Skip to contentSkip to conversation

TDX QEMU support

PATCH In review Xiaoyao Li 8 participants 120 messages Aug 18, 2023 – Oct 10, 2023

Subsystems (3)
lore
Share & tools
lore

Xiaoyao Li

author
#0 lore raw

v2 cover TDX QEMU support

This is v2 series of adding TDX suppot in QEMU.

This patch series aims to enable TDX support to allow creating and booting a TD (TDX VM) with QEMU. It needs to work with corresponding KVM v15 patch [1]. TDX related documents can be found in [2].

This series is based on QEMU gmem implemntation, which is posted at [3]. And This series is also available in github:
https://github.com/intel/qemu-tdx/tree/tdx-qemu-upstream-v2

This version aims to update the TDX QEMU side to match with latest TDX KVM side implementation, which expose gmem for private memory. This version is not targeted as the final version because how to support KVM gmem in QEMU is not finalized yet. Though, any review comment is welcomed.

[1] KVM TDX basic feature support v15
https://lore.kernel.org/kvm/[email protected]/

[2] https://www.intel.com/content/www/us/en/developer/articles/technical/intel-trust-domain-extensions.html

[3] https://lore.kernel.org/all/[email protected]/

== Limitation and future work ==
- Readonly memslot

TDX only support readonly (write protection) memslot for shared memory, but not for private memory. For simplicity, just mark readonly memslot not supported entirely for TDX.

- CPU model

We cannot create a TD with arbitrary CPU model like what for non-TDX VMs, because only a subset of features can be configured for TD.

- It's recommended to use '-cpu host' to create TD;
- '+feature/-feature' might not work as expected;

future work: To introduce specific CPU model for TDs and enhance +/-features
for TDs.

- gdb suppport

gdb support to debug a TD of off-debug mode is future work.

== Change history ==
Changes from v1:
[v1] https://lore.kernel.org/qemu-devel/[email protected]/

- Switch to KVM gmem interface for private memory;
- Add TDVMCALL and its sub leaves support;
- mark LMCE as unsupported for TD VM;
- bing back the support of mrconfigid/mrowner/mrownerconfig;
- update documentation;

Changes from RFC v4:
[RFC v4] https://lore.kernel.org/qemu-devel/[email protected]/

- Add 3 more patches(9, 10, 11) to improve the tdx_get_supported_cpuid();
- make attributes of object tdx-guest not settable by user;
- improve get_tdx_capabilities() by using a known starting value and limiting the loop with a known size;
- clarify why isa.bios needs to be skipped;
- remove the MMIO hob setup since OVMF sets them up itself;

Changes from RFC v3:
[RFC v3] https://lore.kernel.org/qemu-devel/[email protected]/

- Load TDVF with -bios interface;
- Adapt to KVM API changes;
- KVM_TDX_CAPABILITIES changes back to KVM-scope;
- struct kvm_tdx_init_vm changes;
- Define TDX_SUPPORTED_KVM_FEATURES;
- Drop the patch of introducing property sept-ve-disable since it's not public yet;
- some misc cleanups

Changes from RFC v2:
[RFC v2] https://lore.kernel.org/qemu-devel/[email protected]/

- Get vm-type from confidential-guest-support object type;
- Drop machine_init_done_late_notifiers;
- Refactor tdx_ioctl implementation;
- re-use existing pflash interface to load TDVF (i.e., OVMF binaries);
- introduce new date structure to track memory type instead of changing e820 table;
- Force smm to off for TDX VM;
- Drop the patches that suppress level-trigger/SMI/INIT/SIPI since KVM will ingore them;
- Add documentation;

Changes from RFC v1:
[RFC v1] https://lore.kernel.org/qemu-devel/[email protected]/

- suppress level trigger/SMI/INIT/SIPI related to IOAPIC.
- add VM attribute sha384 to TD measurement.
- guest TSC Hz specification

Chao Peng (1):
i386/tdx: register TDVF as private memory

Chenyi Qiang (2):
i386/tdx: register the fd read callback with the main loop to read the
quote data
i386/tdx: setup a timer for the qio channel

Isaku Yamahata (14):
i386/tdx: Make sept_ve_disable set by default
qom: implement property helper for sha384
i386/tdx: Allows mrconfigid/mrowner/mrownerconfig for TDX_INIT_VM i386/tdx: Create kvm gmem for TD
kvm/tdx: Don't complain when converting vMMIO region to shared kvm/tdx: Ignore memory conversion to shared of unassigned region i386/tdvf: Introduce function to parse TDVF metadata i386/tdx: Add TDVF memory via KVM_TDX_INIT_MEM_REGION i386/tdx: handle TDG.VP.VMCALL<SetupEventNotifyInterrupt> i386/tdx: handle TDG.VP.VMCALL<GetQuote>
i386/tdx: handle TDG.VP.VMCALL<MapGPA> hypercall i386/tdx: Limit the range size for MapGPA
hw/i386: add option to forcibly report edge trigger in acpi tables i386/tdx: Don't synchronize guest tsc for TDs

Sean Christopherson (2):
i386/kvm: Move architectural CPUID leaf generation to separate helper i386/tdx: Don't get/put guest state for TDX VMs

Xiaoyao Li (39):
*** HACK *** linux-headers: Update headers to pull in TDX API changes
i386: Introduce tdx-guest object
target/i386: Parse TDX vm type
target/i386: Introduce kvm_confidential_guest_init() i386/tdx: Implement tdx_kvm_init() to initialize TDX VM context i386/tdx: Get tdx_capabilities via KVM_TDX_CAPABILITIES i386/tdx: Introduce is_tdx_vm() helper and cache tdx_guest object i386/tdx: Adjust the supported CPUID based on TDX restrictions i386/tdx: Update tdx_cpuid_lookup[].tdx_fixed0/1 by

    tdx_caps.cpuid_config[]

i386/tdx: Integrate tdx_caps->xfam_fixed0/1 into tdx_cpuid_lookup i386/tdx: Integrate tdx_caps->attrs_fixed0/1 to tdx_cpuid_lookup
kvm: Introduce kvm_arch_pre_create_vcpu()
i386/tdx: Initialize TDX before creating TD vcpus i386/tdx: Add property sept-ve-disable for tdx-guest object i386/tdx: Wire CPU features up with attributes of TD guest i386/tdx: Validate TD attributes
i386/tdx: Implement user specified tsc frequency
i386/tdx: Set kvm_readonly_mem_enabled to false for TDX VM i386/tdx: Make memory type private by default
i386/tdx: Parse TDVF metadata for TDX VM
i386/tdx: Skip BIOS shadowing setup
i386/tdx: Don't initialize pc.rom for TDX VMs
i386/tdx: Track mem_ptr for each firmware entry of TDVF i386/tdx: Track RAM entries for TDX VM
headers: Add definitions from UEFI spec for volumes, resources, etc... i386/tdx: Setup the TD HOB list
memory: Introduce memory_region_init_ram_gmem()
i386/tdx: Call KVM_TDX_INIT_VCPU to initialize TDX vcpu i386/tdx: Finalize TDX VM
i386/tdx: Handle TDG.VP.VMCALL<REPORT_FATAL_ERROR> i386/tdx: Wire REPORT_FATAL_ERROR with GuestPanic facility i386/tdx: Disable SMM for TDX VMs
i386/tdx: Disable PIC for TDX VMs
i386/tdx: Don't allow system reset for TDX VMs
i386/tdx: LMCE is not supported for TDX
hw/i386: add eoi_intercept_unsupported member to X86MachineState i386/tdx: Only configure MSR_IA32_UCODE_REV in kvm_init_msrs() for TDs i386/tdx: Skip kvm_put_apicbase() for TDs
docs: Add TDX documentation

 accel/kvm/kvm-all.c                        |   55 +-
 configs/devices/i386-softmmu/default.mak   |    1 +
 docs/system/confidential-guest-support.rst |    1 +
 docs/system/i386/tdx.rst                   |  114 ++
 docs/system/target-i386.rst                |    1 +
 hw/i386/Kconfig                            |    6 +
 hw/i386/acpi-build.c                       |   99 +-
 hw/i386/acpi-common.c                      |   50 +-
 hw/i386/meson.build                        |    1 +
 hw/i386/pc.c                               |   21 +-
 hw/i386/pc_sysfw.c                         |    7 +
 hw/i386/tdvf-hob.c                         |  147 ++
 hw/i386/tdvf-hob.h                         |   24 +
 hw/i386/tdvf.c                             |  200 +++
 hw/i386/x86.c                              |   38 +-
 include/exec/memory.h                      |    6 +
 include/hw/i386/tdvf.h                     |   58 +
 include/hw/i386/x86.h                      |    1 +
 include/qom/object.h                       |   17 +
 include/standard-headers/uefi/uefi.h       |  198 +++
 include/sysemu/kvm.h                       |    3 +
 linux-headers/asm-x86/kvm.h                |   90 ++
 linux-headers/linux/kvm.h                  |   87 ++
 qapi/qom.json                              |   26 +
 qapi/run-state.json                        |   17 +-
 qom/object.c                               |   76 +
 softmmu/memory.c                           |   52 +
 softmmu/runstate.c                         |   49 +
 target/i386/cpu-internal.h                 |    9 +
 target/i386/cpu.c                          |   12 -
 target/i386/cpu.h                          |   21 +
 target/i386/kvm/kvm-cpu.c                  |    5 +
 target/i386/kvm/kvm.c                      |  586 ++++----
 target/i386/kvm/kvm_i386.h                 |    5 +
 target/i386/kvm/meson.build                |    2 +
 target/i386/kvm/tdx-stub.c                 |   22 +
 target/i386/kvm/tdx.c                      | 1543 ++++++++++++++++++++
 target/i386/kvm/tdx.h                      |   73 +
 target/i386/sev.c                          |    1 -
 target/i386/sev.h                          |    2 +
 40 files changed, 3382 insertions(+), 344 deletions(-)
 create mode 100644 docs/system/i386/tdx.rst
 create mode 100644 hw/i386/tdvf-hob.c
 create mode 100644 hw/i386/tdvf-hob.h
 create mode 100644 hw/i386/tdvf.c
 create mode 100644 include/hw/i386/tdvf.h
 create mode 100644 include/standard-headers/uefi/uefi.h
 create mode 100644 target/i386/kvm/tdx-stub.c
 create mode 100644 target/i386/kvm/tdx.c
 create mode 100644 target/i386/kvm/tdx.h
Signature · 1 line
2.34.1
author
#1 lore raw

PATCH v2 1/58 *** HACK *** linux-headers: Update headers to pull in TDX API changes

Pull in recent TDX updates, which are not backwards compatible.

It's just to make this series runnable. It will be updated by script

scripts/update-linux-headers.sh

once TDX support is upstreamed in linux kernel

---
 linux-headers/asm-x86/kvm.h | 90 +++++++++++++++++++++++++++++++++++++
 linux-headers/linux/kvm.h   | 87 +++++++++++++++++++++++++++++++++++
 2 files changed, 177 insertions(+)
linux-headers/asm-x86/kvm.h +90 -0
index 003fb745347c..4c3deb0e2a75 100644
@@ -562,5 +562,95 @@ struct kvm_pmu_event_filter {
 
 #define KVM_X86_DEFAULT_VM	0
 #define KVM_X86_SW_PROTECTED_VM	1
+#define KVM_X86_TDX_VM		2
+#define KVM_X86_SNP_VM		3
+
+/* Trust Domain eXtension sub-ioctl() commands. */
+enum kvm_tdx_cmd_id {
+	KVM_TDX_CAPABILITIES = 0,
+	KVM_TDX_INIT_VM,
+	KVM_TDX_INIT_VCPU,
+	KVM_TDX_INIT_MEM_REGION,
+	KVM_TDX_FINALIZE_VM,
+
+	KVM_TDX_CMD_NR_MAX,
+};
+
+struct kvm_tdx_cmd {
+	/* enum kvm_tdx_cmd_id */
+	__u32 id;
+	/* flags for sub-commend. If sub-command doesn't use this, set zero. */
+	__u32 flags;
+	/*
+	 * data for each sub-command. An immediate or a pointer to the actual
+	 * data in process virtual address.  If sub-command doesn't use it,
+	 * set zero.
+	 */
+	__u64 data;
+	/*
+	 * Auxiliary error code.  The sub-command may return TDX SEAMCALL
+	 * status code in addition to -Exxx.
+	 * Defined for consistency with struct kvm_sev_cmd.
+	 */
+	__u64 error;
+};
+
+struct kvm_tdx_cpuid_config {
+	__u32 leaf;
+	__u32 sub_leaf;
+	__u32 eax;
+	__u32 ebx;
+	__u32 ecx;
+	__u32 edx;
+};
+
+struct kvm_tdx_capabilities {
+	__u64 attrs_fixed0;
+	__u64 attrs_fixed1;
+	__u64 xfam_fixed0;
+	__u64 xfam_fixed1;
+#define TDX_CAP_GPAW_48	(1 << 0)
+#define TDX_CAP_GPAW_52	(1 << 1)
+	__u32 supported_gpaw;
+	__u32 padding;
+	__u64 reserved[251];
+
+	__u32 nr_cpuid_configs;
+	struct kvm_tdx_cpuid_config cpuid_configs[];
+};
+
+struct kvm_tdx_init_vm {
+	__u64 attributes;
+	__u64 mrconfigid[6];	/* sha384 digest */
+	__u64 mrowner[6];	/* sha384 digest */
+	__u64 mrownerconfig[6];	/* sha348 digest */
+	/*
+	 * For future extensibility to make sizeof(struct kvm_tdx_init_vm) = 8KB.
+	 * This should be enough given sizeof(TD_PARAMS) = 1024.
+	 * 8KB was chosen given because
+	 * sizeof(struct kvm_cpuid_entry2) * KVM_MAX_CPUID_ENTRIES(=256) = 8KB.
+	 */
+	__u64 reserved[1004];
+
+	/*
+	 * Call KVM_TDX_INIT_VM before vcpu creation, thus before
+	 * KVM_SET_CPUID2.
+	 * This configuration supersedes KVM_SET_CPUID2s for VCPUs because the
+	 * TDX module directly virtualizes those CPUIDs without VMM.  The user
+	 * space VMM, e.g. qemu, should make KVM_SET_CPUID2 consistent with
+	 * those values.  If it doesn't, KVM may have wrong idea of vCPUIDs of
+	 * the guest, and KVM may wrongly emulate CPUIDs or MSRs that the TDX
+	 * module doesn't virtualize.
+	 */
+	struct kvm_cpuid2 cpuid;
+};
+
+#define KVM_TDX_MEASURE_MEMORY_REGION	(1UL << 0)
+
+struct kvm_tdx_init_mem_region {
+	__u64 source_addr;
+	__u64 gpa;
+	__u64 nr_pages;
+};
 
 #endif /* _ASM_X86_KVM_H */
linux-headers/linux/kvm.h +87 -0

105 lines not on this page · raw message

Signature · 1 line
2.34.1
author
#2 lore raw

PATCH v2 2/58 i386: Introduce tdx-guest object

In reply to Xiaoyao Li

Introduce tdx-guest object which implements the interface of CONFIDENTIAL_GUEST_SUPPORT, and will be used to create TDX VMs (TDs) by

  qemu -machine ...,confidential-guest-support=tdx0	\
       -object tdx-guset,id=tdx0

It has only one property 'attributes' with fixed value 0 and not configurable so far.

