vhost/net: Clear the pending messages when the backend is removed
Subsystems (5)
cover vhost/net: Clear the pending messages when the backend is removed
When the vhost iotlb is used along with a guest virtual iommu and the guest gets rebooted, some MISS messages may have been recorded just before the reboot and spuriously executed by the virtual iommu after the reboot. This is due to the fact the pending messages are not cleared.
As vhost does not have any explicit reset user API, VHOST_NET_SET_BACKEND looks a reasonable point where to clear the pending messages, in case the backend is removed (fd = -1).
This version is a follow-up on the discussions held in [1].
The first patch removes an unused 'enabled' parameter in vhost_init_device_iotlb().
Best Regards
Eric
History:
[1] RFC: [RFC] vhost: Clear the pending messages on vhost_init_device_iotlb()
https://lore.kernel.org/all/[email protected]/
Eric Auger (2):
vhost: Remove the enabled parameter from vhost_init_device_iotlb vhost/net: Clear the pending messages when the backend is removed
drivers/vhost/net.c | 5 ++++- drivers/vhost/vhost.c | 5 +++-- drivers/vhost/vhost.h | 3 ++- 3 files changed, 9 insertions(+), 4 deletions(-)
Signature · 6 lines
2.31.1 _______________________________________________ Virtualization mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/virtualization
PATCH 1/2 vhost: Remove the enabled parameter from vhost_init_device_iotlb
The 'enabled' parameter is not used by the function. Remove it.
--- drivers/vhost/net.c | 2 +- drivers/vhost/vhost.c | 2 +- drivers/vhost/vhost.h | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-)
drivers/vhost/net.c +1 -1
@@ -1642,7 +1642,7 @@ static int vhost_net_set_features(struct vhost_net *n, u64 features) goto out_unlock;if ((features & (1ULL <<VIRTIO_F_ACCESS_PLATFORM ))) { -if (vhost_init_device_iotlb (&n->dev, true)) +if (vhost_init_device_iotlb (&n->dev)) goto out_unlock; }
drivers/vhost/vhost.c +1 -1
@@ -1729,7 +1729,7 @@ long vhost_vring_ioctl(struct vhost_dev *d, unsigned int ioctl, void __user *arg }EXPORT_SYMBOL_GPL (vhost_vring_ioctl); -int vhost_init_device_iotlb (struct vhost_dev *d,bool enabled) +int vhost_init_device_iotlb (struct vhost_dev *d) {struct vhost_iotlb *niotlb, *oiotlb;int i;
drivers/vhost/vhost.h +1 -1
@@ -221,7 +221,7 @@ ssize_t vhost_chr_read_iter(struct vhost_dev *dev, struct iov_iter *to, int noblock);ssize_t vhost_chr_write_iter (struct vhost_dev *dev,struct iov_iter *from); -int vhost_init_device_iotlb (struct vhost_dev *d,bool enabled); +int vhost_init_device_iotlb (struct vhost_dev *d);void vhost_iotlb_map_free (struct vhost_iotlb *iotlb,struct vhost_iotlb_map *map);
Signature · 6 lines
2.31.1 _______________________________________________ Virtualization mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/virtualization
PATCH 2/2 vhost/net: Clear the pending messages when the backend is removed
In reply to Eric AugerWhen the vhost iotlb is used along with a guest virtual iommu and the guest gets rebooted, some MISS messages may have been recorded just before the reboot and spuriously executed by the virtual iommu after the reboot.
As vhost does not have any explicit reset user API, VHOST_NET_SET_BACKEND looks a reasonable point where to clear the pending messages, in case the backend is removed.
Export vhost_clear_msg() and call it in vhost_net_set_backend() when fd == -1.
