virtio/trace-events b/hw/virtio/trace-events
>index 8f8d05cf9b..6265231683 100644
>--- a/hw/virtio/trace-events
>+++ b/hw/virtio/trace-events
>@@ -44,7 +44,7 @@ vhost_vdpa_set_mem_table(void *dev, uint32_t nregions, uint32_t padding) "dev: %
> vhost_vdpa_dump_regions(void *dev, int i, uint64_t guest_phys_addr, uint64_t memory_size, uint64_t userspace_addr, uint64_t flags_padding) "dev: %p %d: guest_phys_addr: 0x%"PRIx64" memory_size: 0x%"PRIx64" userspace_addr: 0x%"PRIx64" flags_padding: 0x%"PRIx64
> vhost_vdpa_set_features(void *dev, uint64_t features) "dev: %p features: 0x%"PRIx64
> vhost_vdpa_get_device_id(void *dev, uint32_t device_id) "dev: %p device_id %"PRIu32
>-vhost_vdpa_reset_device(void *dev, uint8_t status) "dev: %p status: 0x%"PRIx8
>+vhost_vdpa_reset_device(void *dev) "dev: %p"
> vhost_vdpa_get_vq_index(void *dev, int idx, int vq_idx) "dev: %p idx: %d vq idx: %d"
> vhost_vdpa_set_vring_ready(void *dev) "dev: %p"
> vhost_vdpa_dump_config(void *dev, const char *line) "dev: %p %s"
>-- 
>2.31.1
>

Re: [PATCH for 8.1 v2 2/6] vdpa: add vhost_vdpa_reset_status_fd

[Stefano Garzarella]

On Thu, Mar 23, 2023 at 08:54:00PM +0100, Eugenio Pérez wrote:
>This allows to reset a vhost-vdpa device from external subsystems like
>vhost-net, since it does not have any struct vhost_dev by the time we
>need to use it.
>
>It is used in subsequent patches to negotiate features
>and probe for CVQ ASID isolation.
>
>Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
>---
> include/hw/virtio/vhost-vdpa.h |  1 +
> hw/virtio/vhost-vdpa.c         | 58 +++++++++++++++++++++++-----------
> 2 files changed, 41 insertions(+), 18 deletions(-)

Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>

>
>diff --git a/include/hw/virtio/vhost-vdpa.h b/include/hw/virtio/vhost-vdpa.h
>index c278a2a8de..28de7da91e 100644
>--- a/include/hw/virtio/vhost-vdpa.h
>+++ b/include/hw/virtio/vhost-vdpa.h
>@@ -54,6 +54,7 @@ typedef struct vhost_vdpa {
>     VhostVDPAHostNotifier notifier[VIRTIO_QUEUE_MAX];
> } VhostVDPA;
>
>+void vhost_vdpa_reset_status_fd(int fd);
> int vhost_vdpa_get_iova_range(int fd, struct vhost_vdpa_iova_range *iova_range);
>
> int vhost_vdpa_dma_map(struct vhost_vdpa *v, uint32_t asid, hwaddr iova,
>diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c
>index bbabea18f3..7a2053b8d9 100644
>--- a/hw/virtio/vhost-vdpa.c
>+++ b/hw/virtio/vhost-vdpa.c
>@@ -335,38 +335,45 @@ static const MemoryListener vhost_vdpa_memory_listener = {
>     .region_del = vhost_vdpa_listener_region_del,
> };
>
>-static int vhost_vdpa_call(struct vhost_dev *dev, unsigned long int request,
>-                             void *arg)
>+static int vhost_vdpa_dev_fd(const struct vhost_dev *dev)
> {
>     struct vhost_vdpa *v = dev->opaque;
>-    int fd = v->device_fd;
>-    int ret;
>
>     assert(dev->vhost_ops->backend_type == VHOST_BACKEND_TYPE_VDPA);
>+    return v->device_fd;
>+}
>+
>+static int vhost_vdpa_call_fd(int fd, unsigned long int request, void *arg)
>+{
>+    int ret = ioctl(fd, request, arg);
>
>-    ret = ioctl(fd, request, arg);
>     return ret < 0 ? -errno : ret;
> }
>
>-static int vhost_vdpa_add_status(struct vhost_dev *dev, uint8_t status)
>+static int vhost_vdpa_call(struct vhost_dev *dev, unsigned long int request,
>+                           void *arg)
>+{
>+    return vhost_vdpa_call_fd(vhost_vdpa_dev_fd(dev), request, arg);
>+}
>+
>+static int vhost_vdpa_add_status_fd(int fd, uint8_t status)
> {
>     uint8_t s;
>     int ret;
>
>-    trace_vhost_vdpa_add_status(dev, status);
>-    ret = vhost_vdpa_call(dev, VHOST_VDPA_GET_STATUS, &s);
>+    ret = vhost_vdpa_call_fd(fd, VHOST_VDPA_GET_STATUS, &s);
>     if (ret < 0) {
>         return ret;
>     }
>
>     s |= status;
>
>-    ret = vhost_vdpa_call(dev, VHOST_VDPA_SET_STATUS, &s);
>+    ret = vhost_vdpa_call_fd(fd, VHOST_VDPA_SET_STATUS, &s);
>     if (ret < 0) {
>         return ret;
>     }
>
>-    ret = vhost_vdpa_call(dev, VHOST_VDPA_GET_STATUS, &s);
>+    ret = vhost_vdpa_call_fd(fd, VHOST_VDPA_GET_STATUS, &s);
>     if (ret < 0) {
>         return ret;
>     }
>@@ -378,6 +385,12 @@ static int vhost_vdpa_add_status(struct vhost_dev *dev, uint8_t status)
>     return 0;
> }
>
>+static int vhost_vdpa_add_status(struct vhost_dev *dev, uint8_t status)
>+{
>+    trace_vhost_vdpa_add_status(dev, status);
>+    return vhost_vdpa_add_status_fd(vhost_vdpa_dev_fd(dev), status);
>+}
>+
> int vhost_vdpa_get_iova_range(int fd, struct vhost_vdpa_iova_range *iova_range)
> {
>     int ret = ioctl(fd, VHOST_VDPA_GET_IOVA_RANGE, iova_range);
>@@ -709,16 +722,20 @@ static int vhost_vdpa_get_device_id(struct vhost_dev *dev,
>     return ret;
> }
>
>+static int vhost_vdpa_reset_device_fd(int fd)
>+{
>+    uint8_t status = 0;
>+
>+    return vhost_vdpa_call_fd(fd, VHOST_VDPA_SET_STATUS, &status);
>+}
>+
> static int vhost_vdpa_reset_device(struct vhost_dev *dev)
> {
>     struct vhost_vdpa *v = dev->opaque;
>-    int ret;
>-    uint8_t status = 0;
>
>-    ret = vhost_vdpa_call(dev, VHOST_VDPA_SET_STATUS, &status);
>-    trace_vhost_vdpa_reset_device(dev);
>     v->suspended = false;
>-    return ret;
>+    trace_vhost_vdpa_reset_device(dev);
>+    return vhost_vdpa_reset_device_fd(vhost_vdpa_dev_fd(dev));
> }
>
> static int vhost_vdpa_get_vq_index(struct vhost_dev *dev, int idx)
>@@ -1170,6 +1187,13 @@ static int vhost_vdpa_dev_start(struct vhost_dev *dev, bool started)
>     return 0;
> }
>
>+void vhost_vdpa_reset_status_fd(int fd)
>+{
>+    vhost_vdpa_reset_device_fd(fd);
>+    vhost_vdpa_add_status_fd(fd, VIRTIO_CONFIG_S_ACKNOWLEDGE |
>+                                 VIRTIO_CONFIG_S_DRIVER);
>+}
>+
> static void vhost_vdpa_reset_status(struct vhost_dev *dev)
> {
>     struct vhost_vdpa *v = dev->opaque;
>@@ -1178,9 +1202,7 @@ static void vhost_vdpa_reset_status(struct vhost_dev *dev)
>         return;
>     }
>
>-    vhost_vdpa_reset_device(dev);
>-    vhost_vdpa_add_status(dev, VIRTIO_CONFIG_S_ACKNOWLEDGE |
>-                               VIRTIO_CONFIG_S_DRIVER);
>+    vhost_vdpa_reset_status_fd(vhost_vdpa_dev_fd(dev));
>     memory_listener_unregister(&v->listener);
> }
>
>-- 
>2.31.1
>

Re: [PATCH for 8.1 v2 4/6] vdpa: return errno in vhost_vdpa_get_vring_group error

[Stefano Garzarella]

On Thu, Mar 23, 2023 at 08:54:02PM +0100, Eugenio Pérez wrote:
>We need to tell in the caller, as some errors are expected in a normal
>workflow.  In particular, parent drivers in recent kernels with
>VHOST_BACKEND_F_IOTLB_ASID may not support vring groups.  In that case,
>-ENOTSUP is returned.
>
>This is the case of vp_vdpa in Linux 6.2.
>
>Next patches in this series will use that information to know if it must
>abort or not.  Also, next patches return properly an errp instead of
>printing with error_report.
>
>Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
>---
> net/vhost-vdpa.c | 9 +++++++++
> 1 file changed, 9 insertions(+)

Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>

>
>diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c
>index 99904a0da7..4397c0d4b3 100644
>--- a/net/vhost-vdpa.c
>+++ b/net/vhost-vdpa.c
>@@ -361,6 +361,14 @@ static NetClientInfo net_vhost_vdpa_info = {
>         .check_peer_type = vhost_vdpa_check_peer_type,
> };
>
>+/**
>+ * Get vring virtqueue group
>+ *
>+ * @device_fd  vdpa device fd
>+ * @vq_index   Virtqueue index
>+ *
>+ * Return -errno in case of error, or vq group if success.
>+ */
> static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index)
> {
>     struct vhost_vring_state state = {
>@@ -369,6 +377,7 @@ static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index)
>     int r = ioctl(device_fd, VHOST_VDPA_GET_VRING_GROUP, &state);
>
>     if (unlikely(r < 0)) {
>+        r = -errno;
>         error_report("Cannot get VQ %u group: %s", vq_index,
>                      g_strerror(errno));
>         return r;
>-- 
>2.31.1
>

Re: [PATCH for 8.1 v2 5/6] vdpa: move CVQ isolation check to net_init_vhost_vdpa

[Stefano Garzarella]

On Thu, Mar 23, 2023 at 08:54:03PM +0100, Eugenio Pérez wrote:
>Evaluating it at start time instead of initialization time may make the
>guest capable of dynamically adding or removing migration blockers.
>
>Also, moving to initialization reduces the number of ioctls in the
>migration, reducing failure possibilities.
>
>As a drawback we need to check for CVQ isolation twice: one time with no
>MQ negotiated and another one acking it, as long as the device supports
>it.  This is because Vring ASID / group management is based on vq
>indexes, but we don't know the index of CVQ before negotiating MQ.

