From: Michael Roth <michael.roth@amd.com>

KVM_HC_MAP_GPA_RANGE will be used to send requests to userspace for
private/shared memory attribute updates requested by the guest.
Implement handling for that use-case along with some basic
infrastructure for enabling specific hypercall events.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Signed-off-by: Pankaj Gupta <pankaj.gupta@amd.com>
---
 target/i386/kvm/kvm.c        | 55 ++++++++++++++++++++++++++++++++++++
 target/i386/kvm/kvm_i386.h   |  1 +
 target/i386/kvm/trace-events |  1 +
 3 files changed, 57 insertions(+)

diff --git a/target/i386/kvm/kvm.c b/target/i386/kvm/kvm.c
index 6c864e4611..e72c295f77 100644
--- a/target/i386/kvm/kvm.c
+++ b/target/i386/kvm/kvm.c
@@ -21,6 +21,7 @@
 #include <sys/syscall.h>
 
 #include <linux/kvm.h>
+#include <linux/kvm_para.h>
 #include "standard-headers/asm-x86/kvm_para.h"
 #include "hw/xen/interface/arch-x86/cpuid.h"
 
@@ -208,6 +209,13 @@ int kvm_get_vm_type(MachineState *ms)
     return kvm_type;
 }
 
+bool kvm_enable_hypercall(uint64_t enable_mask)
+{
+    KVMState *s = KVM_STATE(current_accel());
+
+    return !kvm_vm_enable_cap(s, KVM_CAP_EXIT_HYPERCALL, 0, enable_mask);
+}
+
 bool kvm_has_smm(void)
 {
     return kvm_vm_check_extension(kvm_state, KVM_CAP_X86_SMM);
@@ -5321,6 +5329,50 @@ static bool host_supports_vmx(void)
     return ecx & CPUID_EXT_VMX;
 }
 
+/*
+ * Currently the handling here only supports use of KVM_HC_MAP_GPA_RANGE
+ * to service guest-initiated memory attribute update requests so that
+ * KVM_SET_MEMORY_ATTRIBUTES can update whether or not a page should be
+ * backed by the private memory pool provided by guest_memfd, and as such
+ * is only applicable to guest_memfd-backed guests (e.g. SNP/TDX).
+ *
+ * Other other use-cases for KVM_HC_MAP_GPA_RANGE, such as for SEV live
+ * migration, are not implemented here currently.
+ *
+ * For the guest_memfd use-case, these exits will generally be synthesized
+ * by KVM based on platform-specific hypercalls, like GHCB requests in the
+ * case of SEV-SNP, and not issued directly within the guest though the
+ * KVM_HC_MAP_GPA_RANGE hypercall. So in this case, KVM_HC_MAP_GPA_RANGE is
+ * not actually advertised to guests via the KVM CPUID feature bit, as
+ * opposed to SEV live migration where it would be. Since it is unlikely the
+ * SEV live migration use-case would be useful for guest-memfd backed guests,
+ * because private/shared page tracking is already provided through other
+ * means, these 2 use-cases should be treated as being mutually-exclusive.
+ */
+static int kvm_handle_hc_map_gpa_range(struct kvm_run *run)
+{
+    uint64_t gpa, size, attributes;
+
+    if (!machine_require_guest_memfd(current_machine))
+        return -EINVAL;
+
+    gpa = run->hypercall.args[0];
+    size = run->hypercall.args[1] * TARGET_PAGE_SIZE;
+    attributes = run->hypercall.args[2];
+
+    trace_kvm_hc_map_gpa_range(gpa, size, attributes, run->hypercall.flags);
+
+    return kvm_convert_memory(gpa, size, attributes & KVM_MAP_GPA_RANGE_ENCRYPTED);
+}
+
+static int kvm_handle_hypercall(struct kvm_run *run)
+{
+    if (run->hypercall.nr == KVM_HC_MAP_GPA_RANGE)
+        return kvm_handle_hc_map_gpa_range(run);
+
+    return -EINVAL;
+}
+
 #define VMX_INVALID_GUEST_STATE 0x80000021
 