---
Without this patch, with QEMU virtio-iommu, on reboot, we get spurious messages such as
qemu-kvm: virtio_iommu_translate no mapping for 0xff732800 for sid=1536
--- drivers/vhost/net.c | 3 +++ drivers/vhost/vhost.c | 3 ++- drivers/vhost/vhost.h | 1 + 3 files changed, 6 insertions(+), 1 deletion(-)
drivers/vhost/net.c +3 -0
@@ -1511,6 +1511,9 @@ static long vhost_net_set_backend(struct vhost_net *n, unsigned index, int fd) nvq = &n->vqs[index];mutex_lock (&vq->mutex); +if (fd ==-1 ) +vhost_clear_msg (&n->dev); +/* Verify that ring has been setup correctly. */ if (!vhost_vq_access_ok (vq)) { r = -EFAULT ;
drivers/vhost/vhost.c +2 -1
@@ -661,7 +661,7 @@ void vhost_dev_stop(struct vhost_dev *dev) }EXPORT_SYMBOL_GPL (vhost_dev_stop); -static void vhost_clear_msg (struct vhost_dev *dev) +void vhost_clear_msg (struct vhost_dev *dev) {struct vhost_msg_node *node, *n;
@@ -679,6 +679,7 @@ static void vhost_clear_msg(struct vhost_dev *dev)spin_unlock (&dev->iotlb_lock); } +EXPORT_SYMBOL_GPL (vhost_clear_msg);void vhost_dev_cleanup (struct vhost_dev *dev) {
drivers/vhost/vhost.h +1 -0
@@ -181,6 +181,7 @@ long vhost_dev_ioctl(struct vhost_dev *, unsigned int ioctl, void __user *argp);long vhost_vring_ioctl (struct vhost_dev *d,unsigned int ioctl,void __user *argp);bool vhost_vq_access_ok (struct vhost_virtqueue *vq);bool vhost_log_access_ok (struct vhost_dev *); +void vhost_clear_msg (struct vhost_dev *dev); int vhost_get_vq_desc(struct vhost_virtqueue *, struct iovec iov[], unsigned int iov_count,
Signature · 6 lines
2.31.1 _______________________________________________ Virtualization mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/virtualization
Hi,
Quoted 6 lines · Eric Auger
When the vhost iotlb is used along with a guest virtual iommu and the guest gets rebooted, some MISS messages may have been recorded just before the reboot and spuriously executed by the virtual iommu after the reboot. This is due to the fact the pending messages are not cleared.
Quoted 8 lines
As vhost does not have any explicit reset user API,
VHOST_NET_SET_BACKENDlooks a reasonable point where to clear the pending messages, in case the backend is removed (fd = -1).This version is a follow-up on the discussions held in [1].
The first patch removes an unused 'enabled' parameter in
vhost_init_device_iotlb().
Gentle Ping. Does it look a reasonable fix now?
Thanks
Eric
Quoted 9 lines
Best Regards
Eric
History:
[1] RFC: [RFC] vhost: Clear the pending messages onvhost_init_device_iotlb()
https://lore.kernel.org/all/[email protected]/
Quoted 9 lines
Eric Auger (2):
vhost: Remove the enabled parameter fromvhost_init_device_iotlbvhost/net: Clear the pending messages when the backend is removeddrivers/vhost/net.c | 5 ++++- drivers/vhost/vhost.c | 5 +++-- drivers/vhost/vhost.h | 3 ++- 3 files changed, 9 insertions(+), 4 deletions(-)
_______________________________________________
Virtualization mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/virtualization
Michael S. Tsirkin
maintainerQuoted 15 lines · Eric Auger
Hi,
Quoted 12 lines · Eric Auger
When the vhost iotlb is used along with a guest virtual iommu and the guest gets rebooted, some MISS messages may have been recorded just before the reboot and spuriously executed by the virtual iommu after the reboot. This is due to the fact the pending messages are not cleared.
As vhost does not have any explicit reset user API,
VHOST_NET_SET_BACKENDlooks a reasonable point where to clear the pending messages, in case the backend is removed (fd = -1).This version is a follow-up on the discussions held in [1].
Quoted 8 lines
Quoted 2 lines
The first patch removes an unused 'enabled' parameter in
vhost_init_device_iotlb().Gentle Ping. Does it look a reasonable fix now?
Thanks
Eric
Yes I applied this - giving it a bit of time in next.
Quoted 13 lines
Quoted 13 lines
Best Regards
Eric
History:
[1] RFC: [RFC] vhost: Clear the pending messages onvhost_init_device_iotlb()
https://lore.kernel.org/all/[email protected]/Eric Auger (2):
vhost: Remove the enabled parameter fromvhost_init_device_iotlbvhost/net: Clear the pending messages when the backend is removed
Quoted 5 lines
Quoted 5 lines
drivers/vhost/net.c | 5 ++++- drivers/vhost/vhost.c | 5 +++-- drivers/vhost/vhost.h | 3 ++- 3 files changed, 9 insertions(+), 4 deletions(-)
_______________________________________________
Virtualization mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/virtualization