* Eugenio Perez Martin (epere...@redhat.com) wrote: > On Tue, Jan 17, 2023 at 10:58 AM Dr. David Alan Gilbert > <dgilb...@redhat.com> wrote: > > > > * Eugenio Perez Martin (epere...@redhat.com) wrote: > > > On Fri, Jan 13, 2023 at 5:55 AM Jason Wang <jasow...@redhat.com> wrote: > > > > > > > > On Fri, Jan 13, 2023 at 1:25 AM Eugenio Pérez <epere...@redhat.com> > > > > wrote: > > > > > > > > > > This allows net to restart the device backend to configure SVQ on it. > > > > > > > > > > Ideally, these changes should not be net specific. However, the vdpa > > > > > net > > > > > backend is the one with enough knowledge to configure everything > > > > > because > > > > > of some reasons: > > > > > * Queues might need to be shadowed or not depending on its kind > > > > > (control > > > > > vs data). > > > > > * Queues need to share the same map translations (iova tree). > > > > > > > > > > Because of that it is cleaner to restart the whole net backend and > > > > > configure again as expected, similar to how vhost-kernel moves between > > > > > userspace and passthrough. > > > > > > > > > > If more kinds of devices need dynamic switching to SVQ we can create a > > > > > callback struct like VhostOps and move most of the code there. > > > > > VhostOps cannot be reused since all vdpa backend share them, and to > > > > > personalize just for networking would be too heavy. > > > > > > > > > > Signed-off-by: Eugenio Pérez <epere...@redhat.com> > > > > > --- > > > > > net/vhost-vdpa.c | 84 > > > > > ++++++++++++++++++++++++++++++++++++++++++++++++ > > > > > 1 file changed, 84 insertions(+) > > > > > > > > > > diff --git a/net/vhost-vdpa.c b/net/vhost-vdpa.c > > > > > index 5d7ad6e4d7..f38532b1df 100644 > > > > > --- a/net/vhost-vdpa.c > > > > > +++ b/net/vhost-vdpa.c > > > > > @@ -26,6 +26,8 @@ > > > > > #include <err.h> > > > > > #include "standard-headers/linux/virtio_net.h" > > > > > #include "monitor/monitor.h" > > > > > +#include "migration/migration.h" > > > > > +#include "migration/misc.h" > > > > > #include "migration/blocker.h" > > > > > #include "hw/virtio/vhost.h" > > > > > > > > > > @@ -33,6 +35,7 @@ > > > > > typedef struct VhostVDPAState { > > > > > NetClientState nc; > > > > > struct vhost_vdpa vhost_vdpa; > > > > > + Notifier migration_state; > > > > > Error *migration_blocker; > > > > > VHostNetState *vhost_net; > > > > > > > > > > @@ -243,10 +246,86 @@ static VhostVDPAState > > > > > *vhost_vdpa_net_first_nc_vdpa(VhostVDPAState *s) > > > > > return DO_UPCAST(VhostVDPAState, nc, nc0); > > > > > } > > > > > > > > > > +static void vhost_vdpa_net_log_global_enable(VhostVDPAState *s, bool > > > > > enable) > > > > > +{ > > > > > + struct vhost_vdpa *v = &s->vhost_vdpa; > > > > > + VirtIONet *n; > > > > > + VirtIODevice *vdev; > > > > > + int data_queue_pairs, cvq, r; > > > > > + NetClientState *peer; > > > > > + > > > > > + /* We are only called on the first data vqs and only if x-svq is > > > > > not set */ > > > > > + if (s->vhost_vdpa.shadow_vqs_enabled == enable) { > > > > > + return; > > > > > + } > > > > > + > > > > > + vdev = v->dev->vdev; > > > > > + n = VIRTIO_NET(vdev); > > > > > + if (!n->vhost_started) { > > > > > + return; > > > > > + } > > > > > + > > > > > + if (enable) { > > > > > + ioctl(v->device_fd, VHOST_VDPA_SUSPEND); > > > > > > > > Do we need to check if