 int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
@@ -5416,6 +5468,9 @@ int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
         ret = kvm_xen_handle_exit(cpu, &run->xen);
         break;
 #endif
+    case KVM_EXIT_HYPERCALL:
+        ret = kvm_handle_hypercall(run);
+        break;
     default:
         fprintf(stderr, "KVM: unknown exit reason %d\n", run->exit_reason);
         ret = -1;
diff --git a/target/i386/kvm/kvm_i386.h b/target/i386/kvm/kvm_i386.h
index 6b44844d95..34fc60774b 100644
--- a/target/i386/kvm/kvm_i386.h
+++ b/target/i386/kvm/kvm_i386.h
@@ -33,6 +33,7 @@
 bool kvm_has_smm(void);
 bool kvm_enable_x2apic(void);
 bool kvm_hv_vpindex_settable(void);
+bool kvm_enable_hypercall(uint64_t enable_mask);
 
 bool kvm_enable_sgx_provisioning(KVMState *s);
 bool kvm_hyperv_expand_features(X86CPU *cpu, Error **errp);
diff --git a/target/i386/kvm/trace-events b/target/i386/kvm/trace-events
index b365a8e8e2..74a6234ff7 100644
--- a/target/i386/kvm/trace-events
+++ b/target/i386/kvm/trace-events
@@ -5,6 +5,7 @@ kvm_x86_fixup_msi_error(uint32_t gsi) "VT-d failed to remap interrupt for GSI %"
 kvm_x86_add_msi_route(int virq) "Adding route entry for virq %d"
 kvm_x86_remove_msi_route(int virq) "Removing route entry for virq %d"
 kvm_x86_update_msi_routes(int num) "Updated %d MSI routes"
+kvm_hc_map_gpa_range(uint64_t gpa, uint64_t size, uint64_t attributes, uint64_t flags) "gpa 0x%" PRIx64 " size 0x%" PRIx64 " attributes 0x%" PRIx64 " flags 0x%" PRIx64
 
 # xen-emu.c
 kvm_xen_hypercall(int cpu, uint8_t cpl, uint64_t input, uint64_t a0, uint64_t a1, uint64_t a2, uint64_t ret) "xen_hypercall: cpu %d cpl %d input %" PRIu64 " a0 0x%" PRIx64 " a1 0x%" PRIx64 " a2 0x%" PRIx64" ret 0x%" PRIx64
-- 
2.34.1

On 5/30/2024 7:16 PM, Pankaj Gupta wrote:

[...]
> +/*
> + * Currently the handling here only supports use of KVM_HC_MAP_GPA_RANGE
> + * to service guest-initiated memory attribute update requests so that
> + * KVM_SET_MEMORY_ATTRIBUTES can update whether or not a page should be
> + * backed by the private memory pool provided by guest_memfd, and as such
> + * is only applicable to guest_memfd-backed guests (e.g. SNP/TDX).
> + *
> + * Other other use-cases for KVM_HC_MAP_GPA_RANGE, such as for SEV live
            ^
            extra "other"?
> + * migration, are not implemented here currently.
> + *
> + * For the guest_memfd use-case, these exits will generally be synthesized
> + * by KVM based on platform-specific hypercalls, like GHCB requests in the
> + * case of SEV-SNP, and not issued directly within the guest though the
> + * KVM_HC_MAP_GPA_RANGE hypercall. So in this case, KVM_HC_MAP_GPA_RANGE is
> + * not actually advertised to guests via the KVM CPUID feature bit, as
> + * opposed to SEV live migration where it would be. Since it is unlikely the
> + * SEV live migration use-case would be useful for guest-memfd backed guests,
> + * because private/shared page tracking is already provided through other
> + * means, these 2 use-cases should be treated as being mutually-exclusive.
> + */
> +static int kvm_handle_hc_map_gpa_range(struct kvm_run *run)
> +{
> +    uint64_t gpa, size, attributes;
> +
> +    if (!machine_require_guest_memfd(current_machine))
> +        return -EINVAL;
> +
> +    gpa = run->hypercall.args[0];
> +    size = run->hypercall.args[1] * TARGET_PAGE_SIZE;
> +    attributes = run->hypercall.args[2];
> +
> +    trace_kvm_hc_map_gpa_range(gpa, size, attributes, run->hypercall.flags);
> +
> +    return kvm_convert_memory(gpa, size, attributes & KVM_MAP_GPA_RANGE_ENCRYPTED);

run->hypercall.ret should be updated accordingly.
At least for successful case.
For failure case, QEMU will shutdown the VM, is it the expected behavior?


> +}
> +
> +static int kvm_handle_hypercall(struct kvm_run *run)
> +{
> +    if (run->hypercall.nr == KVM_HC_MAP_GPA_RANGE)
> +        return kvm_handle_hc_map_gpa_range(run);
> +
> +    return -EINVAL;
> +}
> +
>
[...]