---
changes from RFC-V4
- make @attributes not user-settable

---
 configs/devices/i386-softmmu/default.mak |  1 +
 hw/i386/Kconfig                          |  5 +++
 qapi/qom.json                            | 12 +++++++
 target/i386/kvm/meson.build              |  2 ++
 target/i386/kvm/tdx.c                    | 40 ++++++++++++++++++++++++
 target/i386/kvm/tdx.h                    | 19 +++++++++++
 6 files changed, 79 insertions(+)
 create mode 100644 target/i386/kvm/tdx.c
 create mode 100644 target/i386/kvm/tdx.h
configs/devices/i386-softmmu/default.mak +1 -0
index 598c6646dfc0..9b5ec59d65b0 100644
@@ -18,6 +18,7 @@
 #CONFIG_QXL=n
 #CONFIG_SEV=n
 #CONFIG_SGA=n
+#CONFIG_TDX=n
 #CONFIG_TEST_DEVICES=n
 #CONFIG_TPM_CRB=n
 #CONFIG_TPM_TIS_ISA=n
hw/i386/Kconfig +5 -0
index 9051083c1e78..929f6c3f0e85 100644
@@ -10,6 +10,10 @@ config SGX
     bool
     depends on KVM
 
+config TDX
+    bool
+    depends on KVM
+
 config PC
     bool
     imply APPLESMC
@@ -26,6 +30,7 @@ config PC
     imply QXL
     imply SEV
     imply SGX
+    imply TDX
     imply TEST_DEVICES
     imply TPM_CRB
     imply TPM_TIS_ISA
qapi/qom.json +12 -0
index e0b2044e3d20..2ca7ce7c0da5 100644
@@ -866,6 +866,16 @@
             'reduced-phys-bits': 'uint32',
             '*kernel-hashes': 'bool' } }
 
+##
+# @TdxGuestProperties:
+#
+# Properties for tdx-guest objects.
+#
+# Since: 8.2
+##
+{ 'struct': 'TdxGuestProperties',
+  'data': { }}
+
 ##
 # @ThreadContextProperties:
 #
@@ -944,6 +954,7 @@
     'sev-guest',
     'thread-context',
     's390-pv-guest',
+    'tdx-guest',
     'throttle-group',
     'tls-creds-anon',
     'tls-creds-psk',
@@ -1010,6 +1021,7 @@
       'secret_keyring':             { 'type': 'SecretKeyringProperties',
                                       'if': 'CONFIG_SECRET_KEYRING' },
       'sev-guest':                  'SevGuestProperties',
+      'tdx-guest':                  'TdxGuestProperties',
       'thread-context':             'ThreadContextProperties',
       'throttle-group':             'ThrottleGroupProperties',
       'tls-creds-anon':             'TlsCredsAnonProperties',
target/i386/kvm/meson.build +2 -0
index 40fbde96cac6..21ab03fe1349 100644
@@ -11,6 +11,8 @@ i386_softmmu_kvm_ss.add(when: 'CONFIG_XEN_EMU', if_true: files('xen-emu.c'))
 
 i386_softmmu_kvm_ss.add(when: 'CONFIG_SEV', if_false: files('sev-stub.c'))
 
+i386_softmmu_kvm_ss.add(when: 'CONFIG_TDX', if_true: files('tdx.c'))
+
 i386_system_ss.add(when: 'CONFIG_HYPERV', if_true: files('hyperv.c'), if_false: files('hyperv-stub.c'))
 
 i386_system_ss.add_all(when: 'CONFIG_KVM', if_true: i386_softmmu_kvm_ss)
target/i386/kvm/tdx.c new file +40 -0
new file mode 100644
index 000000000000..d3792d4a3d56
@@ -0,0 +1,40 @@
+/*
+ * QEMU TDX support
+ *
+ * Copyright Intel
+ *
+ * Author:
+ *      Xiaoyao Li <[email protected]>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qom/object_interfaces.h"
+
+#include "tdx.h"
+
+/* tdx guest */
+OBJECT_DEFINE_TYPE_WITH_INTERFACES(TdxGuest,
+                                   tdx_guest,
+                                   TDX_GUEST,
+                                   CONFIDENTIAL_GUEST_SUPPORT,
+                                   { TYPE_USER_CREATABLE },
+                                   { NULL })
+
+static void tdx_guest_init(Object *obj)
+{
+    TdxGuest *tdx = TDX_GUEST(obj);
+
+    tdx->attributes = 0;
+}
+
+static void tdx_guest_finalize(Object *obj)
+{
+}
+
+static void tdx_guest_class_init(ObjectClass *oc, void *data)
+{
+}
target/i386/kvm/tdx.h new file +19 -0
new file mode 100644
index 000000000000..415aeb5af746
@@ -0,0 +1,19 @@
+#ifndef QEMU_I386_TDX_H
+#define QEMU_I386_TDX_H
+
+#include "exec/confidential-guest-support.h"
+
+#define TYPE_TDX_GUEST "tdx-guest"
+#define TDX_GUEST(obj)  OBJECT_CHECK(TdxGuest, (obj), TYPE_TDX_GUEST)
+
+typedef struct TdxGuestClass {
+    ConfidentialGuestSupportClass parent_class;
+} TdxGuestClass;
+
+typedef struct TdxGuest {
+    ConfidentialGuestSupport parent_obj;
+
+    uint64_t attributes;    /* TD attributes */
+} TdxGuest;
+
+#endif /* QEMU_I386_TDX_H */
Signature · 1 line
2.34.1
Acked-by on v2 2/58
#3 lore raw

Re: [PATCH v2 02/58] i386: Introduce tdx-guest object

Quoted 5 lines · Xiaoyao Li

Introduce tdx-guest object which implements the interface of CONFIDENTIAL_GUEST_SUPPORT, and will be used to create TDX VMs (TDs) by

  qemu -machine ...,confidential-guest-support=tdx0	\
       -object tdx-guset,id=tdx0

Typo: tdx-guest

Quoted 2 lines

It has only one property 'attributes' with fixed value 0 and not configurable so far.

This must refer to TdxGuest member @attributes.

"Property" suggests QOM property, which @attributes isn't, at least not in this patch. Will it become a QOM property later in this series?

Hmm, @attributes appears to remain unused until PATCH 14. Recommend to delay its addition until then.

Quoted 81 lines

---
changes from RFC-V4
- make @attributes not user-settable

---
 configs/devices/i386-softmmu/default.mak |  1 +
 hw/i386/Kconfig                          |  5 +++
 qapi/qom.json                            | 12 +++++++
 target/i386/kvm/meson.build              |  2 ++
 target/i386/kvm/tdx.c                    | 40 ++++++++++++++++++++++++
 target/i386/kvm/tdx.h                    | 19 +++++++++++
 6 files changed, 79 insertions(+)
 create mode 100644 target/i386/kvm/tdx.c
 create mode 100644 target/i386/kvm/tdx.h
configs/devices/i386-softmmu/default.mak +1 -0
index 598c6646dfc0..9b5ec59d65b0 100644
@@ -18,6 +18,7 @@
 #CONFIG_QXL=n
 #CONFIG_SEV=n
 #CONFIG_SGA=n
+#CONFIG_TDX=n
 #CONFIG_TEST_DEVICES=n
 #CONFIG_TPM_CRB=n
 #CONFIG_TPM_TIS_ISA=n
hw/i386/Kconfig +5 -0
index 9051083c1e78..929f6c3f0e85 100644
@@ -10,6 +10,10 @@ config SGX
     bool
     depends on KVM
 
+config TDX
+    bool
+    depends on KVM
+
 config PC
     bool
     imply APPLESMC
@@ -26,6 +30,7 @@ config PC
     imply QXL
     imply SEV
     imply SGX
+    imply TDX
     imply TEST_DEVICES
     imply TPM_CRB
     imply TPM_TIS_ISA
qapi/qom.json +11 -0
index e0b2044e3d20..2ca7ce7c0da5 100644
@@ -866,6 +866,16 @@
             'reduced-phys-bits': 'uint32',
             '*kernel-hashes': 'bool' } }
 
+##
+# @TdxGuestProperties:
+#
+# Properties for tdx-guest objects.
+#
+# Since: 8.2
+##
+{ 'struct': 'TdxGuestProperties',
+  'data': { }}
+
 ##
 # @ThreadContextProperties:
 #
@@ -944,6 +954,7 @@
     'sev-guest',
     'thread-context',
     's390-pv-guest',
+    'tdx-guest',
     'throttle-group',
     'tls-creds-anon',
     'tls-creds-psk',

qapi/qom.json line 1027

Quoted 7 lines
qapi/qom.json +1 -0
@@ -1010,6 +1021,7 @@
       'secret_keyring':             { 'type': 'SecretKeyringProperties',
                                       'if': 'CONFIG_SECRET_KEYRING' },
       'sev-guest':                  'SevGuestProperties',
+      'tdx-guest':                  'TdxGuestProperties',
       'thread-context':             'ThreadContextProperties',
       'throttle-group':             'ThrottleGroupProperties',
       'tls-creds-anon':             'TlsCredsAnonProperties',

Actually useful only when CONFIG_TDX is on, but can't make it conditional here, as CONFIG_TDX is poisoned.

Quoted 74 lines
target/i386/kvm/meson.build +2 -0
index 40fbde96cac6..21ab03fe1349 100644
@@ -11,6 +11,8 @@ i386_softmmu_kvm_ss.add(when: 'CONFIG_XEN_EMU', if_true: files('xen-emu.c'))
 
 i386_softmmu_kvm_ss.add(when: 'CONFIG_SEV', if_false: files('sev-stub.c'))
 
+i386_softmmu_kvm_ss.add(when: 'CONFIG_TDX', if_true: files('tdx.c'))
+
 i386_system_ss.add(when: 'CONFIG_HYPERV', if_true: files('hyperv.c'), if_false: files('hyperv-stub.c'))
 
 i386_system_ss.add_all(when: 'CONFIG_KVM', if_true: i386_softmmu_kvm_ss)
target/i386/kvm/tdx.c new file +40 -0
new file mode 100644
index 000000000000..d3792d4a3d56
@@ -0,0 +1,40 @@
+/*
+ * QEMU TDX support
+ *
+ * Copyright Intel
+ *
+ * Author:
+ *      Xiaoyao Li <[email protected]>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qom/object_interfaces.h"
+
+#include "tdx.h"
+
+/* tdx guest */
+OBJECT_DEFINE_TYPE_WITH_INTERFACES(TdxGuest,
+                                   tdx_guest,
+                                   TDX_GUEST,
+                                   CONFIDENTIAL_GUEST_SUPPORT,
+                                   { TYPE_USER_CREATABLE },
+                                   { NULL })
+
+static void tdx_guest_init(Object *obj)
+{
+    TdxGuest *tdx = TDX_GUEST(obj);
+
+    tdx->attributes = 0;
+}
+
+static void tdx_guest_finalize(Object *obj)
+{
+}
+
+static void tdx_guest_class_init(ObjectClass *oc, void *data)
+{
+}
target/i386/kvm/tdx.h new file +9 -0
new file mode 100644
index 000000000000..415aeb5af746
@@ -0,0 +1,19 @@
+#ifndef QEMU_I386_TDX_H
+#define QEMU_I386_TDX_H
+
+#include "exec/confidential-guest-support.h"
+
+#define TYPE_TDX_GUEST "tdx-guest"
+#define TDX_GUEST(obj)  OBJECT_CHECK(TdxGuest, (obj), TYPE_TDX_GUEST)
+
+typedef struct TdxGuestClass {

target/i386/kvm/tdx.h line 19

Quoted 10 lines
target/i386/kvm/tdx.h new file +10 -0
+    ConfidentialGuestSupportClass parent_class;
+} TdxGuestClass;
+
+typedef struct TdxGuest {
+    ConfidentialGuestSupport parent_obj;
+
+    uint64_t attributes;    /* TD attributes */
+} TdxGuest;
+
+#endif /* QEMU_I386_TDX_H */

QAPI schema

author
#4 lore raw

Re: [PATCH v2 02/58] i386: Introduce tdx-guest object

Quoted 9 lines · Markus Armbruster
Quoted 5 lines · Xiaoyao Li

Introduce tdx-guest object which implements the interface of CONFIDENTIAL_GUEST_SUPPORT, and will be used to create TDX VMs (TDs) by

   qemu -machine ...,confidential-guest-support=tdx0	\
        -object tdx-guset,id=tdx0

Typo: tdx-guest

Will fix it.

Quoted 7 lines
Quoted 2 lines

It has only one property 'attributes' with fixed value 0 and not configurable so far.

This must refer to TdxGuest member @attributes.

"Property" suggests QOM property, which @attributes isn't, at least not in this patch. Will it become a QOM property later in this series?

At least not in this series. Maybe in the future there is request to directly configure the whole attributes via QOM property, but none from now.

I will change the description of it to avoid confusion.

Quoted 2 lines

Hmm, @attributes appears to remain unused until PATCH 14. Recommend to delay its addition until then.

IMHO, it's not suitable to introduce it in patch 14. Using a separate patch seems unnecessary. I'll leave it in this patch unless strong objection on it.

Quoted 83 lines
Quoted 83 lines

---
changes from RFC-V4
- make @attributes not user-settable

---
  configs/devices/i386-softmmu/default.mak |  1 +
  hw/i386/Kconfig                          |  5 +++
  qapi/qom.json                            | 12 +++++++
  target/i386/kvm/meson.build              |  2 ++
  target/i386/kvm/tdx.c                    | 40 ++++++++++++++++++++++++
  target/i386/kvm/tdx.h                    | 19 +++++++++++
  6 files changed, 79 insertions(+)

create mode 100644 target/i386/kvm/tdx.c
create mode 100644 target/i386/kvm/tdx.h

configs/devices/i386-softmmu/default.mak +1 -0
index 598c6646dfc0..9b5ec59d65b0 100644
@@ -18,6 +18,7 @@
  #CONFIG_QXL=n
  #CONFIG_SEV=n
  #CONFIG_SGA=n
+#CONFIG_TDX=n
  #CONFIG_TEST_DEVICES=n
  #CONFIG_TPM_CRB=n
  #CONFIG_TPM_TIS_ISA=n
hw/i386/Kconfig +5 -0
index 9051083c1e78..929f6c3f0e85 100644
@@ -10,6 +10,10 @@ config SGX
      bool
      depends on KVM
 
+config TDX
+    bool
+    depends on KVM
+
  config PC
      bool
      imply APPLESMC
@@ -26,6 +30,7 @@ config PC
      imply QXL
      imply SEV
      imply SGX
+    imply TDX
      imply TEST_DEVICES
      imply TPM_CRB
      imply TPM_TIS_ISA
qapi/qom.json +11 -0
index e0b2044e3d20..2ca7ce7c0da5 100644
@@ -866,6 +866,16 @@
              'reduced-phys-bits': 'uint32',
              '*kernel-hashes': 'bool' } }
 
+##
+# @TdxGuestProperties:
+#
+# Properties for tdx-guest objects.
+#
+# Since: 8.2
+##
+{ 'struct': 'TdxGuestProperties',
+  'data': { }}
+
  ##
  # @ThreadContextProperties:
  #
@@ -944,6 +954,7 @@
      'sev-guest',
      'thread-context',
      's390-pv-guest',
+    'tdx-guest',
      'throttle-group',
      'tls-creds-anon',
      'tls-creds-psk',
@@ -1010,6 +1021,7 @@
        'secret_keyring':             { 'type': 'SecretKeyringProperties',
                                        'if': 'CONFIG_SECRET_KEYRING' },
Quoted 8 lines

qapi/qom.json line 1027

Quoted 5 lines
qapi/qom.json +1 -0
        'sev-guest':                  'SevGuestProperties',
+      'tdx-guest':                  'TdxGuestProperties',
        'thread-context':             'ThreadContextProperties',
        'throttle-group':             'ThrottleGroupProperties',
        'tls-creds-anon':             'TlsCredsAnonProperties',

Actually useful only when CONFIG_TDX is on, but can't make it conditional here, as CONFIG_TDX is poisoned.

