On 01/12/2018 11:13 AM, Zheng Xiang wrote:
When guest driver only setup part of queues declared in QEMU, it
would corrupt guest's physical address 0 when using uninitialized
queues in vhost_virtqueue_start.
In AARCH64 virtual machines, the address of system memory starts at
0x40000000 and the address of rom starts at 0. So, when using qemu
with vhost-scsi, it will fail with below error:
qemu-kvm: Error start vhost dev
qemu-kvm: unable to start vhost-scsi: Cannot allocate memory
This patch fix this issue by skipping calling vhost_virtqueue_start
for uninitialized queues.
Cc: Michael S. Tsirkin <m...@redhat.com>
Signed-off-by: Zheng Xiang <zhengxia...@huawei.com>
---
hw/virtio/vhost.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/hw/virtio/vhost.c b/hw/virtio/vhost.c
index e4290ce..ac79ffd 100644
--- a/hw/virtio/vhost.c
+++ b/hw/virtio/vhost.c
@@ -1532,6 +1532,8 @@ int vhost_dev_start(struct vhost_dev *hdev, VirtIODevice
*vdev)
goto fail_mem;
}
for (i = 0; i < hdev->nvqs; ++i) {
+ if (virtio_queue_get_desc_addr(vdev, hdev->vq_index + i) == 0)
+ continue;
r = vhost_virtqueue_start(hdev,
vdev,
hdev->vqs + i,
Thanks, it fixes the silent corruption that happens with vhost-user net
backend and Windows guests, when the number of queues pairs declared in
Qemu is higher than the number of vcpus.
Tested-by: Maxime Coquelin <maxime.coque...@redhat.com>
Maxime
