On 19/02/2026 11:19, Ani Sinha wrote:
On Thu, Feb 19, 2026 at 4:15 PM Paul Durrant <[email protected]> wrote:
On 19/02/2026 10:31, Ani Sinha wrote:
On 19 Feb 2026, at 3:09 PM, Paul Durrant <[email protected]> wrote:
On 18/02/2026 11:42, Ani Sinha wrote:
On confidential guests KVM virtual machine file descriptor changes as a
part of the guest reset process. Xen capabilities needs to be re-initialized in
KVM against the new file descriptor.
Signed-off-by: Ani Sinha <[email protected]>
---
target/i386/kvm/xen-emu.c | 50 +++++++++++++++++++++++++++++++++++++--
1 file changed, 48 insertions(+), 2 deletions(-)
diff --git a/target/i386/kvm/xen-emu.c b/target/i386/kvm/xen-emu.c
index 52de019834..69527145eb 100644
--- a/target/i386/kvm/xen-emu.c
+++ b/target/i386/kvm/xen-emu.c
@@ -44,9 +44,12 @@
#include "xen-compat.h"
+NotifierWithReturn xen_vmfd_change_notifier;
+static bool hyperv_enabled;
static void xen_vcpu_singleshot_timer_event(void *opaque);
static void xen_vcpu_periodic_timer_event(void *opaque);
static int vcpuop_stop_singleshot_timer(CPUState *cs);
+static int do_initialize_xen_caps(KVMState *s, uint32_t hypercall_msr);
#ifdef TARGET_X86_64
#define hypercall_compat32(longmode) (!(longmode))
@@ -54,6 +57,30 @@ static int vcpuop_stop_singleshot_timer(CPUState *cs);
#define hypercall_compat32(longmode) (false)
#endif
+static int xen_handle_vmfd_change(NotifierWithReturn *n,
+ void *data, Error** errp)
+{
+ int ret;
+
+ /* we are not interested in pre vmfd change notification */
+ if (((VmfdChangeNotifier *)data)->pre) {
+ return 0;
+ }
+
+ ret = do_initialize_xen_caps(kvm_state, XEN_HYPERCALL_MSR);
+ if (ret < 0) {
+ return ret;
+ }
+
+ if (hyperv_enabled) {
+ ret = do_initialize_xen_caps(kvm_state, XEN_HYPERCALL_MSR_HYPERV);
+ if (ret < 0) {
+ return ret;
+ }
+ }
+ return 0;
This seems odd. Why use the hyperv_enabled boolean, rather than simply the msr
value, since when hyperv_enabled is set you will be calling
do_initialize_xen_caps() twice.
I am not sure of enabling capabilities for Xen. I assumed we need to call
kvm_xen_init() twice, once normally with XEN_HYPERCALL_MSR and if hyper is
enabled, again with XEN_HYPERCALL_MSR_HYPERV. Is that not the case? Is it one
or the other but not both? It seems kvm_arch_init() calls kvm_xen_init() once
with XEN_HYPERCALL_MSR and another time vcpu_arch_init() calls it again if
hyperv is enabled with XEN_HYPERCALL_MSR_HYPERV .
Yes, it has to be assumed that XEN_HYPERCALL_MSR is correct until
Hyper-V supported is enabled, which comes later, at which point the MSR
is changed. So you only need save the latest MSR value and use that in
xen_handle_vmfd_change().
ok hopefully this looks good
https://gitlab.com/anisinha/qemu/-/commit/7f7ba25151b6a658c54f95a370f1970c01a6269a
sending this out to minimize churn and to make v6 as close to the
merge worthy as possible.
Yeah, that looks better. I don't think you need to move the `int
xen_caps, ret;` line though so your patch can be even smaller AFAICS.
