On 4/30/24 12:06, Duan, Zhenzhong wrote:
-----Original Message-----
From: Cédric Le Goater <c...@redhat.com>
Subject: Re: [PATCH v3 11/19] backends/iommufd: Implement
HostIOMMUDeviceClass::check_cap() handler
On 4/29/24 08:50, Zhenzhong Duan wrote:
Suggested-by: Cédric Le Goater <c...@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.d...@intel.com>
---
backends/iommufd.c | 18 ++++++++++++++++++
1 file changed, 18 insertions(+)
diff --git a/backends/iommufd.c b/backends/iommufd.c
index d61209788a..28faec528e 100644
--- a/backends/iommufd.c
+++ b/backends/iommufd.c
@@ -233,6 +233,23 @@ int
iommufd_backend_get_device_info(IOMMUFDBackend *be, uint32_t devid,
return ret;
}
+static int hiod_iommufd_check_cap(HostIOMMUDevice *hiod, int cap,
Error **errp)
+{
+ switch (cap) {
+ case HOST_IOMMU_DEVICE_CAP_IOMMUFD:
+ return 1;
I don't understand this value.
1 means this host iommu device is attached to IOMMUFD backend,
or else 0 if attached to legacy backend.
Hmm, this looks hacky to me and it is not used anywhere in the patchset.
Let's reconsider when there is actually a use for it. Until then, please
drop. My feeling is that a new HostIOMMUDeviceClass handler/attributed
should be introduced instead.
Thanks,
C.
Strictly speaking, HOST_IOMMU_DEVICE_CAP_IOMMUFD is not a
hardware capability, I'm trying to put all(sw/hw) in CAPs checking
framework just like KVM<->qemu CAPs does.
Thanks
Zhenzhong
Thanks,
C.
+ default:
+ return host_iommu_device_check_cap_common(hiod, cap, errp);
+ }
+}
+
+static void hiod_iommufd_class_init(ObjectClass *oc, void *data)
+{
+ HostIOMMUDeviceClass *hioc = HOST_IOMMU_DEVICE_CLASS(oc);
+
+ hioc->check_cap = hiod_iommufd_check_cap;
+};
+
static const TypeInfo types[] = {
{
.name = TYPE_IOMMUFD_BACKEND,
@@ -251,6 +268,7 @@ static const TypeInfo types[] = {
.parent = TYPE_HOST_IOMMU_DEVICE,
.instance_size = sizeof(HostIOMMUDeviceIOMMUFD),
.class_size = sizeof(HostIOMMUDeviceIOMMUFDClass),
+ .class_init = hiod_iommufd_class_init,
.abstract = true,
}
};