I don't know this code sufficiently to do a review, but now I understand
the motivation behind it ;-)

Thanks,
Stefano

>
>Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
>---
>v2: Take out the reset of the device from vhost_vdpa_cvq_is_isolated
>---
> net/vhost-vdpa.c | 194 ++++++++++++++++++++++++++++++++++++-----------
> 1 file changed, 151 insertions(+), 43 deletions(-)
>
>diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c
>index 4397c0d4b3..db2c9afcb3 100644
>--- a/net/vhost-vdpa.c
>+++ b/net/vhost-vdpa.c
>@@ -43,6 +43,13 @@ typedef struct VhostVDPAState {
>
>     /* The device always have SVQ enabled */
>     bool always_svq;
>+
>+    /* The device can isolate CVQ in its own ASID if MQ is negotiated */
>+    bool cvq_isolated_mq;
>+
>+    /* The device can isolate CVQ in its own ASID if MQ is not negotiated */
>+    bool cvq_isolated;
>+
>     bool started;
> } VhostVDPAState;
>
>@@ -361,15 +368,8 @@ static NetClientInfo net_vhost_vdpa_info = {
>         .check_peer_type = vhost_vdpa_check_peer_type,
> };
>
>-/**
>- * Get vring virtqueue group
>- *
>- * @device_fd  vdpa device fd
>- * @vq_index   Virtqueue index
>- *
>- * Return -errno in case of error, or vq group if success.
>- */
>-static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index)
>+static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index,
>+                                          Error **errp)
> {
>     struct vhost_vring_state state = {
>         .index = vq_index,
>@@ -378,8 +378,7 @@ static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index)
>
>     if (unlikely(r < 0)) {
>         r = -errno;
>-        error_report("Cannot get VQ %u group: %s", vq_index,
>-                     g_strerror(errno));
>+        error_setg_errno(errp, errno, "Cannot get VQ %u group", vq_index);
>         return r;
>     }
>
>@@ -479,9 +478,9 @@ static int vhost_vdpa_net_cvq_start(NetClientState *nc)
> {
>     VhostVDPAState *s, *s0;
>     struct vhost_vdpa *v;
>-    uint64_t backend_features;
>     int64_t cvq_group;
>-    int cvq_index, r;
>+    int r;
>+    Error *err = NULL;
>
>     assert(nc->info->type == NET_CLIENT_DRIVER_VHOST_VDPA);
>
>@@ -501,42 +500,29 @@ static int vhost_vdpa_net_cvq_start(NetClientState *nc)
>     /*
>      * If we early return in these cases SVQ will not be enabled. The migration
>      * will be blocked as long as vhost-vdpa backends will not offer _F_LOG.
>-     *
>-     * Calling VHOST_GET_BACKEND_FEATURES as they are not available in v->dev
>-     * yet.
>      */
>-    r = ioctl(v->device_fd, VHOST_GET_BACKEND_FEATURES, &backend_features);
>-    if (unlikely(r < 0)) {
>-        error_report("Cannot get vdpa backend_features: %s(%d)",
>-            g_strerror(errno), errno);
>-        return -1;
>-    }
>-    if (!(backend_features & BIT_ULL(VHOST_BACKEND_F_IOTLB_ASID)) ||
>-        !vhost_vdpa_net_valid_svq_features(v->dev->features, NULL)) {
>+    if (!vhost_vdpa_net_valid_svq_features(v->dev->features, NULL)) {
>         return 0;
>     }
>
>-    /*
>-     * Check if all the virtqueues of the virtio device are in a different vq
>-     * than the last vq. VQ group of last group passed in cvq_group.
>-     */
>-    cvq_index = v->dev->vq_index_end - 1;
>-    cvq_group = vhost_vdpa_get_vring_group(v->device_fd, cvq_index);
>-    if (unlikely(cvq_group < 0)) {
>-        return cvq_group;
>-    }
>-    for (int i = 0; i < cvq_index; ++i) {
>-        int64_t group = vhost_vdpa_get_vring_group(v->device_fd, i);
>-
>-        if (unlikely(group < 0)) {
>-            return group;
>+    if (v->dev->features & BIT_ULL(VIRTIO_NET_F_MQ)) {
>+        if (!s->cvq_isolated_mq) {
>+            return 0;
>         }
>-
>-        if (group == cvq_group) {
>+    } else {
>+        if (!s->cvq_isolated) {
>             return 0;
>         }
>     }
>
>+    cvq_group = vhost_vdpa_get_vring_group(v->device_fd,
>+                                           v->dev->vq_index_end - 1,
>+                                           &err);
>+    if (unlikely(cvq_group < 0)) {
>+        error_report_err(err);
>+        return cvq_group;
>+    }
>+
>     r = vhost_vdpa_set_address_space_id(v, cvq_group, VHOST_VDPA_NET_CVQ_ASID);
>     if (unlikely(r < 0)) {
>         return r;
>@@ -798,6 +784,116 @@ static const VhostShadowVirtqueueOps vhost_vdpa_net_svq_ops = {
>     .avail_handler = vhost_vdpa_net_handle_ctrl_avail,
> };
>
>+/**
>+ * Probe the device to check control virtqueue is isolated.
>+ *
>+ * @device_fd vhost-vdpa file descriptor
>+ * @features features to negotiate
>+ * @cvq_index Control vq index
>+ *
>+ * Returns -1 in case of error, 0 if false and 1 if true
>+ */
>+static int vhost_vdpa_cvq_is_isolated(int device_fd, uint64_t features,
>+                                      unsigned cvq_index, Error **errp)
>+{
>+    int64_t cvq_group;
>+    int r;
>+
>+    r = vhost_vdpa_set_dev_features_fd(device_fd, features);
>+    if (unlikely(r < 0)) {
>+        error_setg_errno(errp, -r, "Cannot set device features");
>+        return r;
>+    }
>+
>+    cvq_group = vhost_vdpa_get_vring_group(device_fd, cvq_index, errp);
>+    if (unlikely(cvq_group < 0)) {
>+        return cvq_group;
>+    }
>+
>+    for (int i = 0; i < cvq_index; ++i) {
>+        int64_t group = vhost_vdpa_get_vring_group(device_fd, i, errp);
>+
>+        if (unlikely(group < 0)) {
>+            return group;
>+        }
>+
>+        if (group == (int64_t)cvq_group) {
>+            return 0;
>+        }
>+    }
>+
>+    return 1;
>+}
>+
>+/**
>+ * Probe if CVQ is isolated when the device is MQ and when it is not MQ
>+ *
>+ * @device_fd         The vdpa device fd
>+ * @features          Features offered by the device.
>+ * @cvq_index         The control vq index if mq is negotiated. Ignored
>+ *                    otherwise.
>+ * @cvq_isolated      It'll be set to true if cvq is isolated if mq is not
>+ *                    negotiated.
>+ * @cvq_isolated_mq   It'll be set to true if cvq is isolated if mq is
>+ *                    negotiated.
>+ *
>+ * Returns -1 in case of failure
>+ */
>+static int vhost_vdpa_probe_cvq_isolation(int device_fd, uint64_t features,
>+                                          int cvq_index, bool *cvq_isolated,
>+                                          bool *cvq_isolated_mq, Error **errp)
>+{
>+    uint64_t backend_features;
>+    int r;
>+
>+    ERRP_GUARD();
>+
>+    *cvq_isolated = false;
>+    *cvq_isolated_mq = false;
>+    r = ioctl(device_fd, VHOST_GET_BACKEND_FEATURES, &backend_features);
>+    if (unlikely(r < 0)) {
>+        error_setg_errno(errp, errno, "Cannot get vdpa backend_features");
>+        return r;
>+    }
>+
>+    if (!(backend_features & BIT_ULL(VHOST_BACKEND_F_IOTLB_ASID))) {
>+        return 0;
>+    }
>+
>+    r = vhost_vdpa_cvq_is_isolated(device_fd,
>+                                   features & ~BIT_ULL(VIRTIO_NET_F_MQ), 2,
>+                                   errp);
>+    if (unlikely(r < 0)) {
>+        if (r == -ENOTSUP) {
>+            /*
>+             * The kernel report VHOST_BACKEND_F_IOTLB_ASID if the vdpa
>+             * frontend support ASID but the parent driver does not.  The CVQ
>+             * cannot be isolated in this case.
>+             */
>+            error_free(*errp);
>+            *errp = NULL;
>+            return 0;
>+        }
>+
>+        return r;
>+    }
>+
>+    *cvq_isolated = r == 1;
>+    vhost_vdpa_reset_status_fd(device_fd);
>+    if (!(features & BIT_ULL(VIRTIO_NET_F_MQ))) {
>+        return 0;
>+    }
>+
>+    r = vhost_vdpa_cvq_is_isolated(device_fd, features, cvq_index * 2, errp);
>+    if (unlikely(r < 0)) {
>+        return r;
>+    }
>+
>+    *cvq_isolated_mq = r == 1;
>+    vhost_vdpa_reset_status_fd(device_fd);
>+    return 0;
>+}
>+
> static NetClientState *net_vhost_vdpa_init(NetClientState *peer,
>                                        const char *device,
>                                        const char *name,
>@@ -807,16 +903,26 @@ static NetClientState *net_vhost_vdpa_init(NetClientState *peer,
>                                        bool is_datapath,
>                                        bool svq,
>                                        struct vhost_vdpa_iova_range iova_range,
>-                                       uint64_t features)
>+                                       uint64_t features,
>+                                       Error **errp)
> {
>     NetClientState *nc = NULL;
>     VhostVDPAState *s;
>     int ret = 0;
>     assert(name);
>+    bool cvq_isolated, cvq_isolated_mq;
>+
>     if (is_datapath) {
>         nc = qemu_new_net_client(&net_vhost_vdpa_info, peer, device,
>                                  name);
>     } else {
>+        ret = vhost_vdpa_probe_cvq_isolation(vdpa_device_fd, features,
>+                                             queue_pair_index, &cvq_isolated,
>+                                             &cvq_isolated_mq, errp);
>+        if (unlikely(ret)) {
>+            return NULL;
>+        }
>+
>         nc = qemu_new_net_control_client(&net_vhost_vdpa_cvq_info, peer,
>                                          device, name);
>     }
>@@ -843,6 +949,8 @@ static NetClientState *net_vhost_vdpa_init(NetClientState *peer,
>
>         s->vhost_vdpa.shadow_vq_ops = &vhost_vdpa_net_svq_ops;
>         s->vhost_vdpa.shadow_vq_ops_opaque = s;
>+        s->cvq_isolated = cvq_isolated;
>+        s->cvq_isolated_mq = cvq_isolated_mq;
>
>         /*
>          * TODO: We cannot migrate devices with CVQ as there is no way to set
>@@ -971,7 +1079,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name,
>     for (i = 0; i < queue_pairs; i++) {
>         ncs[i] = net_vhost_vdpa_init(peer, TYPE_VHOST_VDPA, name,
>                                      vdpa_device_fd, i, 2, true, opts->x_svq,
>-                                     iova_range, features);
>+                                     iova_range, features, errp);
>         if (!ncs[i])
>             goto err;
>     }
>@@ -979,7 +1087,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name,
>     if (has_cvq) {
>         nc = net_vhost_vdpa_init(peer, TYPE_VHOST_VDPA, name,
>                                  vdpa_device_fd, i, 1, false,
>-                                 opts->x_svq, iova_range, features);
>+                                 opts->x_svq, iova_range, features, errp);
>         if (!nc)
>             goto err;
>     }
>-- 
>2.31.1
>