In fact, I just followed what SEV did.

To me, it looks OK to make it conditional on CONFIG_TDX. Could you please elaborate "but can't make it conditional here, as CONFIG_TDX is poisoned." ?

Quoted 86 lines
Quoted 74 lines
target/i386/kvm/meson.build +2 -0
index 40fbde96cac6..21ab03fe1349 100644
@@ -11,6 +11,8 @@ i386_softmmu_kvm_ss.add(when: 'CONFIG_XEN_EMU', if_true: files('xen-emu.c'))
 
  i386_softmmu_kvm_ss.add(when: 'CONFIG_SEV', if_false: files('sev-stub.c'))
 
+i386_softmmu_kvm_ss.add(when: 'CONFIG_TDX', if_true: files('tdx.c'))
+
  i386_system_ss.add(when: 'CONFIG_HYPERV', if_true: files('hyperv.c'), if_false: files('hyperv-stub.c'))
 
  i386_system_ss.add_all(when: 'CONFIG_KVM', if_true: i386_softmmu_kvm_ss)
target/i386/kvm/tdx.c new file +40 -0
new file mode 100644
index 000000000000..d3792d4a3d56
@@ -0,0 +1,40 @@
+/*
+ * QEMU TDX support
+ *
+ * Copyright Intel
+ *
+ * Author:
+ *      Xiaoyao Li <[email protected]>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory
+ *
+ */
+
+#include "qemu/osdep.h"
+#include "qom/object_interfaces.h"
+
+#include "tdx.h"
+
+/* tdx guest */
+OBJECT_DEFINE_TYPE_WITH_INTERFACES(TdxGuest,
+                                   tdx_guest,
+                                   TDX_GUEST,
+                                   CONFIDENTIAL_GUEST_SUPPORT,
+                                   { TYPE_USER_CREATABLE },
+                                   { NULL })
+
+static void tdx_guest_init(Object *obj)
+{
+    TdxGuest *tdx = TDX_GUEST(obj);
+
+    tdx->attributes = 0;
+}
+
+static void tdx_guest_finalize(Object *obj)
+{
+}
+
+static void tdx_guest_class_init(ObjectClass *oc, void *data)
+{
+}
target/i386/kvm/tdx.h new file +9 -0
new file mode 100644
index 000000000000..415aeb5af746
@@ -0,0 +1,19 @@
+#ifndef QEMU_I386_TDX_H
+#define QEMU_I386_TDX_H
+
+#include "exec/confidential-guest-support.h"
+
+#define TYPE_TDX_GUEST "tdx-guest"
+#define TDX_GUEST(obj)  OBJECT_CHECK(TdxGuest, (obj), TYPE_TDX_GUEST)
+
+typedef struct TdxGuestClass {

target/i386/kvm/tdx.h line 19

Quoted 10 lines
target/i386/kvm/tdx.h new file +10 -0
+    ConfidentialGuestSupportClass parent_class;
+} TdxGuestClass;
+
+typedef struct TdxGuest {
+    ConfidentialGuestSupport parent_obj;
+
+    uint64_t attributes;    /* TD attributes */
+} TdxGuest;
+
+#endif /* QEMU_I386_TDX_H */

QAPI schema

Quoted 1 line

Thank you!

#5 lore raw

Re: [PATCH v2 02/58] i386: Introduce tdx-guest object

Quoted 20 lines · Xiaoyao Li
Quoted 9 lines · Markus Armbruster
Quoted 5 lines · Xiaoyao Li

Introduce tdx-guest object which implements the interface of CONFIDENTIAL_GUEST_SUPPORT, and will be used to create TDX VMs (TDs) by

   qemu -machine ...,confidential-guest-support=tdx0	\
        -object tdx-guset,id=tdx0

Typo: tdx-guest

Will fix it.

Quoted 6 lines
Quoted 2 lines

It has only one property 'attributes' with fixed value 0 and not configurable so far.

This must refer to TdxGuest member @attributes.
"Property" suggests QOM property, which @attributes isn't, at least not in this patch. Will it become a QOM property later in this series?

Quoted 8 lines

At least not in this series. Maybe in the future there is request to directly configure the whole attributes via QOM property, but none from now.

I will change the description of it to avoid confusion.

Quoted 2 lines

Hmm, @attributes appears to remain unused until PATCH 14. Recommend to delay its addition until then.

IMHO, it's not suitable to introduce it in patch 14. Using a separate patch seems unnecessary. I'll leave it in this patch unless strong objection on it.

Not worth arguing about.

Quoted 90 lines
Quoted 81 lines
Quoted 81 lines

---
changes from RFC-V4
- make @attributes not user-settable

---
  configs/devices/i386-softmmu/default.mak |  1 +
  hw/i386/Kconfig                          |  5 +++
  qapi/qom.json                            | 12 +++++++
  target/i386/kvm/meson.build              |  2 ++
  target/i386/kvm/tdx.c                    | 40 ++++++++++++++++++++++++
  target/i386/kvm/tdx.h                    | 19 +++++++++++
  6 files changed, 79 insertions(+)

create mode 100644 target/i386/kvm/tdx.c
create mode 100644 target/i386/kvm/tdx.h

configs/devices/i386-softmmu/default.mak +1 -0
index 598c6646dfc0..9b5ec59d65b0 100644
@@ -18,6 +18,7 @@
  #CONFIG_QXL=n
  #CONFIG_SEV=n
  #CONFIG_SGA=n
+#CONFIG_TDX=n
  #CONFIG_TEST_DEVICES=n
  #CONFIG_TPM_CRB=n
  #CONFIG_TPM_TIS_ISA=n
hw/i386/Kconfig +4 -0
index 9051083c1e78..929f6c3f0e85 100644
@@ -10,6 +10,10 @@ config SGX
      bool
      depends on KVM
  +config TDX
+    bool
+    depends on KVM
+
  config PC
      bool
      imply APPLESMC
@@ -26,6 +30,7 @@ config PC
      imply QXL
      imply SEV
      imply SGX
+    imply TDX
      imply TEST_DEVICES
      imply TPM_CRB
      imply TPM_TIS_ISA
qapi/qom.json +10 -0
index e0b2044e3d20..2ca7ce7c0da5 100644
@@ -866,6 +866,16 @@
              'reduced-phys-bits': 'uint32',
              '*kernel-hashes': 'bool' } }
  +##
+# @TdxGuestProperties:
+#
+# Properties for tdx-guest objects.
+#
+# Since: 8.2
+##
+{ 'struct': 'TdxGuestProperties',
+  'data': { }}
+
  ##
  # @ThreadContextProperties:
  #
@@ -944,6 +954,7 @@
      'sev-guest',
      'thread-context',
      's390-pv-guest',
+    'tdx-guest',
      'throttle-group',
      'tls-creds-anon',
      'tls-creds-psk',
@@ -1010,6 +1021,7 @@
        'secret_keyring':             { 'type': 'SecretKeyringProperties',
                                        'if': 'CONFIG_SECRET_KEYRING' },
Quoted 8 lines

qapi/qom.json line 1027

Quoted 5 lines
qapi/qom.json +1 -0
        'sev-guest':                  'SevGuestProperties',
+      'tdx-guest':                  'TdxGuestProperties',
        'thread-context':             'ThreadContextProperties',
        'throttle-group':             'ThrottleGroupProperties',
        'tls-creds-anon':             'TlsCredsAnonProperties',

Actually useful only when CONFIG_TDX is on, but can't make it conditional here, as CONFIG_TDX is poisoned.

Quoted 1 line

In fact, I just followed what SEV did.

Yup.

Quoted 1 line

To me, it looks OK to make it conditional on CONFIG_TDX. Could you please elaborate "but can't make it conditional here, as CONFIG_TDX is poisoned." ?

CONFIG_TDX is one of the macros that can only be used in target-dependent code. Enforced by config-poison.h's

#pragma GCC poison CONFIG_TDX

The code generated from qom.json is target-independent.

To use 'if': 'CONFIG_TDX', we'd have to move the definition to a target-dependent QAPI module, say qom-machine.json. Sadly, that's more trouble than it's worth.

[...]

author
#6 lore raw

PATCH v2 3/58 target/i386: Parse TDX vm type

In reply to Xiaoyao Li

TDX VM requires VM type KVM_X86_TDX_VM to be passed to kvm_ioctl(KVM_CREATE_VM).

If tdx-guest object is specified to confidential-guest-support, like,

  qemu -machine ...,confidential-guest-support=tdx0 \
       -object tdx-guest,id=tdx0,...

it parses VM type as KVM_X86_TDX_VM.

---
 target/i386/kvm/kvm.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
target/i386/kvm/kvm.c +9 -1
index 62f237068a3a..77f4772afe6c 100644
@@ -32,6 +32,7 @@
 #include "sysemu/runstate.h"
 #include "kvm_i386.h"
 #include "sev.h"
+#include "tdx.h"
 #include "xen-emu.h"
 #include "hyperv.h"
 #include "hyperv-proto.h"
@@ -158,6 +159,7 @@ static int kvm_get_one_msr(X86CPU *cpu, int index, uint64_t *value);
 static const char* vm_type_name[] = {
     [KVM_X86_DEFAULT_VM] = "default",
     [KVM_X86_SW_PROTECTED_VM] = "sw-protected-vm",
+    [KVM_X86_TDX_VM] = "tdx",
 };
 
 int kvm_get_vm_type(MachineState *ms, const char *vm_type)
@@ -170,12 +172,18 @@ int kvm_get_vm_type(MachineState *ms, const char *vm_type)
             kvm_type = KVM_X86_DEFAULT_VM;
         } else if (!g_ascii_strcasecmp(vm_type, "sw-protected-vm")) {
             kvm_type = KVM_X86_SW_PROTECTED_VM;
-        } else {
+        } else if (!g_ascii_strcasecmp(vm_type, "tdx")) {
+            kvm_type = KVM_X86_TDX_VM;
+        }else {
             error_report("Unknown kvm-type specified '%s'", vm_type);
             exit(1);
         }
     }
 
+    if (ms->cgs && object_dynamic_cast(OBJECT(ms->cgs), TYPE_TDX_GUEST)) {
+        kvm_type = KVM_X86_TDX_VM;
+    }
+
     /*
      * old KVM doesn't support KVM_CAP_VM_TYPES and KVM_X86_DEFAULT_VM
      * is always supported
Signature · 1 line
2.34.1
#7 lore raw

Re: [PATCH v2 03/58] target/i386: Parse TDX vm type

Quoted 37 lines · Xiaoyao Li

TDX VM requires VM type KVM_X86_TDX_VM to be passed to kvm_ioctl(KVM_CREATE_VM).

If tdx-guest object is specified to confidential-guest-support, like,

  qemu -machine ...,confidential-guest-support=tdx0 \
       -object tdx-guest,id=tdx0,...

it parses VM type as KVM_X86_TDX_VM.

---
 target/i386/kvm/kvm.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
target/i386/kvm/kvm.c +2 -0
index 62f237068a3a..77f4772afe6c 100644
@@ -32,6 +32,7 @@
 #include "sysemu/runstate.h"
 #include "kvm_i386.h"
 #include "sev.h"
+#include "tdx.h"
 #include "xen-emu.h"
 #include "hyperv.h"
 #include "hyperv-proto.h"
@@ -158,6 +159,7 @@ static int kvm_get_one_msr(X86CPU *cpu, int index, uint64_t *value);
 static const char* vm_type_name[] = {
     [KVM_X86_DEFAULT_VM] = "default",
     [KVM_X86_SW_PROTECTED_VM] = "sw-protected-vm",
+    [KVM_X86_TDX_VM] = "tdx",
 };
 
 int kvm_get_vm_type(MachineState *ms, const char *vm_type)
@@ -170,12 +172,18 @@ int kvm_get_vm_type(MachineState *ms, const char *vm_type)
             kvm_type = KVM_X86_DEFAULT_VM;

target/i386/kvm/kvm.c int kvm_get_vm_type(MachineState *ms, const char *vm_type) · line 181

Quoted 10 lines
target/i386/kvm/kvm.c +3 -1
         } else if (!g_ascii_strcasecmp(vm_type, "sw-protected-vm")) {
             kvm_type = KVM_X86_SW_PROTECTED_VM;
-        } else {
+        } else if (!g_ascii_strcasecmp(vm_type, "tdx")) {
+            kvm_type = KVM_X86_TDX_VM;
+        }else {
             error_report("Unknown kvm-type specified '%s'", vm_type);
             exit(1);
         }
     }

This whole block of code should go away - as this should not exist as a user visible property. It should be sufficient to use the tdx-guest object type to identify use of TDX.

Quoted 11 lines
target/i386/kvm/kvm.c +4 -0
+    if (ms->cgs && object_dynamic_cast(OBJECT(ms->cgs), TYPE_TDX_GUEST)) {
+        kvm_type = KVM_X86_TDX_VM;
+    }
+
     /*
      * old KVM doesn't support KVM_CAP_VM_TYPES and KVM_X86_DEFAULT_VM
      * is always supported
Signature · 1 line
2.34.1

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
author
#8 lore raw

Re: [PATCH v2 03/58] target/i386: Parse TDX vm type

Quoted 44 lines · Daniel P. Berrangé
Quoted 43 lines · Xiaoyao Li

TDX VM requires VM type KVM_X86_TDX_VM to be passed to kvm_ioctl(KVM_CREATE_VM).

If tdx-guest object is specified to confidential-guest-support, like,

   qemu -machine ...,confidential-guest-support=tdx0 \
        -object tdx-guest,id=tdx0,...

it parses VM type as KVM_X86_TDX_VM.

---
  target/i386/kvm/kvm.c | 10 +++++++++-
  1 file changed, 9 insertions(+), 1 deletion(-)
target/i386/kvm/kvm.c +5 -1
index 62f237068a3a..77f4772afe6c 100644
@@ -32,6 +32,7 @@
  #include "sysemu/runstate.h"
  #include "kvm_i386.h"
  #include "sev.h"
+#include "tdx.h"
  #include "xen-emu.h"
  #include "hyperv.h"
  #include "hyperv-proto.h"
@@ -158,6 +159,7 @@ static int kvm_get_one_msr(X86CPU *cpu, int index, uint64_t *value);
  static const char* vm_type_name[] = {
      [KVM_X86_DEFAULT_VM] = "default",
      [KVM_X86_SW_PROTECTED_VM] = "sw-protected-vm",
+    [KVM_X86_TDX_VM] = "tdx",
  };
 
  int kvm_get_vm_type(MachineState *ms, const char *vm_type)
@@ -170,12 +172,18 @@ int kvm_get_vm_type(MachineState *ms, const char *vm_type)
              kvm_type = KVM_X86_DEFAULT_VM;
          } else if (!g_ascii_strcasecmp(vm_type, "sw-protected-vm")) {
              kvm_type = KVM_X86_SW_PROTECTED_VM;
-        } else {
+        } else if (!g_ascii_strcasecmp(vm_type, "tdx")) {
+            kvm_type = KVM_X86_TDX_VM;
+        }else {
Quoted 9 lines

target/i386/kvm/kvm.c int kvm_get_vm_type(MachineState *ms, const char *vm_type) · line 181

Quoted 4 lines
target/i386/kvm/kvm.c +0 -0
              error_report("Unknown kvm-type specified '%s'", vm_type);
              exit(1);
          }
      }

This whole block of code should go away - as this should not exist as a user visible property. It should be sufficient to use the tdx-guest object type to identify use of TDX.

yes, agreed.

It's here because this series is based on the gmem series, which introduced property. I'm sorry that I forgot to mention it in the commit message.

