hw/virtio: propagate vhost_virtqueue_mask() errors from start path
PATCH 1/1 hw/virtio: propagate vhost_virtqueue_mask() errors from start path
vhost_virtqueue_start() unmasks the call notifier by calling vhost_virtqueue_mask(), whose vhost_set_vring_call ioctl can fail (closed vhost-user socket, kernel ENOMEM, revoked guest_notifier fd) but whose void signature throws the error away. vhost_dev_start() then reports success while the backend has no valid call eventfd for that vq, leaving the guest with a working kick path but no virtqueue interrupts -- a half-up state harder to diagnose than a clean failure.
Make vhost_virtqueue_mask() return int and handle the error in the start path via the existing fail_vector unwind. Other callers reach the function through VirtioDeviceClass.guest_notifier_mask, whose void signature offers no upward error channel; they invoke it as a statement and remain source-compatible.
Resolves a long-standing TODO at the unmask call.
--- hw/virtio/vhost.c | 12 +++++++++--- include/hw/virtio/vhost.h | 4 ++-- 2 files changed, 11 insertions(+), 5 deletions(-)
hw/virtio/vhost.c +9 -3
@@ -1351,8 +1351,11 @@ int vhost_virtqueue_start(struct vhost_dev *dev,* will do it later. */ if (!vdev->use_guest_notifier_mask) { -/* TODO: check and handle errors. */ -vhost_virtqueue_mask (dev, vdev, idx, false); + r =vhost_virtqueue_mask (dev, vdev, idx, false); +if (r <0 ) { +/* No call eventfd means no interrupts; bail out. */ + goto fail_vector; + } } if (k->query_guest_notifiers &&
@@ -1815,7 +1818,7 @@ bool vhost_virtqueue_pending(struct vhost_dev *hdev, int n) }/* Mask/unmask events from this vq. */ -void vhost_virtqueue_mask (struct vhost_dev *hdev,VirtIODevice *vdev,int n, +int vhost_virtqueue_mask (struct vhost_dev *hdev,VirtIODevice *vdev,int n,bool mask) {struct VirtQueue *vvq =virtio_get_queue (vdev, n);
@@ -1836,7 +1839,10 @@ void vhost_virtqueue_mask(struct vhost_dev *hdev, VirtIODevice *vdev, int n, r = hdev->vhost_ops->vhost_set_vring_call (hdev, &file);if (r <0 ) {error_report ("vhost_set_vring_call failed %d" , -r); +return r; } + +return 0 ; }bool vhost_config_pending (struct vhost_dev *hdev)
include/hw/virtio/vhost.h +2 -2
@@ -317,8 +317,8 @@ bool vhost_virtqueue_pending(struct vhost_dev *hdev, int n);/* Mask/unmask events from this vq. */ -void vhost_virtqueue_mask (struct vhost_dev *hdev,VirtIODevice *vdev,int n, -bool mask); +int vhost_virtqueue_mask (struct vhost_dev *hdev,VirtIODevice *vdev,int n, +bool mask);/** * vhost_get_features_ex() - sanitize the extended features set
Signature · 1 line
2.50.1 (Apple Git-155)
Hi,
Quoted 30 lines · Bin Guo
vhost_virtqueue_start()unmasks the call notifier by callingvhost_virtqueue_mask(), whosevhost_set_vring_callioctl can fail (closed vhost-user socket, kernel ENOMEM, revokedguest_notifierfd) but whose void signature throws the error away.vhost_dev_start()then reports success while the backend has no valid call eventfd for that vq, leaving the guest with a working kick path but no virtqueue interrupts -- a half-up state harder to diagnose than a clean failure.Make
vhost_virtqueue_mask()return int and handle the error in the start path via the existingfail_vectorunwind. Other callers reach the function through VirtioDeviceClass.guest_notifier_mask, whose void signature offers no upward error channel; they invoke it as a statement and remain source-compatible.Resolves a long-standing TODO at the unmask call.
Signed-off-by Bin Guo <[email protected]>--- hw/virtio/vhost.c | 12 +++++++++--- include/hw/virtio/vhost.h | 4 ++-- 2 files changed, 11 insertions(+), 5 deletions(-)/* Mask/unmask events from this vq. */
hw/virtio/vhost.c +1
-1-void vhost_virtqueue_mask (struct vhost_dev *hdev,VirtIODevice *vdev,int n, +int vhost_virtqueue_mask (struct vhost_dev *hdev,VirtIODevice *vdev,int n,bool mask) {struct VirtQueue *vvq =virtio_get_queue (vdev, n);
hw/virtio/vhost.c void vhost_virtqueue_mask(struct vhost_dev *hdev, VirtIODevice *vdev, int n, · line 1841
Quoted 3 lines
hw/virtio/vhost.c +0
-0@@ -1836,7 +1839,10 @@ void vhost_virtqueue_mask(struct vhost_dev *hdev, VirtIODevice *vdev, int n, r = hdev->vhost_ops->vhost_set_vring_call (hdev, &file);if (r <0 ) {error_report ("vhost_set_vring_call failed %d" , -r);
IIUC the intention behind this TODO is to convert to "a function returning a boolean and taking Error** as last argument", replacing this call with error_setg_errno().