the device is started or not here? > > > > > > > > > > v->vhost_started is checked right above, right? > > > > > > > > + } > > > > > > > > I'm not sure I understand the reason for vhost_net_stop() after a > > > > VHOST_VDPA_SUSPEND. It looks to me those functions are duplicated. > > > > > > > > > > I think this is really worth exploring, and it would have been clearer > > > if I didn't squash the vhost_reset_status commit by mistake :). > > > > > > Looking at qemu master vhost.c:vhost_dev_stop: > > > if (hdev->vhost_ops->vhost_dev_start) { > > > hdev->vhost_ops->vhost_dev_start(hdev, false); > > > } > > > if (vrings) { > > > vhost_dev_set_vring_enable(hdev, false); > > > } > > > for (i = 0; i < hdev->nvqs; ++i) { > > > vhost_virtqueue_stop(hdev, > > > vdev, > > > hdev->vqs + i, > > > hdev->vq_index + i); > > > } > > > > > > Both vhost-used and vhost-vdpa set_status(0) at > > > ->vhost_dev_start(hdev, false). It cleans virtqueue state in vdpa so > > > they are not recoverable at vhost_virtqueue_stop->get_vring_base, and > > > I think it is too late for vdpa devices to change it. I guess > > > vhost-user devices do not lose the state there, but I did not test. > > > > > > I call VHOST_VDPA_SUSPEND here so vhost_vdpa_dev_start looks more > > > similar to vhost_user_dev_start. We can make > > > vhost_vdpa_dev_start(false) to suspend the device instead. But then we > > > need to reset it after getting the indexes. That's why I added > > > vhost_vdpa_reset_status, but I admit it is neither the cleanest > > > approach nor the best name to it. > > > > > > Adding Maxime, RFC here so we can make -vdpa and -user not to divert too > > > much. > > > > > > > > + data_queue_pairs = n->multiqueue ? n->max_queue_pairs : 1; > > > > > + cvq = virtio_vdev_has_feature(vdev, VIRTIO_NET_F_CTRL_VQ) ? > > > > > + n->max_ncs - n->max_queue_pairs : > > > > > 0; > > > > > + vhost_net_stop(vdev, n->nic->ncs, data_queue_pairs, cvq); > > > > > + > > > > > + peer = s->nc.peer; > > > > > + for (int i = 0; i < data_queue_pairs + cvq; i++) { > > > > > + VhostVDPAState *vdpa_state; > > > > > + NetClientState *nc; > > > > > + > > > > > + if (i < data_queue_pairs) { > > > > > + nc = qemu_get_peer(peer, i); > > > > > + } else { > > > > > + nc = qemu_get_peer(peer, n->max_queue_pairs); > > > > > + } > > > > > + > > > > > + vdpa_state = DO_UPCAST(VhostVDPAState, nc, nc); > > > > > + vdpa_state->vhost_vdpa.shadow_data = enable; > > > > > + > > > > > + if (i < data_queue_pairs) { > > > > > + /* Do not override CVQ shadow_vqs_enabled */ > > > > > + vdpa_state->vhost_vdpa.shadow_vqs_enabled = enable; > > > > > + } > > > > > + } > > > > > + > > > > > + r = vhost_net_start(vdev, n->nic->ncs, data_queue_pairs, cvq); > > > > > + if (unlikely(r < 0)) { > > > > > + error_report("unable to start vhost net: %s(%d)", > > > > > g_strerror(-r), -r); > > > > > + } > > > > > +} > > > > > + > > > > > +static void vdpa_net_migration_state_notifier(Notifier *notifier, > > > > > void *data) > > > > > +{ > > > > > + MigrationState *migration = data; > > > > > + VhostVDPAState *s = container_of(notifier, VhostVDPAState, > > > > > + migration_state); > > > > > + > > > > > + switch (migration->state) { > > > > > + case MIGRATION_STATUS_SETUP: > > > > > + vhost_vdpa_net_log_global_enable(s, true); > > > > > + return; > > > > > + > > > > > + case