Next gmem series will drop the implementation of kvm-type property [1] and above code will be dropped in next version as well.

[1] https://lore.kernel.org/qemu-devel/[email protected]/

author
#9 lore raw

PATCH v2 4/58 target/i386: Introduce kvm_confidential_guest_init()

In reply to Xiaoyao Li

Introduce a separate function kvm_confidential_guest_init() for SEV (and future TDX).

---
 target/i386/kvm/kvm.c | 11 ++++++++++-
 target/i386/sev.c     |  1 -
 target/i386/sev.h     |  2 ++
 3 files changed, 12 insertions(+), 2 deletions(-)
target/i386/kvm/kvm.c +10 -1
index 77f4772afe6c..051307437ecd 100644
@@ -2633,6 +2633,15 @@ static MemoryListener kvm_x86_sw_protected_vm_memory_listener = {
     .priority = MEMORY_LISTENER_PRIORITY_ACCEL_HIGH,
 };
 
+static int kvm_confidential_guest_init(MachineState *ms, Error **errp)
+{
+    if (object_dynamic_cast(OBJECT(ms->cgs), TYPE_SEV_GUEST)) {
+        return sev_kvm_init(ms->cgs, errp);
+    }
+
+    return 0;
+}
+
 int kvm_arch_init(MachineState *ms, KVMState *s)
 {
     X86MachineState *x86ms = X86_MACHINE(ms);
@@ -2654,7 +2663,7 @@ int kvm_arch_init(MachineState *ms, KVMState *s)
      * mechanisms are supported in future (e.g. TDX), they'll need
      * their own initialization either here or elsewhere.
      */
-    ret = sev_kvm_init(ms->cgs, &local_err);
+    ret = kvm_confidential_guest_init(ms, &local_err);
     if (ret < 0) {
         error_report_err(local_err);
         return ret;
target/i386/sev.c +0 -1
index fe2144c0388b..5aa04863846d 100644
@@ -39,7 +39,6 @@
 #include "hw/i386/pc.h"
 #include "exec/address-spaces.h"
 
-#define TYPE_SEV_GUEST "sev-guest"
 OBJECT_DECLARE_SIMPLE_TYPE(SevGuestState, SEV_GUEST)
 
 
target/i386/sev.h +2 -0
index 7b1528248a54..64fbf186dbd2 100644
@@ -20,6 +20,8 @@
 
 #include "exec/confidential-guest-support.h"
 
+#define TYPE_SEV_GUEST "sev-guest"
+
 #define SEV_POLICY_NODBG        0x1
 #define SEV_POLICY_NOKS         0x2
 #define SEV_POLICY_ES           0x4
Signature · 1 line
2.34.1
Reviewed-by on v2 4/58
#10 lore raw

Re: [PATCH v2 04/58] target/i386: Introduce kvm_confidential_guest_init()

Quoted 10 lines · Xiaoyao Li

Introduce a separate function kvm_confidential_guest_init() for SEV (and future TDX).

---
  target/i386/kvm/kvm.c | 11 ++++++++++-
  target/i386/sev.c     |  1 -
  target/i386/sev.h     |  2 ++
  3 files changed, 12 insertions(+), 2 deletions(-)
author
#11 lore raw

PATCH v2 5/58 i386/tdx: Implement tdx_kvm_init() to initialize TDX VM context

In reply to Xiaoyao Li

Introduce tdx_kvm_init() and invoke it in kvm_confidential_guest_init() if it's a TDX VM. More initialization will be added later.

---
 target/i386/kvm/kvm.c       | 15 ++++++---------
 target/i386/kvm/meson.build |  2 +-
 target/i386/kvm/tdx-stub.c  |  8 ++++++++
 target/i386/kvm/tdx.c       |  7 +++++++
 target/i386/kvm/tdx.h       |  2 ++
 5 files changed, 24 insertions(+), 10 deletions(-)
 create mode 100644 target/i386/kvm/tdx-stub.c
target/i386/kvm/kvm.c +6 -9
index 051307437ecd..d6b988d6c2d1 100644
@@ -63,6 +63,7 @@
 #include "migration/blocker.h"
 #include "exec/memattrs.h"
 #include "trace.h"
+#include "tdx.h"
 
 #include CONFIG_DEVICES
 
@@ -2637,6 +2638,8 @@ static int kvm_confidential_guest_init(MachineState *ms, Error **errp)
 {
     if (object_dynamic_cast(OBJECT(ms->cgs), TYPE_SEV_GUEST)) {
         return sev_kvm_init(ms->cgs, errp);
+    } else if (object_dynamic_cast(OBJECT(ms->cgs), TYPE_TDX_GUEST)) {
+        return tdx_kvm_init(ms, errp);
     }
 
     return 0;
@@ -2652,16 +2655,10 @@ int kvm_arch_init(MachineState *ms, KVMState *s)
     Error *local_err = NULL;
 
     /*
-     * Initialize SEV context, if required
+     * Initialize confidential guest (SEV/TDX) context, if required
      *
-     * If no memory encryption is requested (ms->cgs == NULL) this is
-     * a no-op.
-     *
-     * It's also a no-op if a non-SEV confidential guest support
-     * mechanism is selected.  SEV is the only mechanism available to
-     * select on x86 at present, so this doesn't arise, but if new
-     * mechanisms are supported in future (e.g. TDX), they'll need
-     * their own initialization either here or elsewhere.
+     * It's a no-op if a non-SEV/non-tdx confidential guest support
+     * mechanism is selected, i.e., ms->cgs == NULL
      */
     ret = kvm_confidential_guest_init(ms, &local_err);
     if (ret < 0) {
target/i386/kvm/meson.build +1 -1
index 21ab03fe1349..876350a387aa 100644
@@ -11,7 +11,7 @@ i386_softmmu_kvm_ss.add(when: 'CONFIG_XEN_EMU', if_true: files('xen-emu.c'))
 
 i386_softmmu_kvm_ss.add(when: 'CONFIG_SEV', if_false: files('sev-stub.c'))
 
-i386_softmmu_kvm_ss.add(when: 'CONFIG_TDX', if_true: files('tdx.c'))
+i386_softmmu_kvm_ss.add(when: 'CONFIG_TDX', if_true: files('tdx.c'), if_false: files('tdx-stub.c'))
 
 i386_system_ss.add(when: 'CONFIG_HYPERV', if_true: files('hyperv.c'), if_false: files('hyperv-stub.c'))
 
target/i386/kvm/tdx-stub.c new file +8 -0
new file mode 100644
index 000000000000..1d866d5496bf
@@ -0,0 +1,8 @@
+#include "qemu/osdep.h"
+
+#include "tdx.h"
+
+int tdx_kvm_init(MachineState *ms, Error **errp)
+{
+    return -EINVAL;
+}
target/i386/kvm/tdx.c +7 -0
index d3792d4a3d56..77e33ae01147 100644
@@ -12,10 +12,17 @@
  */
 
 #include "qemu/osdep.h"
+#include "qapi/error.h"
 #include "qom/object_interfaces.h"
 
+#include "hw/i386/x86.h"
 #include "tdx.h"
 
+int tdx_kvm_init(MachineState *ms, Error **errp)
+{
+    return 0;
+}
+
 /* tdx guest */
 OBJECT_DEFINE_TYPE_WITH_INTERFACES(TdxGuest,
                                    tdx_guest,
target/i386/kvm/tdx.h +2 -0
index 415aeb5af746..c8a23d95258d 100644
@@ -16,4 +16,6 @@ typedef struct TdxGuest {
     uint64_t attributes;    /* TD attributes */
 } TdxGuest;
 
+int tdx_kvm_init(MachineState *ms, Error **errp);
+
 #endif /* QEMU_I386_TDX_H */
Signature · 1 line
2.34.1
author
#12 lore raw

PATCH v2 6/58 i386/tdx: Get tdx_capabilities via KVM_TDX_CAPABILITIES

In reply to Xiaoyao Li

KVM provides TDX capabilities via sub command KVM_TDX_CAPABILITIES of IOCTL(KVM_MEMORY_ENCRYPT_OP). Get the capabilities when initializing TDX context. It will be used to validate user's setting later.

Since there is no interface reporting how many cpuid configs contains in KVM_TDX_CAPABILITIES, QEMU chooses to try starting with a known number and abort when it exceeds KVM_MAX_CPUID_ENTRIES.

Besides, introduce the interfaces to invoke TDX "ioctls" at different scope (KVM, VM and VCPU) in preparation.

---
changes from v1:
- Make the error message more clear;

changes from RFC v4:
- start from nr_cpuid_configs = 6 for the loop;
- stop the loop when nr_cpuid_configs exceeds KVM_MAX_CPUID_ENTRIES;

---
 target/i386/kvm/kvm.c      |  2 -
 target/i386/kvm/kvm_i386.h |  2 +
 target/i386/kvm/tdx.c      | 93 ++++++++++++++++++++++++++++++++++++++
 3 files changed, 95 insertions(+), 2 deletions(-)
target/i386/kvm/kvm.c +0 -2
index d6b988d6c2d1..ec5c07bffd38 100644
@@ -1751,8 +1751,6 @@ static int hyperv_init_vcpu(X86CPU *cpu)
 
 static Error *invtsc_mig_blocker;
 
-#define KVM_MAX_CPUID_ENTRIES  100
-
 static void kvm_init_xsave(CPUX86State *env)
 {
     if (has_xsave2) {
target/i386/kvm/kvm_i386.h +2 -0
index ea3a5b174ac0..769eadbba56c 100644
@@ -13,6 +13,8 @@
 
 #include "sysemu/kvm.h"
 
+#define KVM_MAX_CPUID_ENTRIES  100
+
 #define kvm_apic_in_kernel() (kvm_irqchip_in_kernel())
 
 #ifdef CONFIG_KVM
target/i386/kvm/tdx.c +93 -0

107 lines not on this page · raw message

Signature · 1 line
2.34.1
#13 lore raw

Re: [PATCH v2 06/58] i386/tdx: Get tdx_capabilities via KVM_TDX_CAPABILITIES

Quoted 68 lines · Xiaoyao Li

KVM provides TDX capabilities via sub command KVM_TDX_CAPABILITIES of IOCTL(KVM_MEMORY_ENCRYPT_OP). Get the capabilities when initializing TDX context. It will be used to validate user's setting later.

Since there is no interface reporting how many cpuid configs contains in KVM_TDX_CAPABILITIES, QEMU chooses to try starting with a known number and abort when it exceeds KVM_MAX_CPUID_ENTRIES.

Besides, introduce the interfaces to invoke TDX "ioctls" at different scope (KVM, VM and VCPU) in preparation.

---
changes from v1:
- Make the error message more clear;

changes from RFC v4:
- start from nr_cpuid_configs = 6 for the loop;
- stop the loop when nr_cpuid_configs exceeds KVM_MAX_CPUID_ENTRIES;

---
 target/i386/kvm/kvm.c      |  2 -
 target/i386/kvm/kvm_i386.h |  2 +
 target/i386/kvm/tdx.c      | 93 ++++++++++++++++++++++++++++++++++++++
 3 files changed, 95 insertions(+), 2 deletions(-)
target/i386/kvm/kvm.c +0 -2
index d6b988d6c2d1..ec5c07bffd38 100644
@@ -1751,8 +1751,6 @@ static int hyperv_init_vcpu(X86CPU *cpu)
 
 static Error *invtsc_mig_blocker;
 
-#define KVM_MAX_CPUID_ENTRIES  100
-
 static void kvm_init_xsave(CPUX86State *env)
 {
     if (has_xsave2) {
target/i386/kvm/kvm_i386.h +2 -0
index ea3a5b174ac0..769eadbba56c 100644
@@ -13,6 +13,8 @@
 
 #include "sysemu/kvm.h"
 
+#define KVM_MAX_CPUID_ENTRIES  100
+
 #define kvm_apic_in_kernel() (kvm_irqchip_in_kernel())
 
 #ifdef CONFIG_KVM
target/i386/kvm/tdx.c +3 -0
index 77e33ae01147..255c47a2a553 100644
@@ -12,14 +12,107 @@
  */
 
 #include "qemu/osdep.h"
+#include "qemu/error-report.h"
 #include "qapi/error.h"
 #include "qom/object_interfaces.h"
+#include "sysemu/kvm.h"
 
 #include "hw/i386/x86.h"
+#include "kvm_i386.h"
 #include "tdx.h"
 

target/i386/kvm/tdx.c line 33

Quoted 10 lines
target/i386/kvm/tdx.c +10 -0
+static struct kvm_tdx_capabilities *tdx_caps;
+
+enum tdx_ioctl_level{
+    TDX_PLATFORM_IOCTL,
+    TDX_VM_IOCTL,
+    TDX_VCPU_IOCTL,
+};
+
+static int __tdx_ioctl(void *state, enum tdx_ioctl_level level, int cmd_id,
+                        __u32 flags, void *data)

Names with an initial double underscore are reserved for us by the platform implementation, so shouldn't be used in userspace app code.

Quoted 35 lines
target/i386/kvm/tdx.c +35 -0
+{
+    struct kvm_tdx_cmd tdx_cmd;
+    int r;
+
+    memset(&tdx_cmd, 0x0, sizeof(tdx_cmd));
+
+    tdx_cmd.id = cmd_id;
+    tdx_cmd.flags = flags;
+    tdx_cmd.data = (__u64)(unsigned long)data;
+
+    switch (level) {
+    case TDX_PLATFORM_IOCTL:
+        r = kvm_ioctl(kvm_state, KVM_MEMORY_ENCRYPT_OP, &tdx_cmd);
+        break;
+    case TDX_VM_IOCTL:
+        r = kvm_vm_ioctl(kvm_state, KVM_MEMORY_ENCRYPT_OP, &tdx_cmd);
+        break;
+    case TDX_VCPU_IOCTL:
+        r = kvm_vcpu_ioctl(state, KVM_MEMORY_ENCRYPT_OP, &tdx_cmd);
+        break;
+    default:
+        error_report("Invalid tdx_ioctl_level %d", level);
+        exit(1);
+    }
+
+    return r;
+}
+
+static inline int tdx_platform_ioctl(int cmd_id, __u32 flags, void *data)
+{
+    return __tdx_ioctl(NULL, TDX_PLATFORM_IOCTL, cmd_id, flags, data);
+}
+
+static inline int tdx_vm_ioctl(int cmd_id, __u32 flags, void *data)
+{

target/i386/kvm/tdx.c line 78

Quoted 10 lines
target/i386/kvm/tdx.c +10 -0
+    return __tdx_ioctl(NULL, TDX_VM_IOCTL, cmd_id, flags, data);
+}
+
+static inline int tdx_vcpu_ioctl(void *vcpu_fd, int cmd_id, __u32 flags,
+                                 void *data)
+{
+    return  __tdx_ioctl(vcpu_fd, TDX_VCPU_IOCTL, cmd_id, flags, data);
+}
+
+static void get_tdx_capabilities(void)

Pass in 'Error **errp'

target/i386/kvm/tdx.c line 83

Quoted 5 lines
target/i386/kvm/tdx.c +5 -0
+{
+    struct kvm_tdx_capabilities *caps;
+    /* 1st generation of TDX reports 6 cpuid configs */
+    int nr_cpuid_configs = 6;
+    int r, size;

It is preferrable to use 'size_t' for memory allocation sizes.

target/i386/kvm/tdx.c line 96

Quoted 13 lines
target/i386/kvm/tdx.c +13 -0
+
+    do {
+        size = sizeof(struct kvm_tdx_capabilities) +
+               nr_cpuid_configs * sizeof(struct kvm_tdx_cpuid_config);
+        caps = g_malloc0(size);
+        caps->nr_cpuid_configs = nr_cpuid_configs;
+
+        r = tdx_vm_ioctl(KVM_TDX_CAPABILITIES, 0, caps);
+        if (r == -E2BIG) {
+            g_free(caps);
+            nr_cpuid_configs *= 2;
+            if (nr_cpuid_configs > KVM_MAX_CPUID_ENTRIES) {
+                error_report("KVM TDX seems broken that number of CPUID entries in kvm_tdx_capabilities exceeds limit");

Include the limit in the error message, so if we ever need to change the limit, it'll be clear what limit the QEMU version was built with.