Re: [PATCH for 8.1 v2 1/6] vdpa: Remove status in reset tracing

[Jason Wang]

On Fri, Mar 24, 2023 at 3:54 AM Eugenio Pérez <eperezma@redhat.com> wrote:
>
> It is always 0 and it is not useful to route call through file
> descriptor.
>
> Signed-off-by: Eugenio Pérez <eperezma@redhat.com>

Acked-by: Jason Wang <jasowang@redhat.com>

Thanks

> ---
>  hw/virtio/vhost-vdpa.c | 2 +-
>  hw/virtio/trace-events | 2 +-
>  2 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c
> index bc6bad23d5..bbabea18f3 100644
> --- a/hw/virtio/vhost-vdpa.c
> +++ b/hw/virtio/vhost-vdpa.c
> @@ -716,7 +716,7 @@ static int vhost_vdpa_reset_device(struct vhost_dev *dev)
>      uint8_t status = 0;
>
>      ret = vhost_vdpa_call(dev, VHOST_VDPA_SET_STATUS, &status);
> -    trace_vhost_vdpa_reset_device(dev, status);
> +    trace_vhost_vdpa_reset_device(dev);
>      v->suspended = false;
>      return ret;
>  }
> diff --git a/hw/virtio/trace-events b/hw/virtio/trace-events
> index 8f8d05cf9b..6265231683 100644
> --- a/hw/virtio/trace-events
> +++ b/hw/virtio/trace-events
> @@ -44,7 +44,7 @@ vhost_vdpa_set_mem_table(void *dev, uint32_t nregions, uint32_t padding) "dev: %
>  vhost_vdpa_dump_regions(void *dev, int i, uint64_t guest_phys_addr, uint64_t memory_size, uint64_t userspace_addr, uint64_t flags_padding) "dev: %p %d: guest_phys_addr: 0x%"PRIx64" memory_size: 0x%"PRIx64" userspace_addr: 0x%"PRIx64" flags_padding: 0x%"PRIx64
>  vhost_vdpa_set_features(void *dev, uint64_t features) "dev: %p features: 0x%"PRIx64
>  vhost_vdpa_get_device_id(void *dev, uint32_t device_id) "dev: %p device_id %"PRIu32
> -vhost_vdpa_reset_device(void *dev, uint8_t status) "dev: %p status: 0x%"PRIx8
> +vhost_vdpa_reset_device(void *dev) "dev: %p"
>  vhost_vdpa_get_vq_index(void *dev, int idx, int vq_idx) "dev: %p idx: %d vq idx: %d"
>  vhost_vdpa_set_vring_ready(void *dev) "dev: %p"
>  vhost_vdpa_dump_config(void *dev, const char *line) "dev: %p %s"
> --
> 2.31.1
>

Re: [PATCH for 8.1 v2 2/6] vdpa: add vhost_vdpa_reset_status_fd

[Jason Wang]

On Fri, Mar 24, 2023 at 3:54 AM Eugenio Pérez <eperezma@redhat.com> wrote:
>

The title needs some tweak, I think the more appropriate one is
"introduce vhost_vdpa_call_fd()"

> This allows to reset a vhost-vdpa device from external subsystems like
> vhost-net, since it does not have any struct vhost_dev by the time we
> need to use it.

This part needs some clarification, vhost-net should be initialized
after vhost_dev, so it can access its parent vhost_dev structure?

>
> It is used in subsequent patches to negotiate features
> and probe for CVQ ASID isolation.
>
> Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
> ---
>  include/hw/virtio/vhost-vdpa.h |  1 +
>  hw/virtio/vhost-vdpa.c         | 58 +++++++++++++++++++++++-----------
>  2 files changed, 41 insertions(+), 18 deletions(-)
>
> diff --git a/include/hw/virtio/vhost-vdpa.h b/include/hw/virtio/vhost-vdpa.h
> index c278a2a8de..28de7da91e 100644
> --- a/include/hw/virtio/vhost-vdpa.h
> +++ b/include/hw/virtio/vhost-vdpa.h
> @@ -54,6 +54,7 @@ typedef struct vhost_vdpa {
>      VhostVDPAHostNotifier notifier[VIRTIO_QUEUE_MAX];
>  } VhostVDPA;
>
> +void vhost_vdpa_reset_status_fd(int fd);
>  int vhost_vdpa_get_iova_range(int fd, struct vhost_vdpa_iova_range *iova_range);
>
>  int vhost_vdpa_dma_map(struct vhost_vdpa *v, uint32_t asid, hwaddr iova,
> diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c
> index bbabea18f3..7a2053b8d9 100644
> --- a/hw/virtio/vhost-vdpa.c
> +++ b/hw/virtio/vhost-vdpa.c
> @@ -335,38 +335,45 @@ static const MemoryListener vhost_vdpa_memory_listener = {
>      .region_del = vhost_vdpa_listener_region_del,
>  };
>
> -static int vhost_vdpa_call(struct vhost_dev *dev, unsigned long int request,
> -                             void *arg)
> +static int vhost_vdpa_dev_fd(const struct vhost_dev *dev)
>  {
>      struct vhost_vdpa *v = dev->opaque;
> -    int fd = v->device_fd;
> -    int ret;
>
>      assert(dev->vhost_ops->backend_type == VHOST_BACKEND_TYPE_VDPA);
> +    return v->device_fd;
> +}
> +
> +static int vhost_vdpa_call_fd(int fd, unsigned long int request, void *arg)
> +{
> +    int ret = ioctl(fd, request, arg);
>
> -    ret = ioctl(fd, request, arg);
>      return ret < 0 ? -errno : ret;
>  }
>
> -static int vhost_vdpa_add_status(struct vhost_dev *dev, uint8_t status)
> +static int vhost_vdpa_call(struct vhost_dev *dev, unsigned long int request,
> +                           void *arg)
> +{
> +    return vhost_vdpa_call_fd(vhost_vdpa_dev_fd(dev), request, arg);
> +}
> +
> +static int vhost_vdpa_add_status_fd(int fd, uint8_t status)
>  {
>      uint8_t s;
>      int ret;
>
> -    trace_vhost_vdpa_add_status(dev, status);
> -    ret = vhost_vdpa_call(dev, VHOST_VDPA_GET_STATUS, &s);

We can stick a vhost_vdpa_call() function that calls
vhost_vdpa_call_fd() then we can avoid a lot of changes?

Thanks

> +    ret = vhost_vdpa_call_fd(fd, VHOST_VDPA_GET_STATUS, &s);
>      if (ret < 0) {
>          return ret;
>      }
>
>      s |= status;
>
> -    ret = vhost_vdpa_call(dev, VHOST_VDPA_SET_STATUS, &s);
> +    ret = vhost_vdpa_call_fd(fd, VHOST_VDPA_SET_STATUS, &s);
>      if (ret < 0) {
>          return ret;
>      }
>
> -    ret = vhost_vdpa_call(dev, VHOST_VDPA_GET_STATUS, &s);
> +    ret = vhost_vdpa_call_fd(fd, VHOST_VDPA_GET_STATUS, &s);
>      if (ret < 0) {
>          return ret;
>      }
> @@ -378,6 +385,12 @@ static int vhost_vdpa_add_status(struct vhost_dev *dev, uint8_t status)
>      return 0;
>  }
>
> +static int vhost_vdpa_add_status(struct vhost_dev *dev, uint8_t status)
> +{
> +    trace_vhost_vdpa_add_status(dev, status);
> +    return vhost_vdpa_add_status_fd(vhost_vdpa_dev_fd(dev), status);
> +}
> +
>  int vhost_vdpa_get_iova_range(int fd, struct vhost_vdpa_iova_range *iova_range)
>  {
>      int ret = ioctl(fd, VHOST_VDPA_GET_IOVA_RANGE, iova_range);
> @@ -709,16 +722,20 @@ static int vhost_vdpa_get_device_id(struct vhost_dev *dev,
>      return ret;
>  }
>
> +static int vhost_vdpa_reset_device_fd(int fd)
> +{
> +    uint8_t status = 0;
> +
> +    return vhost_vdpa_call_fd(fd, VHOST_VDPA_SET_STATUS, &status);
> +}
> +
>  static int vhost_vdpa_reset_device(struct vhost_dev *dev)
>  {
>      struct vhost_vdpa *v = dev->opaque;
> -    int ret;
> -    uint8_t status = 0;
>
> -    ret = vhost_vdpa_call(dev, VHOST_VDPA_SET_STATUS, &status);
> -    trace_vhost_vdpa_reset_device(dev);
>      v->suspended = false;
> -    return ret;
> +    trace_vhost_vdpa_reset_device(dev);
> +    return vhost_vdpa_reset_device_fd(vhost_vdpa_dev_fd(dev));
>  }
>
>  static int vhost_vdpa_get_vq_index(struct vhost_dev *dev, int idx)
> @@ -1170,6 +1187,13 @@ static int vhost_vdpa_dev_start(struct vhost_dev *dev, bool started)
>      return 0;
>  }
>
> +void vhost_vdpa_reset_status_fd(int fd)
> +{
> +    vhost_vdpa_reset_device_fd(fd);
> +    vhost_vdpa_add_status_fd(fd, VIRTIO_CONFIG_S_ACKNOWLEDGE |
> +                                 VIRTIO_CONFIG_S_DRIVER);
> +}
> +
>  static void vhost_vdpa_reset_status(struct vhost_dev *dev)
>  {
>      struct vhost_vdpa *v = dev->opaque;
> @@ -1178,9 +1202,7 @@ static void vhost_vdpa_reset_status(struct vhost_dev *dev)
>          return;
>      }
>
> -    vhost_vdpa_reset_device(dev);
> -    vhost_vdpa_add_status(dev, VIRTIO_CONFIG_S_ACKNOWLEDGE |
> -                               VIRTIO_CONFIG_S_DRIVER);
> +    vhost_vdpa_reset_status_fd(vhost_vdpa_dev_fd(dev));
>      memory_listener_unregister(&v->listener);
>  }
>
> --
> 2.31.1
>

Re: [PATCH for 8.1 v2 2/6] vdpa: add vhost_vdpa_reset_status_fd

[Eugenio Perez Martin]

On Tue, Mar 28, 2023 at 8:53 AM Jason Wang <jasowang@redhat.com> wrote:
>
> On Fri, Mar 24, 2023 at 3:54 AM Eugenio Pérez <eperezma@redhat.com> wrote:
> >
>
> The title needs some tweak, I think the more appropriate one is
> "introduce vhost_vdpa_call_fd()"
>

We can do it otherwise, but the function exported in the
include/hw/virtio/vhost-vdpa.h file is vhost_vdpa_reset_status_fd. I'd
consider vhost_vdpa_call_fd an implementation detail, only in
hw/virtio/vhost-vdpa.c file.

