On Tue, Mar 22, 2022 at 03:54:32PM +0000, Alex Bennée wrote: > > "Michael S. Tsirkin" <m...@redhat.com> writes: > > > On Mon, Mar 21, 2022 at 03:30:36PM +0000, Alex Bennée wrote: > >> Previously we would silently suppress VHOST_USER_PROTOCOL_F_CONFIG > >> during the protocol negotiation if the QEMU stub hadn't implemented > >> the vhost_dev_config_notifier. However this isn't the only way we can > >> handle config messages, the existing vdc->get/set_config can do this > >> as well. > > > > Could you give an example where the problem is encountered please? > > Well I only came across it when I realised by rpmb implementation wasn't > sending config updates to the vhost-user daemon because it only > implemented set/get_config methods. > > I think vhost-user-scsi suffers from this but I'm not able to test it as > it's not clear how to do it. The vhost-user-scsi daemon want to attach > to real SCSI devices rather than a file for the block device. I guess I > need to somehow add "fake" scsi nodes to my host system which SCSI > commands can be passed to? > > When was the last time vhost-user-scsi was tested with anything?
It's for passthrough mostly, yes. I think you can use tcm_loop on top of loopback on top of a file. Didn't try myself. > >> Lightly re-factor the code to check for both potential methods and > >> instead of silently squashing the feature error out. It is unlikely > >> that a vhost-user backend expecting to handle CONFIG messages will > >> behave correctly if they never get sent. > > > > Hmm but are you sure? Most devices work mostly fine without CONFIG > > messages, there's a chance a backend set this flag just in case > > without much thought ... > > But that would be a bug right? Certainly a mismatch if something really > does want to see config messages. Again RPMB needs this because it's the > vhost-user daemon that knows the size of the device. Could you pls clarify how do you create a config that's broken with RPMB? > >> Fixes: 1c3e5a2617 ("vhost-user: back SET/GET_CONFIG requests with a > >> protocol feature") > > > > I'm not sure whether something is broken or this is a cleanup patch. > > Fixes tag means "if you have 1c3e5a2617 you should pick this patch", so > > cleanups don't need a fixes: tag. > > No I think it's broken, we just didn't notice because as you say most > devices don't need to care. I was mostly asking about the fixes tag. If no existing user cares then it's best to avoid the fixes tag, it's a hint for backporters which versions need the patch. > > > > > >> Cc: Maxime Coquelin <maxime.coque...@redhat.com> > >> Cc: Michael S. Tsirkin <m...@redhat.com> > >> Cc: Stefan Hajnoczi <stefa...@redhat.com> > >> Signed-off-by: Alex Bennée <alex.ben...@linaro.org> > >> > >> --- > >> - we can't check for get_config/set_config as the stack squashed vdev > >> - use vhost-user-state to transmit this > >> --- > >> include/hw/virtio/vhost-user.h | 1 + > >> hw/scsi/vhost-user-scsi.c | 1 + > >> hw/virtio/vhost-user.c | 46 ++++++++++++++++++++++++---------- > >> 3 files changed, 35 insertions(+), 13 deletions(-) > >> > >> diff --git a/include/hw/virtio/vhost-user.h > >> b/include/hw/virtio/vhost-user.h > >> index e44a41bb70..6e0e8a71a3 100644 > >> --- a/include/hw/virtio/vhost-user.h > >> +++ b/include/hw/virtio/vhost-user.h > >> @@ -22,6 +22,7 @@ typedef struct VhostUserState { > >> CharBackend *chr; > >> VhostUserHostNotifier notifier[VIRTIO_QUEUE_MAX]; > >> int memory_slots; > >> + bool supports_config; > >> } VhostUserState; > >> > >> bool vhost_user_init(VhostUserState *user, CharBackend *chr, Error > >> **errp); > >> diff --git a/hw/scsi/vhost-user-scsi.c b/hw/scsi/vhost-user-scsi.c > >> index 