Also use error_setg(errp, ...);

target/i386/kvm/tdx.c line 97

Quoted 1 line
target/i386/kvm/tdx.c +1 -0
+                exit(1);

Return -1

target/i386/kvm/tdx.c line 101

Quoted 4 lines
target/i386/kvm/tdx.c +4 -0
+            }
+        } else if (r < 0) {
+            g_free(caps);
+            error_report("KVM_TDX_CAPABILITIES failed: %s", strerror(-r));

Use error_setg_errno(errp, ...) instead of calling strerror yourself;

target/i386/kvm/tdx.c line 102

Quoted 1 line
target/i386/kvm/tdx.c +1 -0
+            exit(1);

Return -1

target/i386/kvm/tdx.c line 107

Quoted 5 lines
target/i386/kvm/tdx.c +5 -0
+        }
+    }
+    while (r == -E2BIG);
+
+    tdx_caps = caps;

Return 0

target/i386/kvm/tdx.c line 113

Quoted 6 lines
target/i386/kvm/tdx.c +4 -0
+}
+
 int tdx_kvm_init(MachineState *ms, Error **errp)
 {
+    if (!tdx_caps) {
+        get_tdx_capabilities();

Pass 'errp' into this method, and check return value for failure

Quoted 8 lines
target/i386/kvm/tdx.c +2 -0
+    }
+
     return 0;
 }
 
Signature · 1 line
2.34.1

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
author
#14 lore raw

Re: [PATCH v2 06/58] i386/tdx: Get tdx_capabilities via KVM_TDX_CAPABILITIES

Quoted 69 lines · Daniel P. Berrangé
Quoted 68 lines · Xiaoyao Li

KVM provides TDX capabilities via sub command KVM_TDX_CAPABILITIES of IOCTL(KVM_MEMORY_ENCRYPT_OP). Get the capabilities when initializing TDX context. It will be used to validate user's setting later.

Since there is no interface reporting how many cpuid configs contains in KVM_TDX_CAPABILITIES, QEMU chooses to try starting with a known number and abort when it exceeds KVM_MAX_CPUID_ENTRIES.

Besides, introduce the interfaces to invoke TDX "ioctls" at different scope (KVM, VM and VCPU) in preparation.

---
changes from v1:
- Make the error message more clear;

changes from RFC v4:
- start from nr_cpuid_configs = 6 for the loop;
- stop the loop when nr_cpuid_configs exceeds KVM_MAX_CPUID_ENTRIES;

---
  target/i386/kvm/kvm.c      |  2 -
  target/i386/kvm/kvm_i386.h |  2 +
  target/i386/kvm/tdx.c      | 93 ++++++++++++++++++++++++++++++++++++++
  3 files changed, 95 insertions(+), 2 deletions(-)
target/i386/kvm/kvm.c +0 -2
index d6b988d6c2d1..ec5c07bffd38 100644
@@ -1751,8 +1751,6 @@ static int hyperv_init_vcpu(X86CPU *cpu)
 
  static Error *invtsc_mig_blocker;
 
-#define KVM_MAX_CPUID_ENTRIES  100
-
  static void kvm_init_xsave(CPUX86State *env)
  {
      if (has_xsave2) {
target/i386/kvm/kvm_i386.h +2 -0
index ea3a5b174ac0..769eadbba56c 100644
@@ -13,6 +13,8 @@
 
  #include "sysemu/kvm.h"
 
+#define KVM_MAX_CPUID_ENTRIES  100
+
  #define kvm_apic_in_kernel() (kvm_irqchip_in_kernel())
 
  #ifdef CONFIG_KVM
target/i386/kvm/tdx.c +3 -0
index 77e33ae01147..255c47a2a553 100644
@@ -12,14 +12,107 @@
   */
 
  #include "qemu/osdep.h"
+#include "qemu/error-report.h"
  #include "qapi/error.h"
  #include "qom/object_interfaces.h"
+#include "sysemu/kvm.h"
 
  #include "hw/i386/x86.h"
+#include "kvm_i386.h"
  #include "tdx.h"
 
Quoted 14 lines

target/i386/kvm/tdx.c line 33

Quoted 10 lines
target/i386/kvm/tdx.c +10 -0
+static struct kvm_tdx_capabilities *tdx_caps;
+
+enum tdx_ioctl_level{
+    TDX_PLATFORM_IOCTL,
+    TDX_VM_IOCTL,
+    TDX_VCPU_IOCTL,
+};
+
+static int __tdx_ioctl(void *state, enum tdx_ioctl_level level, int cmd_id,
+                        __u32 flags, void *data)

Names with an initial double underscore are reserved for us by the platform implementation, so shouldn't be used in userspace app code.

How about tdx_ioctl_internal() ?

Quoted 39 lines
Quoted 39 lines
target/i386/kvm/tdx.c +39 -0
+{
+    struct kvm_tdx_cmd tdx_cmd;
+    int r;
+
+    memset(&tdx_cmd, 0x0, sizeof(tdx_cmd));
+
+    tdx_cmd.id = cmd_id;
+    tdx_cmd.flags = flags;
+    tdx_cmd.data = (__u64)(unsigned long)data;
+
+    switch (level) {
+    case TDX_PLATFORM_IOCTL:
+        r = kvm_ioctl(kvm_state, KVM_MEMORY_ENCRYPT_OP, &tdx_cmd);
+        break;
+    case TDX_VM_IOCTL:
+        r = kvm_vm_ioctl(kvm_state, KVM_MEMORY_ENCRYPT_OP, &tdx_cmd);
+        break;
+    case TDX_VCPU_IOCTL:
+        r = kvm_vcpu_ioctl(state, KVM_MEMORY_ENCRYPT_OP, &tdx_cmd);
+        break;
+    default:
+        error_report("Invalid tdx_ioctl_level %d", level);
+        exit(1);
+    }
+
+    return r;
+}
+
+static inline int tdx_platform_ioctl(int cmd_id, __u32 flags, void *data)
+{
+    return __tdx_ioctl(NULL, TDX_PLATFORM_IOCTL, cmd_id, flags, data);
+}
+
+static inline int tdx_vm_ioctl(int cmd_id, __u32 flags, void *data)
+{
+    return __tdx_ioctl(NULL, TDX_VM_IOCTL, cmd_id, flags, data);
+}
+
+static inline int tdx_vcpu_ioctl(void *vcpu_fd, int cmd_id, __u32 flags,
Quoted 8 lines

target/i386/kvm/tdx.c line 78

Quoted 6 lines
target/i386/kvm/tdx.c +6 -0
+                                 void *data)
+{
+    return  __tdx_ioctl(vcpu_fd, TDX_VCPU_IOCTL, cmd_id, flags, data);
+}
+
+static void get_tdx_capabilities(void)

Pass in 'Error **errp'

OK. Will do it and all the following.

Thanks!

Quoted 70 lines

target/i386/kvm/tdx.c line 83

Quoted 5 lines
target/i386/kvm/tdx.c +5 -0
+{
+    struct kvm_tdx_capabilities *caps;
+    /* 1st generation of TDX reports 6 cpuid configs */
+    int nr_cpuid_configs = 6;
+    int r, size;

It is preferrable to use 'size_t' for memory allocation sizes.

target/i386/kvm/tdx.c line 96

Quoted 13 lines
target/i386/kvm/tdx.c +13 -0
+
+    do {
+        size = sizeof(struct kvm_tdx_capabilities) +
+               nr_cpuid_configs * sizeof(struct kvm_tdx_cpuid_config);
+        caps = g_malloc0(size);
+        caps->nr_cpuid_configs = nr_cpuid_configs;
+
+        r = tdx_vm_ioctl(KVM_TDX_CAPABILITIES, 0, caps);
+        if (r == -E2BIG) {
+            g_free(caps);
+            nr_cpuid_configs *= 2;
+            if (nr_cpuid_configs > KVM_MAX_CPUID_ENTRIES) {
+                error_report("KVM TDX seems broken that number of CPUID entries in kvm_tdx_capabilities exceeds limit");

Include the limit in the error message, so if we ever need to change the limit, it'll be clear what limit the QEMU version was built with.

Also use error_setg(errp, ...);

target/i386/kvm/tdx.c line 97

Quoted 1 line
target/i386/kvm/tdx.c +1 -0
+                exit(1);

Return -1

target/i386/kvm/tdx.c line 101

Quoted 4 lines
target/i386/kvm/tdx.c +4 -0
+            }
+        } else if (r < 0) {
+            g_free(caps);
+            error_report("KVM_TDX_CAPABILITIES failed: %s", strerror(-r));

Use error_setg_errno(errp, ...) instead of calling strerror yourself;

target/i386/kvm/tdx.c line 102

Quoted 1 line
target/i386/kvm/tdx.c +1 -0
+            exit(1);

Return -1

target/i386/kvm/tdx.c line 107

Quoted 5 lines
target/i386/kvm/tdx.c +5 -0
+        }
+    }
+    while (r == -E2BIG);
+
+    tdx_caps = caps;

Return 0

target/i386/kvm/tdx.c line 113

Quoted 6 lines
target/i386/kvm/tdx.c +4 -0
+}
+
  int tdx_kvm_init(MachineState *ms, Error **errp)
  {
+    if (!tdx_caps) {
+        get_tdx_capabilities();

Pass 'errp' into this method, and check return value for failure

Quoted 8 lines
target/i386/kvm/tdx.c +2 -0
+    }
+
      return 0;
  }
 
Signature · 1 line
2.34.1

With regards,
Daniel

#15 lore raw

Re: [PATCH v2 06/58] i386/tdx: Get tdx_capabilities via KVM_TDX_CAPABILITIES

Quoted 77 lines · Xiaoyao Li
Quoted 67 lines · Daniel P. Berrangé
Quoted 66 lines · Xiaoyao Li

KVM provides TDX capabilities via sub command KVM_TDX_CAPABILITIES of IOCTL(KVM_MEMORY_ENCRYPT_OP). Get the capabilities when initializing TDX context. It will be used to validate user's setting later.

Since there is no interface reporting how many cpuid configs contains in KVM_TDX_CAPABILITIES, QEMU chooses to try starting with a known number and abort when it exceeds KVM_MAX_CPUID_ENTRIES.

Besides, introduce the interfaces to invoke TDX "ioctls" at different scope (KVM, VM and VCPU) in preparation.

---
changes from v1:
- Make the error message more clear;

changes from RFC v4:
- start from nr_cpuid_configs = 6 for the loop;
- stop the loop when nr_cpuid_configs exceeds KVM_MAX_CPUID_ENTRIES;

---
  target/i386/kvm/kvm.c      |  2 -
  target/i386/kvm/kvm_i386.h |  2 +
  target/i386/kvm/tdx.c      | 93 ++++++++++++++++++++++++++++++++++++++
  3 files changed, 95 insertions(+), 2 deletions(-)
target/i386/kvm/kvm.c +0 -2
index d6b988d6c2d1..ec5c07bffd38 100644
@@ -1751,8 +1751,6 @@ static int hyperv_init_vcpu(X86CPU *cpu)
  static Error *invtsc_mig_blocker;
-#define KVM_MAX_CPUID_ENTRIES  100
-
  static void kvm_init_xsave(CPUX86State *env)
  {
      if (has_xsave2) {
target/i386/kvm/kvm_i386.h +2 -0
index ea3a5b174ac0..769eadbba56c 100644
@@ -13,6 +13,8 @@
  #include "sysemu/kvm.h"
+#define KVM_MAX_CPUID_ENTRIES  100
+
  #define kvm_apic_in_kernel() (kvm_irqchip_in_kernel())
  #ifdef CONFIG_KVM
target/i386/kvm/tdx.c +9 -0
index 77e33ae01147..255c47a2a553 100644
@@ -12,14 +12,107 @@
   */
  #include "qemu/osdep.h"
+#include "qemu/error-report.h"
  #include "qapi/error.h"
  #include "qom/object_interfaces.h"
+#include "sysemu/kvm.h"
  #include "hw/i386/x86.h"
+#include "kvm_i386.h"
  #include "tdx.h"
+static struct kvm_tdx_capabilities *tdx_caps;
+
+enum tdx_ioctl_level{
+    TDX_PLATFORM_IOCTL,
+    TDX_VM_IOCTL,
+    TDX_VCPU_IOCTL,
Quoted 8 lines

target/i386/kvm/tdx.c line 33

Quoted 4 lines
target/i386/kvm/tdx.c +4 -0
+};
+
+static int __tdx_ioctl(void *state, enum tdx_ioctl_level level, int cmd_id,
+                        __u32 flags, void *data)

Names with an initial double underscore are reserved for us by the platform implementation, so shouldn't be used in userspace app code.

Quoted 1 line

How about tdx_ioctl_internal() ?

Sure, that's fine.

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
author
#16 lore raw

PATCH v2 7/58 i386/tdx: Introduce is_tdx_vm() helper and cache tdx_guest object

In reply to Xiaoyao Li

It will need special handling for TDX VMs all around the QEMU. Introduce is_tdx_vm() helper to query if it's a TDX VM.

Cache tdx_guest object thus no need to cast from ms->cgs every time.

---
 target/i386/kvm/tdx.c | 13 +++++++++++++
 target/i386/kvm/tdx.h | 10 ++++++++++
 2 files changed, 23 insertions(+)
target/i386/kvm/tdx.c +13 -0
index 255c47a2a553..56cb826f6125 100644
@@ -21,8 +21,16 @@
 #include "kvm_i386.h"
 #include "tdx.h"
 
+static TdxGuest *tdx_guest;
+
 static struct kvm_tdx_capabilities *tdx_caps;
 
+/* It's valid after kvm_confidential_guest_init()->kvm_tdx_init() */
+bool is_tdx_vm(void)
+{
+    return !!tdx_guest;
+}
+
 enum tdx_ioctl_level{
     TDX_PLATFORM_IOCTL,
     TDX_VM_IOCTL,
@@ -109,10 +117,15 @@ static void get_tdx_capabilities(void)
 
 int tdx_kvm_init(MachineState *ms, Error **errp)
 {
+    TdxGuest *tdx = (TdxGuest *)object_dynamic_cast(OBJECT(ms->cgs),
+                                                    TYPE_TDX_GUEST);
+
     if (!tdx_caps) {
         get_tdx_capabilities();
     }
 
+    tdx_guest = tdx;
+
     return 0;
 }
 
target/i386/kvm/tdx.h +10 -0
index c8a23d95258d..4036ca2f3f99 100644
@@ -1,6 +1,10 @@
 #ifndef QEMU_I386_TDX_H
 #define QEMU_I386_TDX_H
 
+#ifndef CONFIG_USER_ONLY
+#include CONFIG_DEVICES /* CONFIG_TDX */
+#endif
+
 #include "exec/confidential-guest-support.h"
 
 #define TYPE_TDX_GUEST "tdx-guest"
@@ -16,6 +20,12 @@ typedef struct TdxGuest {
     uint64_t attributes;    /* TD attributes */
 } TdxGuest;
 
+#ifdef CONFIG_TDX
+bool is_tdx_vm(void);
+#else
+#define is_tdx_vm() 0
+#endif /* CONFIG_TDX */
+
 int tdx_kvm_init(MachineState *ms, Error **errp);
 
 #endif /* QEMU_I386_TDX_H */
Signature · 1 line
2.34.1
#17 lore raw

Re: [PATCH v2 07/58] i386/tdx: Introduce is_tdx_vm() helper and cache tdx_guest object

Quoted 34 lines · Xiaoyao Li

It will need special handling for TDX VMs all around the QEMU. Introduce is_tdx_vm() helper to query if it's a TDX VM.