> > This allows to reset a vhost-vdpa device from external subsystems like
> > vhost-net, since it does not have any struct vhost_dev by the time we
> > need to use it.
>
> This part needs some clarification, vhost-net should be initialized
> after vhost_dev, so it can access its parent vhost_dev structure?
>

Next patches probes CVQ vring groups at initialization net-vdpa
initialization time (net_init_vhost_vdpa). It already fetches things
like device features using raw ioctl() calls because vhost_dev is
still not available.

> >
> > It is used in subsequent patches to negotiate features
> > and probe for CVQ ASID isolation.
> >
> > Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
> > ---
> >  include/hw/virtio/vhost-vdpa.h |  1 +
> >  hw/virtio/vhost-vdpa.c         | 58 +++++++++++++++++++++++-----------
> >  2 files changed, 41 insertions(+), 18 deletions(-)
> >
> > diff --git a/include/hw/virtio/vhost-vdpa.h b/include/hw/virtio/vhost-vdpa.h
> > index c278a2a8de..28de7da91e 100644
> > --- a/include/hw/virtio/vhost-vdpa.h
> > +++ b/include/hw/virtio/vhost-vdpa.h
> > @@ -54,6 +54,7 @@ typedef struct vhost_vdpa {
> >      VhostVDPAHostNotifier notifier[VIRTIO_QUEUE_MAX];
> >  } VhostVDPA;
> >
> > +void vhost_vdpa_reset_status_fd(int fd);
> >  int vhost_vdpa_get_iova_range(int fd, struct vhost_vdpa_iova_range *iova_range);
> >
> >  int vhost_vdpa_dma_map(struct vhost_vdpa *v, uint32_t asid, hwaddr iova,
> > diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c
> > index bbabea18f3..7a2053b8d9 100644
> > --- a/hw/virtio/vhost-vdpa.c
> > +++ b/hw/virtio/vhost-vdpa.c
> > @@ -335,38 +335,45 @@ static const MemoryListener vhost_vdpa_memory_listener = {
> >      .region_del = vhost_vdpa_listener_region_del,
> >  };
> >
> > -static int vhost_vdpa_call(struct vhost_dev *dev, unsigned long int request,
> > -                             void *arg)
> > +static int vhost_vdpa_dev_fd(const struct vhost_dev *dev)
> >  {
> >      struct vhost_vdpa *v = dev->opaque;
> > -    int fd = v->device_fd;
> > -    int ret;
> >
> >      assert(dev->vhost_ops->backend_type == VHOST_BACKEND_TYPE_VDPA);
> > +    return v->device_fd;
> > +}
> > +
> > +static int vhost_vdpa_call_fd(int fd, unsigned long int request, void *arg)
> > +{
> > +    int ret = ioctl(fd, request, arg);
> >
> > -    ret = ioctl(fd, request, arg);
> >      return ret < 0 ? -errno : ret;
> >  }
> >
> > -static int vhost_vdpa_add_status(struct vhost_dev *dev, uint8_t status)
> > +static int vhost_vdpa_call(struct vhost_dev *dev, unsigned long int request,
> > +                           void *arg)
> > +{
> > +    return vhost_vdpa_call_fd(vhost_vdpa_dev_fd(dev), request, arg);
> > +}
> > +
> > +static int vhost_vdpa_add_status_fd(int fd, uint8_t status)
> >  {
> >      uint8_t s;
> >      int ret;
> >
> > -    trace_vhost_vdpa_add_status(dev, status);
> > -    ret = vhost_vdpa_call(dev, VHOST_VDPA_GET_STATUS, &s);
>
> We can stick a vhost_vdpa_call() function that calls
> vhost_vdpa_call_fd() then we can avoid a lot of changes?
>

I don't follow this. vhost_vdpa_call already calls vhost_vdpa_call_fd.

Thanks!

> Thanks
>
> > +    ret = vhost_vdpa_call_fd(fd, VHOST_VDPA_GET_STATUS, &s);
> >      if (ret < 0) {
> >          return ret;
> >      }
> >
> >      s |= status;
> >
> > -    ret = vhost_vdpa_call(dev, VHOST_VDPA_SET_STATUS, &s);
> > +    ret = vhost_vdpa_call_fd(fd, VHOST_VDPA_SET_STATUS, &s);
> >      if (ret < 0) {
> >          return ret;
> >      }
> >
> > -    ret = vhost_vdpa_call(dev, VHOST_VDPA_GET_STATUS, &s);
> > +    ret = vhost_vdpa_call_fd(fd, VHOST_VDPA_GET_STATUS, &s);
> >      if (ret < 0) {
> >          return ret;
> >      }
> > @@ -378,6 +385,12 @@ static int vhost_vdpa_add_status(struct vhost_dev *dev, uint8_t status)
> >      return 0;
> >  }
> >
> > +static int vhost_vdpa_add_status(struct vhost_dev *dev, uint8_t status)
> > +{
> > +    trace_vhost_vdpa_add_status(dev, status);
> > +    return vhost_vdpa_add_status_fd(vhost_vdpa_dev_fd(dev), status);
> > +}
> > +
> >  int vhost_vdpa_get_iova_range(int fd, struct vhost_vdpa_iova_range *iova_range)
> >  {
> >      int ret = ioctl(fd, VHOST_VDPA_GET_IOVA_RANGE, iova_range);
> > @@ -709,16 +722,20 @@ static int vhost_vdpa_get_device_id(struct vhost_dev *dev,
> >      return ret;
> >  }
> >
> > +static int vhost_vdpa_reset_device_fd(int fd)
> > +{
> > +    uint8_t status = 0;
> > +
> > +    return vhost_vdpa_call_fd(fd, VHOST_VDPA_SET_STATUS, &status);
> > +}
> > +
> >  static int vhost_vdpa_reset_device(struct vhost_dev *dev)
> >  {
> >      struct vhost_vdpa *v = dev->opaque;
> > -    int ret;
> > -    uint8_t status = 0;
> >
> > -    ret = vhost_vdpa_call(dev, VHOST_VDPA_SET_STATUS, &status);
> > -    trace_vhost_vdpa_reset_device(dev);
> >      v->suspended = false;
> > -    return ret;
> > +    trace_vhost_vdpa_reset_device(dev);
> > +    return vhost_vdpa_reset_device_fd(vhost_vdpa_dev_fd(dev));
> >  }
> >
> >  static int vhost_vdpa_get_vq_index(struct vhost_dev *dev, int idx)
> > @@ -1170,6 +1187,13 @@ static int vhost_vdpa_dev_start(struct vhost_dev *dev, bool started)
> >      return 0;
> >  }
> >
> > +void vhost_vdpa_reset_status_fd(int fd)
> > +{
> > +    vhost_vdpa_reset_device_fd(fd);
> > +    vhost_vdpa_add_status_fd(fd, VIRTIO_CONFIG_S_ACKNOWLEDGE |
> > +                                 VIRTIO_CONFIG_S_DRIVER);
> > +}
> > +
> >  static void vhost_vdpa_reset_status(struct vhost_dev *dev)
> >  {
> >      struct vhost_vdpa *v = dev->opaque;
> > @@ -1178,9 +1202,7 @@ static void vhost_vdpa_reset_status(struct vhost_dev *dev)
> >          return;
> >      }
> >
> > -    vhost_vdpa_reset_device(dev);
> > -    vhost_vdpa_add_status(dev, VIRTIO_CONFIG_S_ACKNOWLEDGE |
> > -                               VIRTIO_CONFIG_S_DRIVER);
> > +    vhost_vdpa_reset_status_fd(vhost_vdpa_dev_fd(dev));
> >      memory_listener_unregister(&v->listener);
> >  }
> >
> > --
> > 2.31.1
> >
>

Re: [PATCH for 8.1 v2 3/6] vdpa: add vhost_vdpa_set_dev_features_fd

[Jason Wang]

On Fri, Mar 24, 2023 at 3:54 AM Eugenio Pérez <eperezma@redhat.com> wrote:
>
> This allows to set the features of a vhost-vdpa device from external
> subsystems like vhost-net.  It is used in subsequent patches to
> negotiate features and probe for CVQ ASID isolation.
>
> Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
> Signed-off-by: Eugenio Pérez <eperezma@redhat.com>