MIGRATION_STATUS_CANCELLING: > > > > > + case MIGRATION_STATUS_CANCELLED: > > > > > + case MIGRATION_STATUS_FAILED: > > > > > + vhost_vdpa_net_log_global_enable(s, false); > > > > > > > > Do we need to recover here? > > > > > > > > > > I may be missing something, but the device is fully reset and restored > > > in these cases. > > > > > > CCing Juan and D. Gilbert, a review would be appreciated to check if > > > this covers all the cases. > > > > I'm surprised I'm not seeing an entry for MIGRATION_STATUS_COMPLETED > > there. > > > > You might consider: > > if (migration_in_setup(s)) { > > vhost_vdpa_net_log_global_enable(s, true); > > } else if (migration_has_finished(s) || migration_has_failed(s)) { > > vhost_vdpa_net_log_global_enable(s, false); > > } > > > > Thank you very much for the input, I see this is definitely cleaner > than my proposal. > > Just for completion here I need to handle differently has_finished vs > has_failed because of recovery. This is easily achievable from your > snippet so thank you very much. > > > I'm not too sure what will happen in your world with postcopy; it's > > worth testing, just remember on the source you don't want to be changing > > guest memory when you're in the postcopy phase. > > > > If I'm not wrong postcopy is forbidden as long as there exists a vdpa > device but I can check it for sure.
Ah yes, we don't want the vdpa writing into the destination RAM during the postcopy phase; I can imagine with shadow-queues you might be able to come up with a solution to that - but that's a complication for another time. Dave > Thanks! > > > > Dave > > > > > Thanks! > > > > > > > > > > Thanks > > > > > > > > > + return; > > > > > + }; > > > > > +} > > > > > + > > > > > static void vhost_vdpa_net_data_start_first(VhostVDPAState *s) > > > > > { > > > > > struct vhost_vdpa *v = &s->vhost_vdpa; > > > > > > > > > > + if (v->feature_log) { > > > > > + add_migration_state_change_notifier(&s->migration_state); > > > > > + } > > > > > + > > > > > if (v->shadow_vqs_enabled) { > > > > > v->iova_tree = vhost_iova_tree_new(v->iova_range.first, > > > > > v->iova_range.last); > > > > > @@ -280,6 +359,10 @@ static void > > > > > vhost_vdpa_net_client_stop(NetClientState *nc) > > > > > > > > > > assert(nc->info->type == NET_CLIENT_DRIVER_VHOST_VDPA); > > > > > > > > > > + if (s->vhost_vdpa.index == 0 && s->vhost_vdpa.feature_log) { > > > > > + remove_migration_state_change_notifier(&s->migration_state); > > > > > + } > > > > > + > > > > > dev = s->vhost_vdpa.dev; > > > > > if (dev->vq_index + dev->nvqs == dev->vq_index_end) { > > > > > g_clear_pointer(&s->vhost_vdpa.iova_tree, > > > > > vhost_iova_tree_delete); > > > > > @@ -767,6 +850,7 @@ static NetClientState > > > > > *net_vhost_vdpa_init(NetClientState *peer, > > > > > s->vhost_vdpa.device_fd = vdpa_device_fd; > > > > > s->vhost_vdpa.index = queue_pair_index; > > > > > s->always_svq = svq; > > > > > + s->migration_state.notify = vdpa_net_migration_state_notifier; > > > > > s->vhost_vdpa.shadow_vqs_enabled = svq; > > > > > s->vhost_vdpa.iova_range = iova_range; > > > > > s->vhost_vdpa.shadow_data = svq; > > > > > -- > > > > > 2.31.1 > > > > > > > > > > > > > > > > > > -- > > Dr. David Alan Gilbert / dgilb...@redhat.com / Manchester, UK > > > -- Dr. David Alan Gilbert / dgilb...@redhat.com / Manchester, UK