1b2f7eed98..9be21d07ee 100644 > >> --- a/hw/scsi/vhost-user-scsi.c > >> +++ b/hw/scsi/vhost-user-scsi.c > >> @@ -121,6 +121,7 @@ static void vhost_user_scsi_realize(DeviceState *dev, > >> Error **errp) > >> vsc->dev.backend_features = 0; > >> vqs = vsc->dev.vqs; > >> > >> + s->vhost_user.supports_config = true; > >> ret = vhost_dev_init(&vsc->dev, &s->vhost_user, > >> VHOST_BACKEND_TYPE_USER, 0, errp); > >> if (ret < 0) { > >> diff --git a/hw/virtio/vhost-user.c b/hw/virtio/vhost-user.c > >> index b27b8c56e2..6ce082861b 100644 > >> --- a/hw/virtio/vhost-user.c > >> +++ b/hw/virtio/vhost-user.c > >> @@ -1949,14 +1949,15 @@ static int > >> vhost_user_postcopy_notifier(NotifierWithReturn *notifier, > >> static int vhost_user_backend_init(struct vhost_dev *dev, void *opaque, > >> Error **errp) > >> { > >> - uint64_t features, protocol_features, ram_slots; > >> + uint64_t features, ram_slots; > >> struct vhost_user *u; > >> + VhostUserState *vus = (VhostUserState *) opaque; > >> int err; > >> > >> assert(dev->vhost_ops->backend_type == VHOST_BACKEND_TYPE_USER); > >> > >> u = g_new0(struct vhost_user, 1); > >> - u->user = opaque; > >> + u->user = vus; > >> u->dev = dev; > >> dev->opaque = u; > >> > >> @@ -1967,6 +1968,10 @@ static int vhost_user_backend_init(struct vhost_dev > >> *dev, void *opaque, > >> } > >> > >> if (virtio_has_feature(features, VHOST_USER_F_PROTOCOL_FEATURES)) { > >> + bool supports_f_config = vus->supports_config || > >> + (dev->config_ops && > >> dev->config_ops->vhost_dev_config_notifier); > >> + uint64_t protocol_features; > >> + > >> dev->backend_features |= 1ULL << VHOST_USER_F_PROTOCOL_FEATURES; > >> > >> err = vhost_user_get_u64(dev, VHOST_USER_GET_PROTOCOL_FEATURES, > >> @@ -1976,19 +1981,34 @@ static int vhost_user_backend_init(struct > >> vhost_dev *dev, void *opaque, > >> return -EPROTO; > >> } > >> > >> - dev->protocol_features = > >> - protocol_features & VHOST_USER_PROTOCOL_FEATURE_MASK; > >> - > >> - if (!dev->config_ops || > >> !dev->config_ops->vhost_dev_config_notifier) { > >> - /* Don't acknowledge CONFIG feature if device doesn't support > >> it */ > >> - dev->protocol_features &= ~(1ULL << > >> VHOST_USER_PROTOCOL_F_CONFIG); > >> - } else if (!(protocol_features & > >> - (1ULL << VHOST_USER_PROTOCOL_F_CONFIG))) { > >> - error_setg(errp, "Device expects VHOST_USER_PROTOCOL_F_CONFIG > >> " > >> - "but backend does not support it."); > >> - return -EINVAL; > >> + /* > >> + * We will use all the protocol features we support - although > >> + * we suppress F_CONFIG if we know QEMUs internal code can not > >> support > >> + * it. > >> + */ > >> + protocol_features &= VHOST_USER_PROTOCOL_FEATURE_MASK; > >> + > >> + if (supports_f_config) { > >> + if (!virtio_has_feature(protocol_features, > >> + VHOST_USER_PROTOCOL_F_CONFIG)) { > >> + error_setg(errp, "vhost-user device %s expecting " > >> + "VHOST_USER_PROTOCOL_F_CONFIG but the > >> vhost-user backend does " > >> + "not support it.", dev->vdev->name); > >> + return -EPROTO; > >> + } > >> + } else { > >> + if (virtio_has_feature(protocol_features, > >> + VHOST_USER_PROTOCOL_F_CONFIG)) { > >> + warn_reportf_err(*errp, "vhost-user backend supports " > >> + "VHOST_USER_PROTOCOL_F_CONFIG for " > >> + "device %s but QEMU does not.", > >> + dev->vdev->name); > >> + protocol_features &= ~(1ULL << > >> VHOST_USER_PROTOCOL_F_CONFIG); > >> + } > >> } > >> > >> + /* final set of protocol features */ > >> + dev->protocol_features = protocol_features; > >> err = vhost_user_set_protocol_features(dev, > >> dev->protocol_features); > >> if (err < 0) { > >> error_setg_errno(errp, EPROTO, "vhost_backend_init failed"); > >> -- > >> 2.30.2 > > > -- > Alex Bennée