Acked-by: Jason Wang <jasowang@redhat.com>

Thanks

> ---
>  include/hw/virtio/vhost-vdpa.h |  1 +
>  hw/virtio/vhost-vdpa.c         | 20 +++++++++++++-------
>  2 files changed, 14 insertions(+), 7 deletions(-)
>
> diff --git a/include/hw/virtio/vhost-vdpa.h b/include/hw/virtio/vhost-vdpa.h
> index 28de7da91e..a9cb6f3a32 100644
> --- a/include/hw/virtio/vhost-vdpa.h
> +++ b/include/hw/virtio/vhost-vdpa.h
> @@ -55,6 +55,7 @@ typedef struct vhost_vdpa {
>  } VhostVDPA;
>
>  void vhost_vdpa_reset_status_fd(int fd);
> +int vhost_vdpa_set_dev_features_fd(int fd, uint64_t features);
>  int vhost_vdpa_get_iova_range(int fd, struct vhost_vdpa_iova_range *iova_range);
>
>  int vhost_vdpa_dma_map(struct vhost_vdpa *v, uint32_t asid, hwaddr iova,
> diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c
> index 7a2053b8d9..acd5be46a9 100644
> --- a/hw/virtio/vhost-vdpa.c
> +++ b/hw/virtio/vhost-vdpa.c
> @@ -651,11 +651,22 @@ static int vhost_vdpa_set_mem_table(struct vhost_dev *dev,
>      return 0;
>  }
>
> +int vhost_vdpa_set_dev_features_fd(int fd, uint64_t features)
> +{
> +    int ret;
> +
> +    ret = vhost_vdpa_call_fd(fd, VHOST_SET_FEATURES, &features);
> +    if (ret) {
> +        return ret;
> +    }
> +
> +    return vhost_vdpa_add_status_fd(fd, VIRTIO_CONFIG_S_FEATURES_OK);
> +}
> +
>  static int vhost_vdpa_set_features(struct vhost_dev *dev,
>                                     uint64_t features)
>  {
>      struct vhost_vdpa *v = dev->opaque;
> -    int ret;
>
>      if (!vhost_vdpa_first_dev(dev)) {
>          return 0;
> @@ -678,12 +689,7 @@ static int vhost_vdpa_set_features(struct vhost_dev *dev,
>      }
>
>      trace_vhost_vdpa_set_features(dev, features);
> -    ret = vhost_vdpa_call(dev, VHOST_SET_FEATURES, &features);
> -    if (ret) {
> -        return ret;
> -    }
> -
> -    return vhost_vdpa_add_status(dev, VIRTIO_CONFIG_S_FEATURES_OK);
> +    return vhost_vdpa_set_dev_features_fd(vhost_vdpa_dev_fd(dev), features);
>  }
>
>  static int vhost_vdpa_set_backend_cap(struct vhost_dev *dev)
> --
> 2.31.1
>

Re: [PATCH for 8.1 v2 4/6] vdpa: return errno in vhost_vdpa_get_vring_group error

[Jason Wang]

On Fri, Mar 24, 2023 at 3:54 AM Eugenio Pérez <eperezma@redhat.com> wrote:
>
> We need to tell in the caller, as some errors are expected in a normal
> workflow.  In particular, parent drivers in recent kernels with
> VHOST_BACKEND_F_IOTLB_ASID may not support vring groups.  In that case,
> -ENOTSUP is returned.
>
> This is the case of vp_vdpa in Linux 6.2.
>
> Next patches in this series will use that information to know if it must
> abort or not.  Also, next patches return properly an errp instead of
> printing with error_report.
>
> Signed-off-by: Eugenio Pérez <eperezma@redhat.com>

Acked-by: Jason Wang <jasowang@redhat.com>

Thanks

> ---
>  net/vhost-vdpa.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
>
> diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c
> index 99904a0da7..4397c0d4b3 100644
> --- a/net/vhost-vdpa.c
> +++ b/net/vhost-vdpa.c
> @@ -361,6 +361,14 @@ static NetClientInfo net_vhost_vdpa_info = {
>          .check_peer_type = vhost_vdpa_check_peer_type,
>  };
>
> +/**
> + * Get vring virtqueue group
> + *
> + * @device_fd  vdpa device fd
> + * @vq_index   Virtqueue index
> + *
> + * Return -errno in case of error, or vq group if success.
> + */
>  static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index)
>  {
>      struct vhost_vring_state state = {
> @@ -369,6 +377,7 @@ static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index)
>      int r = ioctl(device_fd, VHOST_VDPA_GET_VRING_GROUP, &state);
>
>      if (unlikely(r < 0)) {
> +        r = -errno;
>          error_report("Cannot get VQ %u group: %s", vq_index,
>                       g_strerror(errno));
>          return r;
> --
> 2.31.1
>

Re: [PATCH for 8.1 v2 5/6] vdpa: move CVQ isolation check to net_init_vhost_vdpa

[Jason Wang]

On Fri, Mar 24, 2023 at 3:54 AM Eugenio Pérez <eperezma@redhat.com> wrote:
>
> Evaluating it at start time instead of initialization time may make the
> guest capable of dynamically adding or removing migration blockers.
>
> Also, moving to initialization reduces the number of ioctls in the
> migration, reducing failure possibilities.
>
> As a drawback we need to check for CVQ isolation twice: one time with no
> MQ negotiated and another one acking it, as long as the device supports
> it.  This is because Vring ASID / group management is based on vq
> indexes, but we don't know the index of CVQ before negotiating MQ.

We need to fail if we see a device that can isolate cvq without MQ but
not with MQ.

>
> Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
> ---
> v2: Take out the reset of the device from vhost_vdpa_cvq_is_isolated
> ---
>  net/vhost-vdpa.c | 194 ++++++++++++++++++++++++++++++++++++-----------
>  1 file changed, 151 insertions(+), 43 deletions(-)
>
> diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c
> index 4397c0d4b3..db2c9afcb3 100644
> --- a/net/vhost-vdpa.c
> +++ b/net/vhost-vdpa.c
> @@ -43,6 +43,13 @@ typedef struct VhostVDPAState {
>
>      /* The device always have SVQ enabled */
>      bool always_svq;
> +
> +    /* The device can isolate CVQ in its own ASID if MQ is negotiated */
> +    bool cvq_isolated_mq;
> +
> +    /* The device can isolate CVQ in its own ASID if MQ is not negotiated */
> +    bool cvq_isolated;

As stated above, if we need a device that cvq_isolated_mq^cvq_isolated
== true, we need to fail. This may reduce the complexity of the code?

