Re: [Intel-gfx] [PATCH 00/10] Connect VFIO to IOMMUFD

[Nicolin Chen <nicolinc@nvidia.com>]

On Tue, Nov 01, 2022 at 11:04:38AM +0800, Yi Liu wrote:
> On 2022/11/1 07:24, Jason Gunthorpe wrote:
> > On Mon, Oct 31, 2022 at 08:25:39PM +0800, Yi Liu wrote:
> > > > There is something wrong with the test suite that it isn't covering
> > > > the above, I'm going to look into that today.
> > > 
> > > sounds to be the cause. I didn't see any significant change in vfio_main.c
> > > that may fail gvt. So should the iommufd changes. Then we will re-run the
> > > test after your update.:-)
> > 
> > I updated the github with all the changes made so far, it is worth
> > trying again!
> 
> gvt is still failing with below call trace in host side. vfio_unpin_pages()
> is still in problem. Any idea on it?

> [  206.464318] WARNING: CPU: 9 PID: 3362 at
> drivers/iommu/iommufd/device.c:591 iommufd_access_pin_pages+0x337/0x360

Judging from this WARNING, and since gvt (mdev) needs pin_pages(),
I assume this might be a fix, though Jason's latest change for the
iova_alignment seems to be added for CONFIG_IOMMUFD_TEST only.

------
diff --git a/drivers/vfio/iommufd.c b/drivers/vfio/iommufd.c
index 72a289c5f8c9..185075528d5e 100644
--- a/drivers/vfio/iommufd.c
+++ b/drivers/vfio/iommufd.c
@@ -120,6 +120,7 @@ static void vfio_emulated_unmap(void *data, unsigned long iova,
 }
 
 static const struct iommufd_access_ops vfio_user_ops = {
+	.needs_pin_pages = 1,
 	.unmap = vfio_emulated_unmap,
 };
 
------

Perhaps you can try it first to see if we can test the rest part of
the routine for now, till Jason acks tomorrow.

Thanks
Nic