Xen devel - Sep 2012 - [PATCH 1 of 6 V6] amd iommu: Add 2 hypercalls for libxc

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

>>> On 26.09.12 at 16:46, Wei Wang <wei.wang2@amd.com> wrote:
>-static unsigned int machine_bdf(struct domain *d, uint16_t guest_bdf)
>+static unsigned int machine_bdf(struct domain *d, uint16_t guest_seg,
>+                                uint16_t guest_bdf)
> {
>-    return guest_bdf;
>+    struct pci_dev *pdev;
>+    uint16_t mbdf = 0;
Is that really a suitable error indicator?
>@@ -207,7 +224,7 @@ void guest_iommu_add_ppr_log(struct domain *d, u32
entry[])
>     log = log_base + tail % (PAGE_SIZE / sizeof(ppr_entry_t));
> 
>     /* Convert physical device id back into virtual device id */
>-    gdev_id = guest_bdf(d, iommu_get_devid_from_cmd(entry[0]));
>+    gdev_id = guest_bdf(d, 0, iommu_get_devid_from_cmd(entry[0]));
Zero (here and below)? You''ll have to get the segment out of
the physical IOMMU''s control structure. Using plain zero is only
okay (for the time being) for guest devices (under the
assumption that they will all get surfaced on segment zero), i.e.
for calls to machine_bdf() (and you could shrink the patch by not
doing the adjustment there in the first place if only segment 0 is
possible right now).
>+/* Support for guest iommu emulation */
>+struct xen_domctl_guest_iommu_op {
>+    /* XEN_DOMCTL_GUEST_IOMMU_OP_* */
>+#define XEN_DOMCTL_GUEST_IOMMU_OP_SET_MSI               0
>+#define XEN_DOMCTL_GUEST_IOMMU_OP_BIND_BDF              1
>+    uint8_t op;
>+    union {
>+        struct iommu_msi {
>+            uint8_t     vector;
>+            uint8_t     dest;
For virtual x2APIC support, this ought to be 32 bits wide.
>+            uint8_t     dest_mode;
>+            uint8_t     delivery_mode;
>+            uint8_t     trig_mode;
>+        } msi;
Jan