Thanks

> +
>      bool started;
>  } VhostVDPAState;
>
> @@ -361,15 +368,8 @@ static NetClientInfo net_vhost_vdpa_info = {
>          .check_peer_type = vhost_vdpa_check_peer_type,
>  };
>
> -/**
> - * Get vring virtqueue group
> - *
> - * @device_fd  vdpa device fd
> - * @vq_index   Virtqueue index
> - *
> - * Return -errno in case of error, or vq group if success.
> - */
> -static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index)
> +static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index,
> +                                          Error **errp)
>  {
>      struct vhost_vring_state state = {
>          .index = vq_index,
> @@ -378,8 +378,7 @@ static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index)
>
>      if (unlikely(r < 0)) {
>          r = -errno;
> -        error_report("Cannot get VQ %u group: %s", vq_index,
> -                     g_strerror(errno));
> +        error_setg_errno(errp, errno, "Cannot get VQ %u group", vq_index);
>          return r;
>      }
>
> @@ -479,9 +478,9 @@ static int vhost_vdpa_net_cvq_start(NetClientState *nc)
>  {
>      VhostVDPAState *s, *s0;
>      struct vhost_vdpa *v;
> -    uint64_t backend_features;
>      int64_t cvq_group;
> -    int cvq_index, r;
> +    int r;
> +    Error *err = NULL;
>
>      assert(nc->info->type == NET_CLIENT_DRIVER_VHOST_VDPA);
>
> @@ -501,42 +500,29 @@ static int vhost_vdpa_net_cvq_start(NetClientState *nc)
>      /*
>       * If we early return in these cases SVQ will not be enabled. The migration
>       * will be blocked as long as vhost-vdpa backends will not offer _F_LOG.
> -     *
> -     * Calling VHOST_GET_BACKEND_FEATURES as they are not available in v->dev
> -     * yet.
>       */
> -    r = ioctl(v->device_fd, VHOST_GET_BACKEND_FEATURES, &backend_features);
> -    if (unlikely(r < 0)) {
> -        error_report("Cannot get vdpa backend_features: %s(%d)",
> -            g_strerror(errno), errno);
> -        return -1;
> -    }
> -    if (!(backend_features & BIT_ULL(VHOST_BACKEND_F_IOTLB_ASID)) ||
> -        !vhost_vdpa_net_valid_svq_features(v->dev->features, NULL)) {
> +    if (!vhost_vdpa_net_valid_svq_features(v->dev->features, NULL)) {
>          return 0;
>      }
>
> -    /*
> -     * Check if all the virtqueues of the virtio device are in a different vq
> -     * than the last vq. VQ group of last group passed in cvq_group.
> -     */
> -    cvq_index = v->dev->vq_index_end - 1;
> -    cvq_group = vhost_vdpa_get_vring_group(v->device_fd, cvq_index);
> -    if (unlikely(cvq_group < 0)) {
> -        return cvq_group;
> -    }
> -    for (int i = 0; i < cvq_index; ++i) {
> -        int64_t group = vhost_vdpa_get_vring_group(v->device_fd, i);
> -
> -        if (unlikely(group < 0)) {
> -            return group;
> +    if (v->dev->features & BIT_ULL(VIRTIO_NET_F_MQ)) {
> +        if (!s->cvq_isolated_mq) {
> +            return 0;
>          }
> -
> -        if (group == cvq_group) {
> +    } else {
> +        if (!s->cvq_isolated) {
>              return 0;
>          }
>      }
>
> +    cvq_group = vhost_vdpa_get_vring_group(v->device_fd,
> +                                           v->dev->vq_index_end - 1,
> +                                           &err);
> +    if (unlikely(cvq_group < 0)) {
> +        error_report_err(err);
> +        return cvq_group;
> +    }
> +
>      r = vhost_vdpa_set_address_space_id(v, cvq_group, VHOST_VDPA_NET_CVQ_ASID);
>      if (unlikely(r < 0)) {
>          return r;
> @@ -798,6 +784,116 @@ static const VhostShadowVirtqueueOps vhost_vdpa_net_svq_ops = {
>      .avail_handler = vhost_vdpa_net_handle_ctrl_avail,
>  };
>
> +/**
> + * Probe the device to check control virtqueue is isolated.
> + *
> + * @device_fd vhost-vdpa file descriptor
> + * @features features to negotiate
> + * @cvq_index Control vq index
> + *
> + * Returns -1 in case of error, 0 if false and 1 if true
> + */
> +static int vhost_vdpa_cvq_is_isolated(int device_fd, uint64_t features,
> +                                      unsigned cvq_index, Error **errp)
> +{
> +    int64_t cvq_group;
> +    int r;
> +
> +    r = vhost_vdpa_set_dev_features_fd(device_fd, features);
> +    if (unlikely(r < 0)) {
> +        error_setg_errno(errp, -r, "Cannot set device features");
> +        return r;
> +    }
> +
> +    cvq_group = vhost_vdpa_get_vring_group(device_fd, cvq_index, errp);
> +    if (unlikely(cvq_group < 0)) {
> +        return cvq_group;
> +    }
> +
> +    for (int i = 0; i < cvq_index; ++i) {
> +        int64_t group = vhost_vdpa_get_vring_group(device_fd, i, errp);
> +
> +        if (unlikely(group < 0)) {
> +            return group;
> +        }
> +
> +        if (group == (int64_t)cvq_group) {
> +            return 0;
> +        }
> +    }
> +
> +    return 1;
> +}
> +
> +/**
> + * Probe if CVQ is isolated when the device is MQ and when it is not MQ
> + *
> + * @device_fd         The vdpa device fd
> + * @features          Features offered by the device.
> + * @cvq_index         The control vq index if mq is negotiated. Ignored
> + *                    otherwise.
> + * @cvq_isolated      It'll be set to true if cvq is isolated if mq is not
> + *                    negotiated.
> + * @cvq_isolated_mq   It'll be set to true if cvq is isolated if mq is
> + *                    negotiated.
> + *
> + * Returns -1 in case of failure
> + */
> +static int vhost_vdpa_probe_cvq_isolation(int device_fd, uint64_t features,
> +                                          int cvq_index, bool *cvq_isolated,
> +                                          bool *cvq_isolated_mq, Error **errp)
> +{
> +    uint64_t backend_features;
> +    int r;
> +
> +    ERRP_GUARD();
> +
> +    *cvq_isolated = false;
> +    *cvq_isolated_mq = false;
> +    r = ioctl(device_fd, VHOST_GET_BACKEND_FEATURES, &backend_features);
> +    if (unlikely(r < 0)) {
> +        error_setg_errno(errp, errno, "Cannot get vdpa backend_features");
> +        return r;
> +    }
> +
> +    if (!(backend_features & BIT_ULL(VHOST_BACKEND_F_IOTLB_ASID))) {
> +        return 0;
> +    }
> +
> +    r = vhost_vdpa_cvq_is_isolated(device_fd,
> +                                   features & ~BIT_ULL(VIRTIO_NET_F_MQ), 2,
> +                                   errp);
> +    if (unlikely(r < 0)) {
> +        if (r == -ENOTSUP) {
> +            /*
> +             * The kernel report VHOST_BACKEND_F_IOTLB_ASID if the vdpa
> +             * frontend support ASID but the parent driver does not.  The CVQ
> +             * cannot be isolated in this case.
> +             */
> +            error_free(*errp);
> +            *errp = NULL;
> +            return 0;
> +        }
> +
> +        return r;
> +    }
> +
> +    *cvq_isolated = r == 1;
> +    vhost_vdpa_reset_status_fd(device_fd);
> +    if (!(features & BIT_ULL(VIRTIO_NET_F_MQ))) {
> +        return 0;
> +    }
> +
> +    r = vhost_vdpa_cvq_is_isolated(device_fd, features, cvq_index * 2, errp);
> +    if (unlikely(r < 0)) {
> +        return r;
> +    }
> +
> +    *cvq_isolated_mq = r == 1;
> +    vhost_vdpa_reset_status_fd(device_fd);
> +    return 0;
> +}
> +
>  static NetClientState *net_vhost_vdpa_init(NetClientState *peer,
>                                         const char *device,
>                                         const char *name,
> @@ -807,16 +903,26 @@ static NetClientState *net_vhost_vdpa_init(NetClientState *peer,
>                                         bool is_datapath,
>                                         bool svq,
>                                         struct vhost_vdpa_iova_range iova_range,
> -                                       uint64_t features)
> +                                       uint64_t features,
> +                                       Error **errp)
>  {
>      NetClientState *nc = NULL;
>      VhostVDPAState *s;
>      int ret = 0;
>      assert(name);
> +    bool cvq_isolated, cvq_isolated_mq;
> +
>      if (is_datapath) {
>          nc = qemu_new_net_client(&net_vhost_vdpa_info, peer, device,
>                                   name);
>      } else {
> +        ret = vhost_vdpa_probe_cvq_isolation(vdpa_device_fd, features,
> +                                             queue_pair_index, &cvq_isolated,
> +                                             &cvq_isolated_mq, errp);
> +        if (unlikely(ret)) {
> +            return NULL;
> +        }
> +
>          nc = qemu_new_net_control_client(&net_vhost_vdpa_cvq_info, peer,
>                                           device, name);
>      }
> @@ -843,6 +949,8 @@ static NetClientState *net_vhost_vdpa_init(NetClientState *peer,
>
>          s->vhost_vdpa.shadow_vq_ops = &vhost_vdpa_net_svq_ops;
>          s->vhost_vdpa.shadow_vq_ops_opaque = s;
> +        s->cvq_isolated = cvq_isolated;
> +        s->cvq_isolated_mq = cvq_isolated_mq;
>
>          /*
>           * TODO: We cannot migrate devices with CVQ as there is no way to set
> @@ -971,7 +1079,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name,
>      for (i = 0; i < queue_pairs; i++) {
>          ncs[i] = net_vhost_vdpa_init(peer, TYPE_VHOST_VDPA, name,
>                                       vdpa_device_fd, i, 2, true, opts->x_svq,
> -                                     iova_range, features);
> +                                     iova_range, features, errp);
>          if (!ncs[i])
>              goto err;
>      }
> @@ -979,7 +1087,7 @@ int net_init_vhost_vdpa(const Netdev *netdev, const char *name,
>      if (has_cvq) {
>          nc = net_vhost_vdpa_init(peer, TYPE_VHOST_VDPA, name,
>                                   vdpa_device_fd, i, 1, false,
> -                                 opts->x_svq, iova_range, features);
> +                                 opts->x_svq, iova_range, features, errp);
>          if (!nc)
>              goto err;
>      }
> --
> 2.31.1
>

Re: [PATCH for 8.1 v2 5/6] vdpa: move CVQ isolation check to net_init_vhost_vdpa

[Jason Wang]

On Thu, Mar 30, 2023 at 2:20 PM Jason Wang <jasowang@redhat.com> wrote:
>
> On Fri, Mar 24, 2023 at 3:54 AM Eugenio Pérez <eperezma@redhat.com> wrote:
> >
> > Evaluating it at start time instead of initialization time may make the
> > guest capable of dynamically adding or removing migration blockers.
> >
> > Also, moving to initialization reduces the number of ioctls in the
> > migration, reducing failure possibilities.
> >
> > As a drawback we need to check for CVQ isolation twice: one time with no
> > MQ negotiated and another one acking it, as long as the device supports
> > it.  This is because Vring ASID / group management is based on vq
> > indexes, but we don't know the index of CVQ before negotiating MQ.
>
> We need to fail if we see a device that can isolate cvq without MQ but
> not with MQ.
>
> >
> > Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
> > ---
> > v2: Take out the reset of the device from vhost_vdpa_cvq_is_isolated
> > ---
> >  net/vhost-vdpa.c | 194 ++++++++++++++++++++++++++++++++++++-----------
> >  1 file changed, 151 insertions(+), 43 deletions(-)
> >
> > diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c
> > index 4397c0d4b3..db2c9afcb3 100644
> > --- a/net/vhost-vdpa.c
> > +++ b/net/vhost-vdpa.c
> > @@ -43,6 +43,13 @@ typedef struct VhostVDPAState {
> >
> >      /* The device always have SVQ enabled */
> >      bool always_svq;
> > +
> > +    /* The device can isolate CVQ in its own ASID if MQ is negotiated */
> > +    bool cvq_isolated_mq;
> > +
> > +    /* The device can isolate CVQ in its own ASID if MQ is not negotiated */
> > +    bool cvq_isolated;
>
> As stated above, if we need a device that cvq_isolated_mq^cvq_isolated
> == true, we need to fail. This may reduce the complexity of the code?
>
> Thanks

Since we are the mediation layer, Qemu can alway choose to negotiate
MQ regardless whether or not it is supported by the guest. In this
way, we can have a stable virtqueue index for cvq.

Thanks

Re: [PATCH for 8.1 v2 5/6] vdpa: move CVQ isolation check to net_init_vhost_vdpa

[Eugenio Perez Martin]

On Thu, Mar 30, 2023 at 8:23 AM Jason Wang <jasowang@redhat.com> wrote:
>
> On Thu, Mar 30, 2023 at 2:20 PM Jason Wang <jasowang@redhat.com> wrote:
> >
> > On Fri, Mar 24, 2023 at 3:54 AM Eugenio Pérez <eperezma@redhat.com> wrote:
> > >
> > > Evaluating it at start time instead of initialization time may make the
> > > guest capable of dynamically adding or removing migration blockers.
> > >
> > > Also, moving to initialization reduces the number of ioctls in the
> > > migration, reducing failure possibilities.
> > >
> > > As a drawback we need to check for CVQ isolation twice: one time with no
> > > MQ negotiated and another one acking it, as long as the device supports
> > > it.  This is because Vring ASID / group management is based on vq
> > > indexes, but we don't know the index of CVQ before negotiating MQ.
> >
> > We need to fail if we see a device that can isolate cvq without MQ but
> > not with MQ.
> >
> > >
> > > Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
> > > ---
> > > v2: Take out the reset of the device from vhost_vdpa_cvq_is_isolated
> > > ---
> > >  net/vhost-vdpa.c | 194 ++++++++++++++++++++++++++++++++++++-----------
> > >  1 file changed, 151 insertions(+), 43 deletions(-)
> > >
> > > diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c
> > > index 4397c0d4b3..db2c9afcb3 100644
> > > --- a/net/vhost-vdpa.c
> > > +++ b/net/vhost-vdpa.c
> > > @@ -43,6 +43,13 @@ typedef struct VhostVDPAState {
> > >
> > >      /* The device always have SVQ enabled */
> > >      bool always_svq;
> > > +
> > > +    /* The device can isolate CVQ in its own ASID if MQ is negotiated */
> > > +    bool cvq_isolated_mq;
> > > +
> > > +    /* The device can isolate CVQ in its own ASID if MQ is not negotiated */
> > > +    bool cvq_isolated;
> >
> > As stated above, if we need a device that cvq_isolated_mq^cvq_isolated
> > == true, we need to fail. This may reduce the complexity of the code?
> >
> > Thanks
>
> Since we are the mediation layer, Qemu can alway choose to negotiate
> MQ regardless whether or not it is supported by the guest. In this
> way, we can have a stable virtqueue index for cvq.
>

I think it is a great idea and it simplifies this patch somehow.
However, we need something like the queue mapping [1] to do so :).

To double confirm:
* If the device supports MQ, only probe MQ. If not, only probe !MQ.
* Only store cvq_isolated in VhostVDPAState.

