On 12/06/2017 12:05 AM, Stefan Hajnoczi wrote:
On Tue, Dec 05, 2017 at 11:33:16AM +0800, Wei Wang wrote:
The vhost-pci driver uses the remote guest physical address to send/receive
packets from the remote guest, so when sending the ving info to the vhost-pci
device, send the guest physical adress directly.
Signed-off-by: Wei Wang <wei.w.w...@intel.com>
---
hw/virtio/vhost.c | 56 +++++++++++++++++++++++++++++++++++--------------------
1 file changed, 36 insertions(+), 20 deletions(-)
Can you do it inside vhost_memory_map()/vhost_memory_unmap() instead of
modifying callers?
Looks like vhost_dev_has_iommu() already takes this approach:
static void *vhost_memory_map(struct vhost_dev *dev, hwaddr addr,
hwaddr *plen, int is_write)
{
if (!vhost_dev_has_iommu(dev)) {
return cpu_physical_memory_map(addr, plen, is_write);
} else {
return (void *)(uintptr_t)addr;
}
}
static void vhost_memory_unmap(struct vhost_dev *dev, void *buffer,
hwaddr len, int is_write,
hwaddr access_len)
{
if (!vhost_dev_has_iommu(dev)) {
cpu_physical_memory_unmap(buffer, len, is_write, access_len);
}
}
Thanks for the reminder. I think this patch may be not needed with
adding the F_IOMMU_PLATFORM feature.
Best,
Wei