Cache tdx_guest object thus no need to cast from ms->cgs every time.

---
 target/i386/kvm/tdx.c | 13 +++++++++++++
 target/i386/kvm/tdx.h | 10 ++++++++++
 2 files changed, 23 insertions(+)
target/i386/kvm/tdx.c +8 -0
index 255c47a2a553..56cb826f6125 100644
@@ -21,8 +21,16 @@
 #include "kvm_i386.h"
 #include "tdx.h"
 
+static TdxGuest *tdx_guest;
+
 static struct kvm_tdx_capabilities *tdx_caps;
 
+/* It's valid after kvm_confidential_guest_init()->kvm_tdx_init() */
+bool is_tdx_vm(void)
+{
+    return !!tdx_guest;
+}
+
 enum tdx_ioctl_level{
     TDX_PLATFORM_IOCTL,
     TDX_VM_IOCTL,

target/i386/kvm/tdx.c static void get_tdx_capabilities(void) · line 121

Quoted 5 lines
target/i386/kvm/tdx.c +2 -0
@@ -109,10 +117,15 @@ static void get_tdx_capabilities(void)
 
 int tdx_kvm_init(MachineState *ms, Error **errp)
 {
+    TdxGuest *tdx = (TdxGuest *)object_dynamic_cast(OBJECT(ms->cgs),
+                                                    TYPE_TDX_GUEST);

This method can return NULL. Presumably tdx_kvm_init() should only be called if we already checked ms->cgs == TYPE_TDX_GUEST. If so then use object_dynamic_cast_assert() instead.

Quoted 41 lines
target/i386/kvm/tdx.c +3 -0
+
     if (!tdx_caps) {
         get_tdx_capabilities();
     }
 
+    tdx_guest = tdx;
+
     return 0;
 }
 
target/i386/kvm/tdx.h +10 -0
index c8a23d95258d..4036ca2f3f99 100644
@@ -1,6 +1,10 @@
 #ifndef QEMU_I386_TDX_H
 #define QEMU_I386_TDX_H
 
+#ifndef CONFIG_USER_ONLY
+#include CONFIG_DEVICES /* CONFIG_TDX */
+#endif
+
 #include "exec/confidential-guest-support.h"
 
 #define TYPE_TDX_GUEST "tdx-guest"
@@ -16,6 +20,12 @@ typedef struct TdxGuest {
     uint64_t attributes;    /* TD attributes */
 } TdxGuest;
 
+#ifdef CONFIG_TDX
+bool is_tdx_vm(void);
+#else
+#define is_tdx_vm() 0
+#endif /* CONFIG_TDX */
+
 int tdx_kvm_init(MachineState *ms, Error **errp);
 
 #endif /* QEMU_I386_TDX_H */
Signature · 1 line
2.34.1

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
author
#18 lore raw

Re: [PATCH v2 07/58] i386/tdx: Introduce is_tdx_vm() helper and cache tdx_guest object

Quoted 36 lines · Daniel P. Berrangé
Quoted 35 lines · Xiaoyao Li

It will need special handling for TDX VMs all around the QEMU. Introduce is_tdx_vm() helper to query if it's a TDX VM.

Cache tdx_guest object thus no need to cast from ms->cgs every time.

---
  target/i386/kvm/tdx.c | 13 +++++++++++++
  target/i386/kvm/tdx.h | 10 ++++++++++
  2 files changed, 23 insertions(+)
target/i386/kvm/tdx.c +8 -0
index 255c47a2a553..56cb826f6125 100644
@@ -21,8 +21,16 @@
  #include "kvm_i386.h"
  #include "tdx.h"
 
+static TdxGuest *tdx_guest;
+
  static struct kvm_tdx_capabilities *tdx_caps;
 
+/* It's valid after kvm_confidential_guest_init()->kvm_tdx_init() */
+bool is_tdx_vm(void)
+{
+    return !!tdx_guest;
+}
+
  enum tdx_ioctl_level{
      TDX_PLATFORM_IOCTL,
      TDX_VM_IOCTL,
@@ -109,10 +117,15 @@ static void get_tdx_capabilities(void)
 
Quoted 9 lines

target/i386/kvm/tdx.c static void get_tdx_capabilities(void) · line 121

Quoted 4 lines
target/i386/kvm/tdx.c +2 -0
  int tdx_kvm_init(MachineState *ms, Error **errp)
  {
+    TdxGuest *tdx = (TdxGuest *)object_dynamic_cast(OBJECT(ms->cgs),
+                                                    TYPE_TDX_GUEST);

This method can return NULL. Presumably tdx_kvm_init() should only be called if we already checked ms->cgs == TYPE_TDX_GUEST. If so then use object_dynamic_cast_assert() instead.

object_dynamic_cast_assert() is for OBJECT_CHECK() and INTERFACE_CHECK().

So I will use TDX_GUEST(OBJECT(ms->cgs)) (introduced in patch 2) instead, which is the wrapper of OBJECT_CHECK().

author
#19 lore raw

PATCH v2 8/58 i386/tdx: Adjust the supported CPUID based on TDX restrictions

In reply to Xiaoyao Li

According to Chapter "CPUID Virtualization" in TDX module spec, CPUID bits of TD can be classified into 6 types:

------------------------------------------------------------------------
1 | As configured | configurable by VMM, independent of native value;
------------------------------------------------------------------------
2 | As configured | configurable by VMM if the bit is supported natively
(if native) | Otherwise it equals as native(0).
------------------------------------------------------------------------
3 | Fixed | fixed to 0/1
------------------------------------------------------------------------
4 | Native | reflect the native value
------------------------------------------------------------------------
5 | Calculated | calculated by TDX module.
------------------------------------------------------------------------
6 | Inducing #VE | get #VE exception
------------------------------------------------------------------------

Note:
1. All the configurable XFAM related features and TD attributes related features fall into type #2. And fixed0/1 bits of XFAM and TD attributes fall into type #3.

2. For CPUID leaves not listed in "CPUID virtualization Overview" table in TDX module spec, TDX module injects #VE to TDs when those are queried. For this case, TDs can request CPUID emulation from VMM via TDVMCALL and the values are fully controlled by VMM.

Due to TDX module has its own virtualization policy on CPUID bits, it leads to what reported via KVM_GET_SUPPORTED_CPUID diverges from the supported CPUID bits for TDs. In order to keep a consistent CPUID configuration between VMM and TDs. Adjust supported CPUID for TDs based on TDX restrictions.

Currently only focus on the CPUID leaves recognized by QEMU's feature_word_info[] that are indexed by a FeatureWord.

Introduce a TDX CPUID lookup table, which maintains 1 entry for each FeatureWord. Each entry has below fields:

- tdx_fixed0/1: The bits that are fixed as 0/1;

- vmm_fixup: The bits that are configurable from the view of TDX module.
But they requires emulation of VMM when they are configured as enabled. For those, they are not supported if VMM doesn't report them as supported. So they need be fixed up by checking if VMM supports them.

- inducing_ve: TD gets #VE when querying this CPUID leaf. The result is totally configurable by VMM.

- supported_on_ve: It's valid only when @inducing_ve is true. It represents the maximum feature set supported that be emulated for TDs.

By applying TDX CPUID lookup table and TDX capabilities reported from TDX module, the supported CPUID for TDs can be obtained from following steps:

- get the base of VMM supported feature set;

- if the leaf is not a FeatureWord just return VMM's value without modification;

- if the leaf is an inducing_ve type, applying supported_on_ve mask and return;

- include all native bits, it covers type #2, #4, and parts of type #1. (it also includes some unsupported bits. The following step will correct it.)

- apply fixed0/1 to it (it covers #3, and rectifies the previous step);

- add configurable bits (it covers the other part of type #1);

- fix the ones in vmm_fixup;

- filter the one has valid .supported field;

(Calculated type is ignored since it's determined at runtime).

---
 target/i386/cpu.h     |  16 +++
 target/i386/kvm/kvm.c |   4 +
 target/i386/kvm/tdx.c | 254 ++++++++++++++++++++++++++++++++++++++++++
 target/i386/kvm/tdx.h |   2 +
 4 files changed, 276 insertions(+)
target/i386/cpu.h +16 -0
index e0771a10433b..c93dcd274531 100644
@@ -780,6 +780,8 @@ uint64_t x86_cpu_get_supported_feature_word(FeatureWord w,
 
 /* Support RDFSBASE/RDGSBASE/WRFSBASE/WRGSBASE */
 #define CPUID_7_0_EBX_FSGSBASE          (1U << 0)
+/* Support for TSC adjustment MSR 0x3B */
+#define CPUID_7_0_EBX_TSC_ADJUST        (1U << 1)
 /* Support SGX */
 #define CPUID_7_0_EBX_SGX               (1U << 2)
 /* 1st Group of Advanced Bit Manipulation Extensions */
@@ -798,8 +800,12 @@ uint64_t x86_cpu_get_supported_feature_word(FeatureWord w,
 #define CPUID_7_0_EBX_INVPCID           (1U << 10)
 /* Restricted Transactional Memory */
 #define CPUID_7_0_EBX_RTM               (1U << 11)
+/* Cache QoS Monitoring */
+#define CPUID_7_0_EBX_PQM               (1U << 12)
 /* Memory Protection Extension */
 #define CPUID_7_0_EBX_MPX               (1U << 14)
+/* Resource Director Technology Allocation */
+#define CPUID_7_0_EBX_RDT_A             (1U << 15)
 /* AVX-512 Foundation */
 #define CPUID_7_0_EBX_AVX512F           (1U << 16)
 /* AVX-512 Doubleword & Quadword Instruction */
@@ -855,10 +861,16 @@ uint64_t x86_cpu_get_supported_feature_word(FeatureWord w,
 #define CPUID_7_0_ECX_AVX512VNNI        (1U << 11)
 /* Support for VPOPCNT[B,W] and VPSHUFBITQMB */
 #define CPUID_7_0_ECX_AVX512BITALG      (1U << 12)
+/* Intel Total Memory Encryption */
+#define CPUID_7_0_ECX_TME               (1U << 13)
 /* POPCNT for vectors of DW/QW */
 #define CPUID_7_0_ECX_AVX512_VPOPCNTDQ  (1U << 14)
+/* Placeholder for bit 15 */
+#define CPUID_7_0_ECX_FZM               (1U << 15)
 /* 5-level Page Tables */
 #define CPUID_7_0_ECX_LA57              (1U << 16)
+/* MAWAU for MPX */
+#define CPUID_7_0_ECX_MAWAU             (31U << 17)
 /* Read Processor ID */
 #define CPUID_7_0_ECX_RDPID             (1U << 22)
 /* Bus Lock Debug Exception */
@@ -869,6 +881,8 @@ uint64_t x86_cpu_get_supported_feature_word(FeatureWord w,
 #define CPUID_7_0_ECX_MOVDIRI           (1U << 27)
 /* Move 64 Bytes as Direct Store Instruction */
 #define CPUID_7_0_ECX_MOVDIR64B         (1U << 28)
+/* ENQCMD and ENQCMDS instructions */
+#define CPUID_7_0_ECX_ENQCMD            (1U << 29)
 /* Support SGX Launch Control */
 #define CPUID_7_0_ECX_SGX_LC            (1U << 30)
 /* Protection Keys for Supervisor-mode Pages */
@@ -886,6 +900,8 @@ uint64_t x86_cpu_get_supported_feature_word(FeatureWord w,
 #define CPUID_7_0_EDX_SERIALIZE         (1U << 14)
 /* TSX Suspend Load Address Tracking instruction */
 #define CPUID_7_0_EDX_TSX_LDTRK         (1U << 16)
+/* PCONFIG instruction */
+#define CPUID_7_0_EDX_PCONFIG           (1U << 18)
 /* Architectural LBRs */
 #define CPUID_7_0_EDX_ARCH_LBR          (1U << 19)
 /* AMX_BF16 instruction */
target/i386/kvm/kvm.c +4 -0
index ec5c07bffd38..46a455a1e331 100644
@@ -539,6 +539,10 @@ uint32_t kvm_arch_get_supported_cpuid(KVMState *s, uint32_t function,
         ret |= 1U << KVM_HINTS_REALTIME;
     }
 
+    if (is_tdx_vm()) {
+        tdx_get_supported_cpuid(function, index, reg, &ret);
+    }
+
     return ret;
 }
 
target/i386/kvm/tdx.c +254 -0

271 lines not on this page · raw message

target/i386/kvm/tdx.h +2 -0
index 4036ca2f3f99..06599b65b827 100644
@@ -27,5 +27,7 @@ bool is_tdx_vm(void);
 #endif /* CONFIG_TDX */
 
 int tdx_kvm_init(MachineState *ms, Error **errp);
+void tdx_get_supported_cpuid(uint32_t function, uint32_t index, int reg,
+                             uint32_t *ret);
 
 #endif /* QEMU_I386_TDX_H */
Signature · 1 line
2.34.1
#20 lore raw

Re: [PATCH v2 08/58] i386/tdx: Adjust the supported CPUID based on TDX restrictions

Quoted 4 lines · Xiaoyao Li
target/i386/kvm/tdx.c +0 -0
index 56cb826f6125..3198bc9fd5fb 100644

...

Quoted 10 lines
target/i386/kvm/tdx.c +10 -0
+static inline uint32_t host_cpuid_reg(uint32_t function,
+                                      uint32_t index, int reg)
+{
+    uint32_t eax, ebx, ecx, edx;
+    uint32_t ret = 0;
+
+    host_cpuid(function, index, &eax, &ebx, &ecx, &edx);
+
+    switch (reg) {
+    case R_EAX:

target/i386/kvm/tdx.c bool is_tdx_vm(void) · line 171

Quoted 10 lines
target/i386/kvm/tdx.c +10 -0
+        ret |= eax;
+        break;
+    case R_EBX:
+        ret |= ebx;
+        break;
+    case R_ECX:
+        ret |= ecx;
+        break;
+    case R_EDX:
+        ret |= edx;

Nitpick: "|" isn't needed as we initialize ret = 0 above. Just '='.

Signature · 1 line
Isaku Yamahata <[email protected]>
author
#21 lore raw

Re: [PATCH v2 08/58] i386/tdx: Adjust the supported CPUID based on TDX restrictions

Quoted 22 lines · Isaku Yamahata
Quoted 4 lines · Xiaoyao Li
target/i386/kvm/tdx.c +0 -0
index 56cb826f6125..3198bc9fd5fb 100644

...

Quoted 14 lines
target/i386/kvm/tdx.c +14 -0
+static inline uint32_t host_cpuid_reg(uint32_t function,
+                                      uint32_t index, int reg)
+{
+    uint32_t eax, ebx, ecx, edx;
+    uint32_t ret = 0;
+
+    host_cpuid(function, index, &eax, &ebx, &ecx, &edx);
+
+    switch (reg) {
+    case R_EAX:
+        ret |= eax;
+        break;
+    case R_EBX:
+        ret |= ebx;
Quoted 8 lines

target/i386/kvm/tdx.c bool is_tdx_vm(void) · line 171

Quoted 6 lines
target/i386/kvm/tdx.c +6 -0
+        break;
+    case R_ECX:
+        ret |= ecx;
+        break;
+    case R_EDX:
+        ret |= edx;

Nitpick: "|" isn't needed as we initialize ret = 0 above. Just '='.

