From c1a1008685af0327d9d03f03d43bdb77e7af5bea Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Eugenio=20P=C3=A9rez?= Date: Thu, 15 Dec 2022 12:31:44 +0100 Subject: [PATCH] vdpa: always start CVQ in SVQ mode if possible MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Isolate control virtqueue in its own group, allowing to intercept control commands but letting dataplane run totally passthrough to the guest. Signed-off-by: Eugenio PĂ©rez Message-Id: <20221215113144.322011-13-eperezma@redhat.com> Reviewed-by: Michael S. Tsirkin Signed-off-by: Michael S. Tsirkin Acked-by: Jason Wang --- hw/virtio/vhost-vdpa.c | 3 +- net/vhost-vdpa.c | 110 ++++++++++++++++++++++++++++++++++++++++- 2 files changed, 111 insertions(+), 2 deletions(-) diff --git a/hw/virtio/vhost-vdpa.c b/hw/virtio/vhost-vdpa.c index 20d9575e4f..fd0c33b0e1 100644 --- a/hw/virtio/vhost-vdpa.c +++ b/hw/virtio/vhost-vdpa.c @@ -638,7 +638,8 @@ static int vhost_vdpa_set_backend_cap(struct vhost_dev *dev) { uint64_t features; uint64_t f = 0x1ULL << VHOST_BACKEND_F_IOTLB_MSG_V2 | - 0x1ULL << VHOST_BACKEND_F_IOTLB_BATCH; + 0x1ULL << VHOST_BACKEND_F_IOTLB_BATCH | + 0x1ULL << VHOST_BACKEND_F_IOTLB_ASID; int r; if (vhost_vdpa_call(dev, VHOST_GET_BACKEND_FEATURES, &features)) { diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c index 710c5efe96..d36664f33a 100644 --- a/net/vhost-vdpa.c +++ b/net/vhost-vdpa.c @@ -102,6 +102,8 @@ static const uint64_t vdpa_svq_device_features = BIT_ULL(VIRTIO_NET_F_RSC_EXT) | BIT_ULL(VIRTIO_NET_F_STANDBY); +#define VHOST_VDPA_NET_CVQ_ASID 1 + VHostNetState *vhost_vdpa_get_vhost_net(NetClientState *nc) { VhostVDPAState *s = DO_UPCAST(VhostVDPAState, nc, nc); @@ -243,6 +245,40 @@ static NetClientInfo net_vhost_vdpa_info = { .check_peer_type = vhost_vdpa_check_peer_type, }; +static int64_t vhost_vdpa_get_vring_group(int device_fd, unsigned vq_index) +{ + struct vhost_vring_state state = { + .index = vq_index, + }; + int r = ioctl(device_fd, VHOST_VDPA_GET_VRING_GROUP, &state); + + if (unlikely(r < 0)) { + error_report("Cannot get VQ %u group: %s", vq_index, + g_strerror(errno)); + return r; + } + + return state.num; +} + +static int vhost_vdpa_set_address_space_id(struct vhost_vdpa *v, + unsigned vq_group, + unsigned asid_num) +{ + struct vhost_vring_state asid = { + .index = vq_group, + .num = asid_num, + }; + int r; + + r = ioctl(v->device_fd, VHOST_VDPA_SET_GROUP_ASID, &asid); + if (unlikely(r < 0)) { + error_report("Can't set vq group %u asid %u, errno=%d (%s)", + asid.index, asid.num, errno, g_strerror(errno)); + } + return r; +} + static void vhost_vdpa_cvq_unmap_buf(struct vhost_vdpa *v, void *addr) { VhostIOVATree *tree = v->iova_tree; @@ -317,11 +353,75 @@ dma_map_err: static int vhost_vdpa_net_cvq_start(NetClientState *nc) { VhostVDPAState *s; - int r; + struct vhost_vdpa *v; + uint64_t backend_features; + int64_t cvq_group; + int cvq_index, r; assert(nc->info->type == NET_CLIENT_DRIVER_VHOST_VDPA); s = DO_UPCAST(VhostVDPAState, nc, nc); + v = &s->vhost_vdpa; + + v->shadow_data = s->always_svq; + v->shadow_vqs_enabled = s->always_svq; + s->vhost_vdpa.address_space_id = VHOST_VDPA_GUEST_PA_ASID; + + if (s->always_svq) { + /* SVQ is already configured for all virtqueues */ + goto out; + } + + /* + * 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 & VHOST_BACKEND_F_IOTLB_ASID) || + !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 (group == cvq_group) { + return 0; + } + } + + r = vhost_vdpa_set_address_space_id(v, cvq_group, VHOST_VDPA_NET_CVQ_ASID); + if (unlikely(r < 0)) { + return r; + } + + v->iova_tree = vhost_iova_tree_new(v->iova_range.first, + v->iova_range.last); + v->shadow_vqs_enabled = true; + s->vhost_vdpa.address_space_id = VHOST_VDPA_NET_CVQ_ASID; + +out: if (!s->vhost_vdpa.shadow_vqs_enabled) { return 0; } @@ -350,6 +450,14 @@ static void vhost_vdpa_net_cvq_stop(NetClientState *nc) if (s->vhost_vdpa.shadow_vqs_enabled) { vhost_vdpa_cvq_unmap_buf(&s->vhost_vdpa, s->cvq_cmd_out_buffer); vhost_vdpa_cvq_unmap_buf(&s->vhost_vdpa, s->status); + if (!s->always_svq) { + /* + * If only the CVQ is shadowed we can delete this safely. + * If all the VQs are shadows this will be needed by the time the + * device is started again to register SVQ vrings and similar. + */ + g_clear_pointer(&s->vhost_vdpa.iova_tree, vhost_iova_tree_delete); + } } }