On Tue, 2023-02-14 at 10:10 +0000, Paul Durrant wrote: > There's a nasty little semantic subtlety here which has been missed... > If the guest is not in longmode then the timer value is actually split > across param[0] and param[1] (so that it is still 64-bits).
I'll work this in. I should be able to exercise this with the XTF tests.
diff --git a/target/i386/kvm/xen-emu.c b/target/i386/kvm/xen-emu.c
index b537d03be7..4781b1fa97 100644
--- a/target/i386/kvm/xen-emu.c
+++ b/target/i386/kvm/xen-emu.c
@@ -1467,8 +1467,15 @@ static bool do_kvm_xen_handle_exit(X86CPU *cpu, struct
kvm_xen_exit *exit)
switch (code) {
case __HYPERVISOR_set_timer_op:
- return kvm_xen_hcall_set_timer_op(exit, cpu,
- exit->u.hcall.params[0]);
+ if (exit->u.hcall.longmode) {
+ return kvm_xen_hcall_set_timer_op(exit, cpu,
+ exit->u.hcall.params[0]);
+ } else {
+ /* In 32-bit mode, the 64-bit timer value is in two args. */
+ uint64_t val = ((uint64_t)exit->u.hcall.params[1]) << 32 |
+ (uint32_t)exit->u.hcall.params[0];
+ return kvm_xen_hcall_set_timer_op(exit, cpu, val);
+ }
case __HYPERVISOR_grant_table_op:
return kvm_xen_hcall_gnttab_op(exit, cpu, exit->u.hcall.params[0],
exit->u.hcall.params[1],
smime.p7s
Description: S/MIME cryptographic signature