Will fix it in next version.

thanks!

#22 lore raw

Re: [PATCH v2 08/58] i386/tdx: Adjust the supported CPUID based on TDX restrictions

In reply to Xiaoyao Li

Hi,

Quoted 221 lines · Xiaoyao Li

According to Chapter "CPUID Virtualization" in TDX module spec, CPUID bits of TD can be classified into 6 types:

------------------------------------------------------------------------
1 | As configured | configurable by VMM, independent of native value;
------------------------------------------------------------------------
2 | As configured | configurable by VMM if the bit is supported natively
(if native) | Otherwise it equals as native(0).
------------------------------------------------------------------------
3 | Fixed | fixed to 0/1
------------------------------------------------------------------------
4 | Native | reflect the native value
------------------------------------------------------------------------
5 | Calculated | calculated by TDX module.
------------------------------------------------------------------------
6 | Inducing #VE | get #VE exception
------------------------------------------------------------------------

Note:
1. All the configurable XFAM related features and TD attributes related features fall into type #2. And fixed0/1 bits of XFAM and TD attributes fall into type #3.

2. For CPUID leaves not listed in "CPUID virtualization Overview" table in TDX module spec, TDX module injects #VE to TDs when those are queried. For this case, TDs can request CPUID emulation from VMM via TDVMCALL and the values are fully controlled by VMM.

Due to TDX module has its own virtualization policy on CPUID bits, it leads to what reported via KVM_GET_SUPPORTED_CPUID diverges from the supported CPUID bits for TDs. In order to keep a consistent CPUID configuration between VMM and TDs. Adjust supported CPUID for TDs based on TDX restrictions.

Currently only focus on the CPUID leaves recognized by QEMU's feature_word_info[] that are indexed by a FeatureWord.

Introduce a TDX CPUID lookup table, which maintains 1 entry for each FeatureWord. Each entry has below fields:

- tdx_fixed0/1: The bits that are fixed as 0/1;

- vmm_fixup: The bits that are configurable from the view of TDX module.
But they requires emulation of VMM when they are configured
as enabled. For those, they are not supported if VMM doesn't report them as supported. So they need be fixed up by checking if VMM supports them.

- inducing_ve: TD gets #VE when querying this CPUID leaf. The result is totally configurable by VMM.

- supported_on_ve: It's valid only when @inducing_ve is true. It represents the maximum feature set supported that be emulated for TDs.

By applying TDX CPUID lookup table and TDX capabilities reported from TDX module, the supported CPUID for TDs can be obtained from following steps:

- get the base of VMM supported feature set;

- if the leaf is not a FeatureWord just return VMM's value without modification;

- if the leaf is an inducing_ve type, applying supported_on_ve mask and return;

- include all native bits, it covers type #2, #4, and parts of type #1. (it also includes some unsupported bits. The following step will correct it.)

- apply fixed0/1 to it (it covers #3, and rectifies the previous step);

- add configurable bits (it covers the other part of type #1);

- fix the ones in vmm_fixup;

- filter the one has valid .supported field;

(Calculated type is ignored since it's determined at runtime).

---
  target/i386/cpu.h     |  16 +++
  target/i386/kvm/kvm.c |   4 +
  target/i386/kvm/tdx.c | 254 ++++++++++++++++++++++++++++++++++++++++++
  target/i386/kvm/tdx.h |   2 +
  4 files changed, 276 insertions(+)
target/i386/cpu.h +16 -0
index e0771a10433b..c93dcd274531 100644
@@ -780,6 +780,8 @@ uint64_t x86_cpu_get_supported_feature_word(FeatureWord w,
 
  /* Support RDFSBASE/RDGSBASE/WRFSBASE/WRGSBASE */
  #define CPUID_7_0_EBX_FSGSBASE          (1U << 0)
+/* Support for TSC adjustment MSR 0x3B */
+#define CPUID_7_0_EBX_TSC_ADJUST        (1U << 1)
  /* Support SGX */
  #define CPUID_7_0_EBX_SGX               (1U << 2)
  /* 1st Group of Advanced Bit Manipulation Extensions */
@@ -798,8 +800,12 @@ uint64_t x86_cpu_get_supported_feature_word(FeatureWord w,
  #define CPUID_7_0_EBX_INVPCID           (1U << 10)
  /* Restricted Transactional Memory */
  #define CPUID_7_0_EBX_RTM               (1U << 11)
+/* Cache QoS Monitoring */
+#define CPUID_7_0_EBX_PQM               (1U << 12)
  /* Memory Protection Extension */
  #define CPUID_7_0_EBX_MPX               (1U << 14)
+/* Resource Director Technology Allocation */
+#define CPUID_7_0_EBX_RDT_A             (1U << 15)
  /* AVX-512 Foundation */
  #define CPUID_7_0_EBX_AVX512F           (1U << 16)
  /* AVX-512 Doubleword & Quadword Instruction */
@@ -855,10 +861,16 @@ uint64_t x86_cpu_get_supported_feature_word(FeatureWord w,
  #define CPUID_7_0_ECX_AVX512VNNI        (1U << 11)
  /* Support for VPOPCNT[B,W] and VPSHUFBITQMB */
  #define CPUID_7_0_ECX_AVX512BITALG      (1U << 12)
+/* Intel Total Memory Encryption */
+#define CPUID_7_0_ECX_TME               (1U << 13)
  /* POPCNT for vectors of DW/QW */
  #define CPUID_7_0_ECX_AVX512_VPOPCNTDQ  (1U << 14)
+/* Placeholder for bit 15 */
+#define CPUID_7_0_ECX_FZM               (1U << 15)
  /* 5-level Page Tables */
  #define CPUID_7_0_ECX_LA57              (1U << 16)
+/* MAWAU for MPX */
+#define CPUID_7_0_ECX_MAWAU             (31U << 17)
  /* Read Processor ID */
  #define CPUID_7_0_ECX_RDPID             (1U << 22)
  /* Bus Lock Debug Exception */
@@ -869,6 +881,8 @@ uint64_t x86_cpu_get_supported_feature_word(FeatureWord w,
  #define CPUID_7_0_ECX_MOVDIRI           (1U << 27)
  /* Move 64 Bytes as Direct Store Instruction */
  #define CPUID_7_0_ECX_MOVDIR64B         (1U << 28)
+/* ENQCMD and ENQCMDS instructions */
+#define CPUID_7_0_ECX_ENQCMD            (1U << 29)
  /* Support SGX Launch Control */
  #define CPUID_7_0_ECX_SGX_LC            (1U << 30)
  /* Protection Keys for Supervisor-mode Pages */
@@ -886,6 +900,8 @@ uint64_t x86_cpu_get_supported_feature_word(FeatureWord w,
  #define CPUID_7_0_EDX_SERIALIZE         (1U << 14)
  /* TSX Suspend Load Address Tracking instruction */
  #define CPUID_7_0_EDX_TSX_LDTRK         (1U << 16)
+/* PCONFIG instruction */
+#define CPUID_7_0_EDX_PCONFIG           (1U << 18)
  /* Architectural LBRs */
  #define CPUID_7_0_EDX_ARCH_LBR          (1U << 19)
  /* AMX_BF16 instruction */
target/i386/kvm/kvm.c +4 -0
index ec5c07bffd38..46a455a1e331 100644
@@ -539,6 +539,10 @@ uint32_t kvm_arch_get_supported_cpuid(KVMState *s, uint32_t function,
          ret |= 1U << KVM_HINTS_REALTIME;
      }
 
+    if (is_tdx_vm()) {
+        tdx_get_supported_cpuid(function, index, reg, &ret);
+    }
+
      return ret;
  }
 
target/i386/kvm/tdx.c +41 -0
index 56cb826f6125..3198bc9fd5fb 100644
@@ -15,11 +15,129 @@
  #include "qemu/error-report.h"
  #include "qapi/error.h"
  #include "qom/object_interfaces.h"
+#include "standard-headers/asm-x86/kvm_para.h"
  #include "sysemu/kvm.h"
+#include "sysemu/sysemu.h"
 
  #include "hw/i386/x86.h"
  #include "kvm_i386.h"
  #include "tdx.h"
+#include "../cpu-internal.h"
+
+#define TDX_SUPPORTED_KVM_FEATURES  ((1U << KVM_FEATURE_NOP_IO_DELAY) | \
+                                     (1U << KVM_FEATURE_PV_UNHALT) | \
+                                     (1U << KVM_FEATURE_PV_TLB_FLUSH) | \
+                                     (1U << KVM_FEATURE_PV_SEND_IPI) | \
+                                     (1U << KVM_FEATURE_POLL_CONTROL) | \
+                                     (1U << KVM_FEATURE_PV_SCHED_YIELD) | \
+                                     (1U << KVM_FEATURE_MSI_EXT_DEST_ID))
+
+typedef struct KvmTdxCpuidLookup {
+    uint32_t tdx_fixed0;
+    uint32_t tdx_fixed1;
+
+    /*
+     * The CPUID bits that are configurable from the view of TDX module
+     * but require VMM emulation if configured to enabled by VMM.
+     *
+     * For those bits, they cannot be enabled actually if VMM (KVM/QEMU) cannot
+     * virtualize them.
+     */
+    uint32_t vmm_fixup;
+
+    bool inducing_ve;
+    /*
+     * The maximum supported feature set for given inducing-#VE leaf.
+     * It's valid only when .inducing_ve is true.
+     */
+    uint32_t supported_on_ve;
+} KvmTdxCpuidLookup;
+
+ /*
+  * QEMU maintained TDX CPUID lookup tables, which reflects how CPUIDs are
+  * virtualized for guest TDs based on "CPUID virtualization" of TDX spec.
+  *
+  * Note:
+  *
+  * This table will be updated runtime by tdx_caps reported by platform.
+  *

target/i386/kvm/tdx.c line 73

Quoted 10 lines
target/i386/kvm/tdx.c +10 -0
+  */
+static KvmTdxCpuidLookup tdx_cpuid_lookup[FEATURE_WORDS] = {
+    [FEAT_1_EDX] = {
+        .tdx_fixed0 =
+            BIT(10) /* Reserved */ | BIT(20) /* Reserved */ | CPUID_IA64,
+        .tdx_fixed1 =
+            CPUID_MSR | CPUID_PAE | CPUID_MCE | CPUID_APIC |
+            CPUID_MTRR | CPUID_MCA | CPUID_CLFLUSH | CPUID_DTS,
+        .vmm_fixup =
+            CPUID_ACPI | CPUID_PBE,

CPUID_HT might also be needed here, as it's disabled by QEMU when TD guest only has a single processor core.

Regards,
-Tina

author
#23 lore raw

Re: [PATCH v2 08/58] i386/tdx: Adjust the supported CPUID based on TDX restrictions

Quoted 238 lines · Tina Zhang

Hi,

Quoted 235 lines · Xiaoyao Li

According to Chapter "CPUID Virtualization" in TDX module spec, CPUID bits of TD can be classified into 6 types:

------------------------------------------------------------------------
1 | As configured | configurable by VMM, independent of native value;
------------------------------------------------------------------------
2 | As configured | configurable by VMM if the bit is supported natively (if native)   | Otherwise it equals as native(0).
------------------------------------------------------------------------
3 | Fixed         | fixed to 0/1
------------------------------------------------------------------------
4 | Native        | reflect the native value
------------------------------------------------------------------------
5 | Calculated    | calculated by TDX module.
------------------------------------------------------------------------
6 | Inducing #VE  | get #VE exception
------------------------------------------------------------------------

Note:
1. All the configurable XFAM related features and TD attributes related features fall into type #2. And fixed0/1 bits of XFAM and TD attributes fall into type #3.

2. For CPUID leaves not listed in "CPUID virtualization Overview" table in TDX module spec, TDX module injects #VE to TDs when those are queried. For this case, TDs can request CPUID emulation from VMM via TDVMCALL and the values are fully controlled by VMM.

Due to TDX module has its own virtualization policy on CPUID bits, it leads to what reported via KVM_GET_SUPPORTED_CPUID diverges from the supported CPUID bits for TDs. In order to keep a consistent CPUID configuration between VMM and TDs. Adjust supported CPUID for TDs based on TDX restrictions.

Currently only focus on the CPUID leaves recognized by QEMU's feature_word_info[] that are indexed by a FeatureWord.

Introduce a TDX CPUID lookup table, which maintains 1 entry for each FeatureWord. Each entry has below fields:

  - tdx_fixed0/1: The bits that are fixed as 0/1;

  - vmm_fixup:   The bits that are configurable from the view of TDX module. But they requires emulation of VMM when they are configured as enabled. For those, they are not supported if VMM doesn't report them as supported. So they need be fixed up by checking if VMM supports them.

  - inducing_ve: TD gets #VE when querying this CPUID leaf. The result is totally configurable by VMM.

  - supported_on_ve: It's valid only when @inducing_ve is true. It represents the maximum feature set supported that be emulated for TDs.

By applying TDX CPUID lookup table and TDX capabilities reported from TDX module, the supported CPUID for TDs can be obtained from following steps:

- get the base of VMM supported feature set;

- if the leaf is not a FeatureWord just return VMM's value without modification;

- if the leaf is an inducing_ve type, applying supported_on_ve mask and return;

- include all native bits, it covers type #2, #4, and parts of type #1. (it also includes some unsupported bits. The following step will correct it.)

- apply fixed0/1 to it (it covers #3, and rectifies the previous step);

- add configurable bits (it covers the other part of type #1);

- fix the ones in vmm_fixup;

- filter the one has valid .supported field;

(Calculated type is ignored since it's determined at runtime).