See in "Error reporting system loosely patterned after Glib's GError" rules in "qapi/error.h".
Quoted 5 lines
hw/virtio/vhost.c +3
-0+return r; } + +return 0 ; }
Quoted 5 lines · Philippe Mathieu-Daudé
IIUC the intention behind this TODO is to convert to a function returning a boolean and taking Error** as last argument, replacing this call with
error_setg_errno().
See in 'Error reporting system loosely patterned after Glib's GError' rules in 'qapi/error.h'.
Thanks for the suggestion. I considered changing the signature to bool + Error **errp, but decided to keep the current int + error_report() approach. vhost_set_vring_call is unlikely to fail in practice, so this is defensive programming rather than a regular error path. Changing the signature would propagate to ~15 call sites across multiple files for what is a rare edge case, which seems disproportionate. The patch already resolves the TODO and confines the change to the start path.
Quoted 30 lines · Bin Guo
vhost_virtqueue_start()unmasks the call notifier by callingvhost_virtqueue_mask(), whosevhost_set_vring_callioctl can fail (closed vhost-user socket, kernel ENOMEM, revokedguest_notifierfd) but whose void signature throws the error away.vhost_dev_start()then reports success while the backend has no valid call eventfd for that vq, leaving the guest with a working kick path but no virtqueue interrupts -- a half-up state harder to diagnose than a clean failure.Make
vhost_virtqueue_mask()return int and handle the error in the start path via the existingfail_vectorunwind. Other callers reach the function through VirtioDeviceClass.guest_notifier_mask, whose void signature offers no upward error channel; they invoke it as a statement and remain source-compatible.Resolves a long-standing TODO at the unmask call.
Signed-off-by Bin Guo <[email protected]>--- hw/virtio/vhost.c | 12 +++++++++--- include/hw/virtio/vhost.h | 4 ++-- 2 files changed, 11 insertions(+), 5 deletions(-)hw/virtio/vhost.c +0
-0@@ -1351,8 +1351,11 @@ int vhost_virtqueue_start(struct vhost_dev *dev,* will do it later. */ if (!vdev->use_guest_notifier_mask) {
hw/virtio/vhost.c int vhost_virtqueue_start(struct vhost_dev *dev, · line 1361
Quoted 10 lines
hw/virtio/vhost.c +5
-2-/* TODO: check and handle errors. */ -vhost_virtqueue_mask (dev, vdev, idx, false); + r =vhost_virtqueue_mask (dev, vdev, idx, false); +if (r <0 ) { +/* No call eventfd means no interrupts; bail out. */ + goto fail_vector; + } } if (k->query_guest_notifiers &&
Can you rebase please? Thanks!
Quoted 37 lines
hw/virtio/vhost.c +4
-1@@ -1815,7 +1818,7 @@ bool vhost_virtqueue_pending(struct vhost_dev *hdev, int n) }/* Mask/unmask events from this vq. */ -void vhost_virtqueue_mask (struct vhost_dev *hdev,VirtIODevice *vdev,int n, +int vhost_virtqueue_mask (struct vhost_dev *hdev,VirtIODevice *vdev,int n,bool mask) {struct VirtQueue *vvq =virtio_get_queue (vdev, n);@@ -1836,7 +1839,10 @@ void vhost_virtqueue_mask(struct vhost_dev *hdev, VirtIODevice *vdev, int n, r = hdev->vhost_ops->vhost_set_vring_call (hdev, &file);if (r <0 ) {error_report ("vhost_set_vring_call failed %d" , -r); +return r; } + +return 0 ; }bool vhost_config_pending (struct vhost_dev *hdev)include/hw/virtio/vhost.h +2
-2@@ -317,8 +317,8 @@ bool vhost_virtqueue_pending(struct vhost_dev *hdev, int n);/* Mask/unmask events from this vq. */ -void vhost_virtqueue_mask (struct vhost_dev *hdev,VirtIODevice *vdev,int n, -bool mask); +int vhost_virtqueue_mask (struct vhost_dev *hdev,VirtIODevice *vdev,int n, +bool mask);/** * vhost_get_features_ex() - sanitize the extended features set Signature · 1 line
2.50.1 (Apple Git-155)