Now, if the device does not negotiate MQ but the device supports MQ:
* All the requests to queue 3 must be redirected to the last queue in
the device. That includes set_vq_address, notifiers regions, etc.

I'm totally ok to go this route but it's not immediate.

Thanks!

[1] https://lists.gnu.org/archive/html/qemu-devel/2023-01/msg07157.html

Re: [PATCH for 8.1 v2 5/6] vdpa: move CVQ isolation check to net_init_vhost_vdpa

[Jason Wang]

在 2023/3/30 18:42, Eugenio Perez Martin 写道:
> On Thu, Mar 30, 2023 at 8:23 AM Jason Wang <jasowang@redhat.com> wrote:
>> On Thu, Mar 30, 2023 at 2:20 PM Jason Wang <jasowang@redhat.com> wrote:
>>> On Fri, Mar 24, 2023 at 3:54 AM Eugenio Pérez <eperezma@redhat.com> wrote:
>>>> Evaluating it at start time instead of initialization time may make the
>>>> guest capable of dynamically adding or removing migration blockers.
>>>>
>>>> Also, moving to initialization reduces the number of ioctls in the
>>>> migration, reducing failure possibilities.
>>>>
>>>> As a drawback we need to check for CVQ isolation twice: one time with no
>>>> MQ negotiated and another one acking it, as long as the device supports
>>>> it.  This is because Vring ASID / group management is based on vq
>>>> indexes, but we don't know the index of CVQ before negotiating MQ.
>>> We need to fail if we see a device that can isolate cvq without MQ but
>>> not with MQ.
>>>
>>>> Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
>>>> ---
>>>> v2: Take out the reset of the device from vhost_vdpa_cvq_is_isolated
>>>> ---
>>>>   net/vhost-vdpa.c | 194 ++++++++++++++++++++++++++++++++++++-----------
>>>>   1 file changed, 151 insertions(+), 43 deletions(-)
>>>>
>>>> diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c
>>>> index 4397c0d4b3..db2c9afcb3 100644
>>>> --- a/net/vhost-vdpa.c
>>>> +++ b/net/vhost-vdpa.c
>>>> @@ -43,6 +43,13 @@ typedef struct VhostVDPAState {
>>>>
>>>>       /* The device always have SVQ enabled */
>>>>       bool always_svq;
>>>> +
>>>> +    /* The device can isolate CVQ in its own ASID if MQ is negotiated */
>>>> +    bool cvq_isolated_mq;
>>>> +
>>>> +    /* The device can isolate CVQ in its own ASID if MQ is not negotiated */
>>>> +    bool cvq_isolated;
>>> As stated above, if we need a device that cvq_isolated_mq^cvq_isolated
>>> == true, we need to fail. This may reduce the complexity of the code?
>>>
>>> Thanks
>> Since we are the mediation layer, Qemu can alway choose to negotiate
>> MQ regardless whether or not it is supported by the guest. In this
>> way, we can have a stable virtqueue index for cvq.
>>
> I think it is a great idea and it simplifies this patch somehow.
> However, we need something like the queue mapping [1] to do so :).
>
> To double confirm:
> * If the device supports MQ, only probe MQ. If not, only probe !MQ.
> * Only store cvq_isolated in VhostVDPAState.
>
> Now, if the device does not negotiate MQ but the device supports MQ:


I'm not sure I understand here, if device supports MQ it should accepts 
MQ or we can fail the initialization here.


> * All the requests to queue 3 must be redirected to the last queue in
> the device. That includes set_vq_address, notifiers regions, etc.


This also means we will only mediate the case:

1) Qemu emulated virtio-net has 1 queue but device support multiple queue

but not

2) Qemu emulated virtio-net has M queue but device support N queue (N>M)


>
> I'm totally ok to go this route but it's not immediate.


Yes but I mean, we can start from failing the device if 
cvq_isolated_mq^cvq_isolated == true (or I wonder if we can meet this 
condition for any existing parents).

Thanks


>
> Thanks!
>
> [1] https://lists.gnu.org/archive/html/qemu-devel/2023-01/msg07157.html
>

Re: [PATCH for 8.1 v2 5/6] vdpa: move CVQ isolation check to net_init_vhost_vdpa

[Eugenio Perez Martin]

On Fri, Mar 31, 2023 at 10:00 AM Jason Wang <jasowang@redhat.com> wrote:
>
>
> 在 2023/3/30 18:42, Eugenio Perez Martin 写道:
> > On Thu, Mar 30, 2023 at 8:23 AM Jason Wang <jasowang@redhat.com> wrote:
> >> On Thu, Mar 30, 2023 at 2:20 PM Jason Wang <jasowang@redhat.com> wrote:
> >>> On Fri, Mar 24, 2023 at 3:54 AM Eugenio Pérez <eperezma@redhat.com> wrote:
> >>>> Evaluating it at start time instead of initialization time may make the
> >>>> guest capable of dynamically adding or removing migration blockers.
> >>>>
> >>>> Also, moving to initialization reduces the number of ioctls in the
> >>>> migration, reducing failure possibilities.
> >>>>
> >>>> As a drawback we need to check for CVQ isolation twice: one time with no
> >>>> MQ negotiated and another one acking it, as long as the device supports
> >>>> it.  This is because Vring ASID / group management is based on vq
> >>>> indexes, but we don't know the index of CVQ before negotiating MQ.
> >>> We need to fail if we see a device that can isolate cvq without MQ but
> >>> not with MQ.
> >>>
> >>>> Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
> >>>> ---
> >>>> v2: Take out the reset of the device from vhost_vdpa_cvq_is_isolated
> >>>> ---
> >>>>   net/vhost-vdpa.c | 194 ++++++++++++++++++++++++++++++++++++-----------
> >>>>   1 file changed, 151 insertions(+), 43 deletions(-)
> >>>>
> >>>> diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c
> >>>> index 4397c0d4b3..db2c9afcb3 100644
> >>>> --- a/net/vhost-vdpa.c
> >>>> +++ b/net/vhost-vdpa.c
> >>>> @@ -43,6 +43,13 @@ typedef struct VhostVDPAState {
> >>>>
> >>>>       /* The device always have SVQ enabled */
> >>>>       bool always_svq;
> >>>> +
> >>>> +    /* The device can isolate CVQ in its own ASID if MQ is negotiated */
> >>>> +    bool cvq_isolated_mq;
> >>>> +
> >>>> +    /* The device can isolate CVQ in its own ASID if MQ is not negotiated */
> >>>> +    bool cvq_isolated;
> >>> As stated above, if we need a device that cvq_isolated_mq^cvq_isolated
> >>> == true, we need to fail. This may reduce the complexity of the code?
> >>>
> >>> Thanks
> >> Since we are the mediation layer, Qemu can alway choose to negotiate
> >> MQ regardless whether or not it is supported by the guest. In this
> >> way, we can have a stable virtqueue index for cvq.
> >>
> > I think it is a great idea and it simplifies this patch somehow.
> > However, we need something like the queue mapping [1] to do so :).
> >
> > To double confirm:
> > * If the device supports MQ, only probe MQ. If not, only probe !MQ.
> > * Only store cvq_isolated in VhostVDPAState.
> >
> > Now, if the device does not negotiate MQ but the device supports MQ:
>
>
> I'm not sure I understand here, if device supports MQ it should accepts
> MQ or we can fail the initialization here.
>

My fault, I wanted to say "if the device offers MQ but the driver does
not acks it".

>
> > * All the requests to queue 3 must be redirected to the last queue in
> > the device. That includes set_vq_address, notifiers regions, etc.
>
>
> This also means we will only mediate the case:
>
> 1) Qemu emulated virtio-net has 1 queue but device support multiple queue
>
> but not
>
> 2) Qemu emulated virtio-net has M queue but device support N queue (N>M)
>

Right.

>
> >
> > I'm totally ok to go this route but it's not immediate.
>
>
> Yes but I mean, we can start from failing the device if
> cvq_isolated_mq^cvq_isolated == true
>

So probe the two cases but set VhostVDPAState->cvq_isolated =
cvq_isolated && cvq_mq_isolated then? No map involved that way, and
all parents should behave that way.

> (or I wonder if we can meet this condition for any existing parents).

I don't think so, but I think we need to probe the two anyway.
Otherwise we may change the dataplane asid too.

Thanks!

Re: [PATCH for 8.1 v2 5/6] vdpa: move CVQ isolation check to net_init_vhost_vdpa

[Jason Wang]

On Fri, Mar 31, 2023 at 6:12 PM Eugenio Perez Martin
<eperezma@redhat.com> wrote:
>
> On Fri, Mar 31, 2023 at 10:00 AM Jason Wang <jasowang@redhat.com> wrote:
> >
> >
> > 在 2023/3/30 18:42, Eugenio Perez Martin 写道:
> > > On Thu, Mar 30, 2023 at 8:23 AM Jason Wang <jasowang@redhat.com> wrote:
> > >> On Thu, Mar 30, 2023 at 2:20 PM Jason Wang <jasowang@redhat.com> wrote:
> > >>> On Fri, Mar 24, 2023 at 3:54 AM Eugenio Pérez <eperezma@redhat.com> wrote:
> > >>>> Evaluating it at start time instead of initialization time may make the
> > >>>> guest capable of dynamically adding or removing migration blockers.
> > >>>>
> > >>>> Also, moving to initialization reduces the number of ioctls in the
> > >>>> migration, reducing failure possibilities.
> > >>>>
> > >>>> As a drawback we need to check for CVQ isolation twice: one time with no
> > >>>> MQ negotiated and another one acking it, as long as the device supports
> > >>>> it.  This is because Vring ASID / group management is based on vq
> > >>>> indexes, but we don't know the index of CVQ before negotiating MQ.
> > >>> We need to fail if we see a device that can isolate cvq without MQ but
> > >>> not with MQ.
> > >>>
> > >>>> Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
> > >>>> ---
> > >>>> v2: Take out the reset of the device from vhost_vdpa_cvq_is_isolated
> > >>>> ---
> > >>>>   net/vhost-vdpa.c | 194 ++++++++++++++++++++++++++++++++++++-----------
> > >>>>   1 file changed, 151 insertions(+), 43 deletions(-)
> > >>>>
> > >>>> diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c
> > >>>> index 4397c0d4b3..db2c9afcb3 100644
> > >>>> --- a/net/vhost-vdpa.c
> > >>>> +++ b/net/vhost-vdpa.c
> > >>>> @@ -43,6 +43,13 @@ typedef struct VhostVDPAState {
> > >>>>
> > >>>>       /* The device always have SVQ enabled */
> > >>>>       bool always_svq;
> > >>>> +
> > >>>> +    /* The device can isolate CVQ in its own ASID if MQ is negotiated */
> > >>>> +    bool cvq_isolated_mq;
> > >>>> +
> > >>>> +    /* The device can isolate CVQ in its own ASID if MQ is not negotiated */
> > >>>> +    bool cvq_isolated;
> > >>> As stated above, if we need a device that cvq_isolated_mq^cvq_isolated
> > >>> == true, we need to fail. This may reduce the complexity of the code?
> > >>>
> > >>> Thanks
> > >> Since we are the mediation layer, Qemu can alway choose to negotiate
> > >> MQ regardless whether or not it is supported by the guest. In this
> > >> way, we can have a stable virtqueue index for cvq.
> > >>
> > > I think it is a great idea and it simplifies this patch somehow.
> > > However, we need something like the queue mapping [1] to do so :).
> > >
> > > To double confirm:
> > > * If the device supports MQ, only probe MQ. If not, only probe !MQ.
> > > * Only store cvq_isolated in VhostVDPAState.
> > >
> > > Now, if the device does not negotiate MQ but the device supports MQ:
> >
> >
> > I'm not sure I understand here, if device supports MQ it should accepts
> > MQ or we can fail the initialization here.
> >
>
> My fault, I wanted to say "if the device offers MQ but the driver does
> not acks it".
>
> >
> > > * All the requests to queue 3 must be redirected to the last queue in
> > > the device. That includes set_vq_address, notifiers regions, etc.
> >
> >
> > This also means we will only mediate the case:
> >
> > 1) Qemu emulated virtio-net has 1 queue but device support multiple queue
> >
> > but not
> >
> > 2) Qemu emulated virtio-net has M queue but device support N queue (N>M)
> >
>
> Right.
>
> >
> > >
> > > I'm totally ok to go this route but it's not immediate.
> >
> >
> > Yes but I mean, we can start from failing the device if
> > cvq_isolated_mq^cvq_isolated == true
> >
>
> So probe the two cases but set VhostVDPAState->cvq_isolated =
> cvq_isolated && cvq_mq_isolated then? No map involved that way, and
> all parents should behave that way.
>
> > (or I wonder if we can meet this condition for any existing parents).
>
> I don't think so, but I think we need to probe the two anyway.
> Otherwise we may change the dataplane asid too.

Just to make sure we are at the same page, I meant we could fail the
initialization of vhost-vDPA is the device:

1) can isolate cvq in the case of singqueue but not multiqueue

or

2) can isolate cvq in the case of multiqueue but not single queue

Because I don't think there are any parents that have such a buggy
implementation.

Thanks

>
> Thanks!
>

Re: [PATCH for 8.1 v2 5/6] vdpa: move CVQ isolation check to net_init_vhost_vdpa

[Eugenio Perez Martin]

On Mon, Apr 3, 2023 at 7:32 AM Jason Wang <jasowang@redhat.com> wrote:
>
> On Fri, Mar 31, 2023 at 6:12 PM Eugenio Perez Martin
> <eperezma@redhat.com> wrote:
> >
> > On Fri, Mar 31, 2023 at 10:00 AM Jason Wang <jasowang@redhat.com> wrote:
> > >
> > >
> > > 在 2023/3/30 18:42, Eugenio Perez Martin 写道:
> > > > On Thu, Mar 30, 2023 at 8:23 AM Jason Wang <jasowang@redhat.com> wrote:
> > > >> On Thu, Mar 30, 2023 at 2:20 PM Jason Wang <jasowang@redhat.com> wrote:
> > > >>> On Fri, Mar 24, 2023 at 3:54 AM Eugenio Pérez <eperezma@redhat.com> wrote:
> > > >>>> Evaluating it at start time instead of initialization time may make the
> > > >>>> guest capable of dynamically adding or removing migration blockers.
> > > >>>>
> > > >>>> Also, moving to initialization reduces the number of ioctls in the
> > > >>>> migration, reducing failure possibilities.
> > > >>>>
> > > >>>> As a drawback we need to check for CVQ isolation twice: one time with no
> > > >>>> MQ negotiated and another one acking it, as long as the device supports
> > > >>>> it.  This is because Vring ASID / group management is based on vq
> > > >>>> indexes, but we don't know the index of CVQ before negotiating MQ.
> > > >>> We need to fail if we see a device that can isolate cvq without MQ but
> > > >>> not with MQ.
> > > >>>
> > > >>>> Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
> > > >>>> ---
> > > >>>> v2: Take out the reset of the device from vhost_vdpa_cvq_is_isolated
> > > >>>> ---
> > > >>>>   net/vhost-vdpa.c | 194 ++++++++++++++++++++++++++++++++++++-----------
> > > >>>>   1 file changed, 151 insertions(+), 43 deletions(-)
> > > >>>>
> > > >>>> diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c
> > > >>>> index 4397c0d4b3..db2c9afcb3 100644
> > > >>>> --- a/net/vhost-vdpa.c
> > > >>>> +++ b/net/vhost-vdpa.c
> > > >>>> @@ -43,6 +43,13 @@ typedef struct VhostVDPAState {
> > > >>>>
> > > >>>>       /* The device always have SVQ enabled */
> > > >>>>       bool always_svq;
> > > >>>> +
> > > >>>> +    /* The device can isolate CVQ in its own ASID if MQ is negotiated */
> > > >>>> +    bool cvq_isolated_mq;
> > > >>>> +
> > > >>>> +    /* The device can isolate CVQ in its own ASID if MQ is not negotiated */
> > > >>>> +    bool cvq_isolated;
> > > >>> As stated above, if we need a device that cvq_isolated_mq^cvq_isolated
> > > >>> == true, we need to fail. This may reduce the complexity of the code?
> > > >>>
> > > >>> Thanks
> > > >> Since we are the mediation layer, Qemu can alway choose to negotiate
> > > >> MQ regardless whether or not it is supported by the guest. In this
> > > >> way, we can have a stable virtqueue index for cvq.
> > > >>
> > > > I think it is a great idea and it simplifies this patch somehow.
> > > > However, we need something like the queue mapping [1] to do so :).
> > > >
> > > > To double confirm:
> > > > * If the device supports MQ, only probe MQ. If not, only probe !MQ.
> > > > * Only store cvq_isolated in VhostVDPAState.
> > > >
> > > > Now, if the device does not negotiate MQ but the device supports MQ:
> > >
> > >
> > > I'm not sure I understand here, if device supports MQ it should accepts
> > > MQ or we can fail the initialization here.
> > >
> >
> > My fault, I wanted to say "if the device offers MQ but the driver does
> > not acks it".
> >
> > >
> > > > * All the requests to queue 3 must be redirected to the last queue in
> > > > the device. That includes set_vq_address, notifiers regions, etc.
> > >
> > >
> > > This also means we will only mediate the case:
> > >
> > > 1) Qemu emulated virtio-net has 1 queue but device support multiple queue
> > >
> > > but not
> > >
> > > 2) Qemu emulated virtio-net has M queue but device support N queue (N>M)
> > >
> >
> > Right.
> >
> > >
> > > >
> > > > I'm totally ok to go this route but it's not immediate.
> > >
> > >
> > > Yes but I mean, we can start from failing the device if
> > > cvq_isolated_mq^cvq_isolated == true
> > >
> >
> > So probe the two cases but set VhostVDPAState->cvq_isolated =
> > cvq_isolated && cvq_mq_isolated then? No map involved that way, and
> > all parents should behave that way.
> >
> > > (or I wonder if we can meet this condition for any existing parents).
> >
> > I don't think so, but I think we need to probe the two anyway.
> > Otherwise we may change the dataplane asid too.
>
> Just to make sure we are at the same page, I meant we could fail the
> initialization of vhost-vDPA is the device:
>
> 1) can isolate cvq in the case of singqueue but not multiqueue
>
> or
>
> 2) can isolate cvq in the case of multiqueue but not single queue
>
> Because I don't think there are any parents that have such a buggy
> implementation.
>

Got it.

Leaving out the queue multiplex for the moment, as it adds complexity
and we can add it on top.

Thanks!

Re: [PATCH for 8.1 v2 2/6] vdpa: add vhost_vdpa_reset_status_fd

[Michael S. Tsirkin]

On Tue, Mar 28, 2023 at 09:37:16AM +0200, Eugenio Perez Martin wrote:
> > > diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c
> > > index bbabea18f3..7a2053b8d9 100644
> > > --- a/hw/virtio/vhost-vdpa.c
> > > +++ b/hw/virtio/vhost-vdpa.c
> > > @@ -335,38 +335,45 @@ static const MemoryListener vhost_vdpa_memory_listener = {
> > >      .region_del = vhost_vdpa_listener_region_del,
> > >  };
> > >
> > > -static int vhost_vdpa_call(struct vhost_dev *dev, unsigned long int request,
> > > -                             void *arg)
> > > +static int vhost_vdpa_dev_fd(const struct vhost_dev *dev)
> > >  {
> > >      struct vhost_vdpa *v = dev->opaque;
> > > -    int fd = v->device_fd;
> > > -    int ret;
> > >
> > >      assert(dev->vhost_ops->backend_type == VHOST_BACKEND_TYPE_VDPA);
> > > +    return v->device_fd;
> > > +}
> > > +
> > > +static int vhost_vdpa_call_fd(int fd, unsigned long int request, void *arg)
> > > +{
> > > +    int ret = ioctl(fd, request, arg);
> > >
> > > -    ret = ioctl(fd, request, arg);
> > >      return ret < 0 ? -errno : ret;
> > >  }
> > >
> > > -static int vhost_vdpa_add_status(struct vhost_dev *dev, uint8_t status)
> > > +static int vhost_vdpa_call(struct vhost_dev *dev, unsigned long int request,
> > > +                           void *arg)
> > > +{
> > > +    return vhost_vdpa_call_fd(vhost_vdpa_dev_fd(dev), request, arg);
> > > +}
> > > +
> > > +static int vhost_vdpa_add_status_fd(int fd, uint8_t status)
> > >  {
> > >      uint8_t s;
> > >      int ret;
> > >
> > > -    trace_vhost_vdpa_add_status(dev, status);
> > > -    ret = vhost_vdpa_call(dev, VHOST_VDPA_GET_STATUS, &s);
> >
> > We can stick a vhost_vdpa_call() function that calls
> > vhost_vdpa_call_fd() then we can avoid a lot of changes?
> >
> 
> I don't follow this. vhost_vdpa_call already calls vhost_vdpa_call_fd.
> 
> Thanks!


Jason were you going to reply?