---
  target/i386/cpu.h     |  16 +++

  target/i386/kvm/kvm.c |   4 +
  target/i386/kvm/tdx.c | 254 ++++++++++++++++++++++++++++++++++++++++++
  target/i386/kvm/tdx.h |   2 +
  4 files changed, 276 insertions(+)
target/i386/cpu.h +6 -0
index e0771a10433b..c93dcd274531 100644
@@ -780,6 +780,8 @@ uint64_t 
 x86_cpu_get_supported_feature_word(FeatureWord w,
   /* Support RDFSBASE/RDGSBASE/WRFSBASE/WRGSBASE */
   #define CPUID_7_0_EBX_FSGSBASE          (1U << 0)
+/* Support for TSC adjustment MSR 0x3B */
+#define CPUID_7_0_EBX_TSC_ADJUST        (1U << 1)
   /* Support SGX */
   #define CPUID_7_0_EBX_SGX               (1U << 2)
   /* 1st Group of Advanced Bit Manipulation Extensions */
@@ -798,8 +800,12 @@ uint64_t 
 x86_cpu_get_supported_feature_word(FeatureWord w,
   #define CPUID_7_0_EBX_INVPCID           (1U << 10)
   /* Restricted Transactional Memory */
   #define CPUID_7_0_EBX_RTM               (1U << 11)
+/* Cache QoS Monitoring */
+#define CPUID_7_0_EBX_PQM               (1U << 12)
   /* Memory Protection Extension */
   #define CPUID_7_0_EBX_MPX               (1U << 14)
+/* Resource Director Technology Allocation */
+#define CPUID_7_0_EBX_RDT_A             (1U << 15)
   /* AVX-512 Foundation */
   #define CPUID_7_0_EBX_AVX512F           (1U << 16)

  /* AVX-512 Doubleword & Quadword Instruction */

target/i386/cpu.h +6 -0
@@ -855,10 +861,16 @@ uint64_t 
 x86_cpu_get_supported_feature_word(FeatureWord w,
   #define CPUID_7_0_ECX_AVX512VNNI        (1U << 11)
   /* Support for VPOPCNT[B,W] and VPSHUFBITQMB */
   #define CPUID_7_0_ECX_AVX512BITALG      (1U << 12)
+/* Intel Total Memory Encryption */
+#define CPUID_7_0_ECX_TME               (1U << 13)
   /* POPCNT for vectors of DW/QW */
   #define CPUID_7_0_ECX_AVX512_VPOPCNTDQ  (1U << 14)
+/* Placeholder for bit 15 */
+#define CPUID_7_0_ECX_FZM               (1U << 15)
   /* 5-level Page Tables */
   #define CPUID_7_0_ECX_LA57              (1U << 16)
+/* MAWAU for MPX */
+#define CPUID_7_0_ECX_MAWAU             (31U << 17)
   /* Read Processor ID */
   #define CPUID_7_0_ECX_RDPID             (1U << 22)

  /* Bus Lock Debug Exception */

target/i386/cpu.h +2 -0
@@ -869,6 +881,8 @@ uint64_t 
 x86_cpu_get_supported_feature_word(FeatureWord w,
   #define CPUID_7_0_ECX_MOVDIRI           (1U << 27)
   /* Move 64 Bytes as Direct Store Instruction */
   #define CPUID_7_0_ECX_MOVDIR64B         (1U << 28)
+/* ENQCMD and ENQCMDS instructions */
+#define CPUID_7_0_ECX_ENQCMD            (1U << 29)
   /* Support SGX Launch Control */
   #define CPUID_7_0_ECX_SGX_LC            (1U << 30)

  /* Protection Keys for Supervisor-mode Pages */

target/i386/cpu.h +2 -0
@@ -886,6 +900,8 @@ uint64_t 
 x86_cpu_get_supported_feature_word(FeatureWord w,
   #define CPUID_7_0_EDX_SERIALIZE         (1U << 14)
   /* TSX Suspend Load Address Tracking instruction */
   #define CPUID_7_0_EDX_TSX_LDTRK         (1U << 16)
+/* PCONFIG instruction */
+#define CPUID_7_0_EDX_PCONFIG           (1U << 18)
   /* Architectural LBRs */
   #define CPUID_7_0_EDX_ARCH_LBR          (1U << 19)

  /* AMX_BF16 instruction */

target/i386/kvm/kvm.c +4 -0
index ec5c07bffd38..46a455a1e331 100644
@@ -539,6 +539,10 @@ uint32_t kvm_arch_get_supported_cpuid(KVMState 
 *s, uint32_t function,
           ret |= 1U << KVM_HINTS_REALTIME;
       }
+    if (is_tdx_vm()) {
+        tdx_get_supported_cpuid(function, index, reg, &ret);
+    }
+
       return ret;
   }
target/i386/kvm/tdx.c +11 -0
index 56cb826f6125..3198bc9fd5fb 100644
@@ -15,11 +15,129 @@
   #include "qemu/error-report.h"
   #include "qapi/error.h"
   #include "qom/object_interfaces.h"
+#include "standard-headers/asm-x86/kvm_para.h"
   #include "sysemu/kvm.h"
+#include "sysemu/sysemu.h"
   #include "hw/i386/x86.h"
   #include "kvm_i386.h"
   #include "tdx.h"
+#include "../cpu-internal.h"
+
+#define TDX_SUPPORTED_KVM_FEATURES  ((1U << KVM_FEATURE_NOP_IO_DELAY) 
 | \
+                                     (1U << KVM_FEATURE_PV_UNHALT) | \
+                                     (1U << KVM_FEATURE_PV_TLB_FLUSH) 
 | \
+                                     (1U << KVM_FEATURE_PV_SEND_IPI) | \
+                                     (1U << KVM_FEATURE_POLL_CONTROL) 
 | \
+                                     (1U << 
 KVM_FEATURE_PV_SCHED_YIELD) | \
+                                     (1U << 

KVM_FEATURE_MSI_EXT_DEST_ID))

target/i386/kvm/tdx.c +10 -0
+
+typedef struct KvmTdxCpuidLookup {
+    uint32_t tdx_fixed0;
+    uint32_t tdx_fixed1;
+
+    /*
+     * The CPUID bits that are configurable from the view of TDX module
+     * but require VMM emulation if configured to enabled by VMM.
+     *
+     * For those bits, they cannot be enabled actually if VMM 

(KVM/QEMU) cannot
+     * virtualize them.
+     */
+    uint32_t vmm_fixup;

target/i386/kvm/tdx.c +11 -0
+
+    bool inducing_ve;
+    /*
+     * The maximum supported feature set for given inducing-#VE leaf.
+     * It's valid only when .inducing_ve is true.
+     */
+    uint32_t supported_on_ve;
+} KvmTdxCpuidLookup;
+
+ /*
+  * QEMU maintained TDX CPUID lookup tables, which reflects how 

CPUIDs are
+  * virtualized for guest TDs based on "CPUID virtualization" of TDX spec.
+  *
+  * Note:
+  *
+  * This table will be updated runtime by tdx_caps reported by platform. +  *
+  */
+static KvmTdxCpuidLookup tdx_cpuid_lookup[FEATURE_WORDS] = { +    [FEAT_1_EDX] = {
+        .tdx_fixed0 =

Quoted 8 lines
Quoted 7 lines

+            BIT(10) /* Reserved */ | BIT(20) /* Reserved */ | CPUID_IA64,
+        .tdx_fixed1 =
+            CPUID_MSR | CPUID_PAE | CPUID_MCE | CPUID_APIC | +            CPUID_MTRR | CPUID_MCA | CPUID_CLFLUSH | CPUID_DTS, +        .vmm_fixup =
+            CPUID_ACPI | CPUID_PBE,

CPUID_HT might also be needed here, as it's disabled by QEMU when TD guest only has a single processor core.

Add CPUID_HT here seems not correct fix. The root cause is that CPUID_HT is wrongly treated as auto_enabled bit, I will sent a fix separately.

Quoted 3 lines

Regards,
-Tina

author
#24 lore raw

PATCH v2 9/58 i386/tdx: Update tdx_cpuid_lookup[].tdx_fixed0/1 by tdx_caps.cpuid_config[]

In reply to Xiaoyao Li

tdx_cpuid_lookup[].tdx_fixed0/1 is QEMU maintained data which reflects TDX restrictions regrading how some CPUIDs are virtualized by TDX.

It's retrieved from TDX spec. However, TDX may change some fixed fields to configurable in the future. Update tdx_cpuid.lookup[].tdx_fixed0/1 fields by removing the bits that reported from TDX module as configurable. This can adapt with the updated TDX (module) automatically.

---
 target/i386/kvm/tdx.c | 30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)
target/i386/kvm/tdx.c +30 -0
index 3198bc9fd5fb..4518c79aecc8 100644
@@ -369,6 +369,34 @@ static void get_tdx_capabilities(void)
     tdx_caps = caps;
 }
 
+static void update_tdx_cpuid_lookup_by_tdx_caps(void)
+{
+    KvmTdxCpuidLookup *entry;
+    FeatureWordInfo *fi;
+    uint32_t config;
+    FeatureWord w;
+
+    /*
+     * Patch tdx_fixed0/1 by tdx_caps that what TDX module reports as
+     * configurable is not fixed.
+     */
+    for (w = 0; w < FEATURE_WORDS; w++) {
+        fi = &feature_word_info[w];
+        entry = &tdx_cpuid_lookup[w];
+
+        if (fi->type != CPUID_FEATURE_WORD) {
+            continue;
+        }
+
+        config = tdx_cap_cpuid_config(fi->cpuid.eax,
+                                      fi->cpuid.needs_ecx ? fi->cpuid.ecx : ~0u,
+                                      fi->cpuid.reg);
+
+        entry->tdx_fixed0 &= ~config;
+        entry->tdx_fixed1 &= ~config;
+    }
+}
+
 int tdx_kvm_init(MachineState *ms, Error **errp)
 {
     TdxGuest *tdx = (TdxGuest *)object_dynamic_cast(OBJECT(ms->cgs),
@@ -378,6 +406,8 @@ int tdx_kvm_init(MachineState *ms, Error **errp)
         get_tdx_capabilities();
     }
 
+    update_tdx_cpuid_lookup_by_tdx_caps();
+
     tdx_guest = tdx;
 
     return 0;
Signature · 1 line
2.34.1
author
#25 lore raw

PATCH v2 10/58 i386/tdx: Integrate tdx_caps->xfam_fixed0/1 into tdx_cpuid_lookup

In reply to Xiaoyao Li

KVM requires userspace to pass XFAM configuration via CPUID 0xD leaves.

Convert tdx_caps->xfam_fixed0/1 into corresponding tdx_cpuid_lookup[].tdx_fixed0/1 field of CPUID 0xD leaves. Thus the requirement can be applied naturally.

---
 target/i386/cpu.c     |  3 ---
 target/i386/cpu.h     |  3 +++
 target/i386/kvm/tdx.c | 24 ++++++++++++++++++++++++
 3 files changed, 27 insertions(+), 3 deletions(-)
target/i386/cpu.c +0 -3
index 97ad229d8ba3..b8850a02539a 100644
@@ -1568,9 +1568,6 @@ static const X86RegisterInfo32 x86_reg_info_32[CPU_NB_REGS32] = {
 };
 #undef REGISTER
 
-/* CPUID feature bits available in XSS */
-#define CPUID_XSTATE_XSS_MASK    (XSTATE_ARCH_LBR_MASK)
-
 ExtSaveArea x86_ext_save_areas[XSAVE_STATE_AREA_COUNT] = {
     [XSTATE_FP_BIT] = {
         /* x87 FP state component is always enabled if XSAVE is supported */
target/i386/cpu.h +3 -0
index c93dcd274531..d4a7f9e3f54c 100644
@@ -588,6 +588,9 @@ typedef enum X86Seg {
                                  XSTATE_Hi16_ZMM_MASK | XSTATE_PKRU_MASK | \
                                  XSTATE_XTILE_CFG_MASK | XSTATE_XTILE_DATA_MASK)
 
+/* CPUID feature bits available in XSS */
+#define CPUID_XSTATE_XSS_MASK    (XSTATE_ARCH_LBR_MASK)
+
 /* CPUID feature words */
 typedef enum FeatureWord {
     FEAT_1_EDX,         /* CPUID[1].EDX */
target/i386/kvm/tdx.c +24 -0
index 4518c79aecc8..19aed03f12c6 100644
@@ -395,6 +395,30 @@ static void update_tdx_cpuid_lookup_by_tdx_caps(void)
         entry->tdx_fixed0 &= ~config;
         entry->tdx_fixed1 &= ~config;
     }
+
+    /*
+     * Because KVM gets XFAM settings via CPUID leaves 0xD,  map
+     * tdx_caps->xfam_fixed{0, 1} into tdx_cpuid_lookup[].tdx_fixed{0, 1}.
+     *
+     * Then the enforment applies in tdx_get_configurable_cpuid() naturally.
+     */
+    tdx_cpuid_lookup[FEAT_XSAVE_XCR0_LO].tdx_fixed0 =
+            (uint32_t)~tdx_caps->xfam_fixed0 & CPUID_XSTATE_XCR0_MASK;
+    tdx_cpuid_lookup[FEAT_XSAVE_XCR0_LO].tdx_fixed1 =
+            (uint32_t)tdx_caps->xfam_fixed1 & CPUID_XSTATE_XCR0_MASK;
+    tdx_cpuid_lookup[FEAT_XSAVE_XCR0_HI].tdx_fixed0 =
+            (~tdx_caps->xfam_fixed0 & CPUID_XSTATE_XCR0_MASK) >> 32;
+    tdx_cpuid_lookup[FEAT_XSAVE_XCR0_HI].tdx_fixed1 =
+            (tdx_caps->xfam_fixed1 & CPUID_XSTATE_XCR0_MASK) >> 32;
+
+    tdx_cpuid_lookup[FEAT_XSAVE_XSS_LO].tdx_fixed0 =
+            (uint32_t)~tdx_caps->xfam_fixed0 & CPUID_XSTATE_XSS_MASK;
+    tdx_cpuid_lookup[FEAT_XSAVE_XSS_LO].tdx_fixed1 =
+            (uint32_t)tdx_caps->xfam_fixed1 & CPUID_XSTATE_XSS_MASK;
+    tdx_cpuid_lookup[FEAT_XSAVE_XSS_HI].tdx_fixed0 =
+            (~tdx_caps->xfam_fixed0 & CPUID_XSTATE_XSS_MASK) >> 32;
+    tdx_cpuid_lookup[FEAT_XSAVE_XSS_HI].tdx_fixed1 =
+            (tdx_caps->xfam_fixed1 & CPUID_XSTATE_XSS_MASK) >> 32;
 }
 
 int tdx_kvm_init(MachineState *ms, Error **errp)
Signature · 1 line
2.34.1
author
#26 lore raw

PATCH v2 11/58 i386/tdx: Integrate tdx_caps->attrs_fixed0/1 to tdx_cpuid_lookup

In reply to Xiaoyao Li

Some bits in TD attributes have corresponding CPUID feature bits. Reflect the fixed0/1 restriction on TD attributes to their corresponding CPUID bits in tdx_cpuid_lookup[] as well.

---
 target/i386/cpu-internal.h |  9 +++++++++
 target/i386/cpu.c          |  9 ---------
 target/i386/cpu.h          |  2 ++
 target/i386/kvm/tdx.c      | 21 +++++++++++++++++++++
 4 files changed, 32 insertions(+), 9 deletions(-)
target/i386/cpu-internal.h +9 -0
index 9baac5c0b450..e980f6e3147f 100644
@@ -20,6 +20,15 @@
 #ifndef I386_CPU_INTERNAL_H
 #define I386_CPU_INTERNAL_H
 
+typedef struct FeatureMask {
+    FeatureWord index;
+    uint64_t mask;
+} FeatureMask;
+
+typedef struct FeatureDep {
+    FeatureMask from, to;
+} FeatureDep;
+
 typedef enum FeatureWordType {
    CPUID_FEATURE_WORD,
    MSR_FEATURE_WORD,
target/i386/cpu.c +0 -9
index b8850a02539a..a529c07e749c 100644
@@ -1439,15 +1439,6 @@ FeatureWordInfo feature_word_info[FEATURE_WORDS] = {
     },
 };
 
-typedef struct FeatureMask {
-    FeatureWord index;
-    uint64_t mask;
-} FeatureMask;
-
-typedef struct FeatureDep {
-    FeatureMask from, to;
-} FeatureDep;
-
 static FeatureDep feature_dependencies[] = {
     {
         .from = { FEAT_7_0_EDX,             CPUID_7_0_EDX_ARCH_CAPABILITIES },
target/i386/cpu.h +2 -0
index d4a7f9e3f54c..f51e93e21f73 100644
@@ -876,6 +876,8 @@ uint64_t x86_cpu_get_supported_feature_word(FeatureWord w,
 #define CPUID_7_0_ECX_MAWAU             (31U << 17)
 /* Read Processor ID */
 #define CPUID_7_0_ECX_RDPID             (1U << 22)
+/* KeyLocker */
+#define CPUID_7_0_ECX_KeyLocker         (1U << 23)
 /* Bus Lock Debug Exception */
 #define CPUID_7_0_ECX_BUS_LOCK_DETECT   (1U << 24)
 /* Cache Line Demote Instruction */
target/i386/kvm/tdx.c +21 -0
index 19aed03f12c6..29f50fb9529e 100644
@@ -32,6 +32,13 @@
                                      (1U << KVM_FEATURE_PV_SCHED_YIELD) | \
                                      (1U << KVM_FEATURE_MSI_EXT_DEST_ID))
 
+#define TDX_ATTRIBUTES_MAX_BITS      64
+
+static FeatureMask tdx_attrs_ctrl_fields[TDX_ATTRIBUTES_MAX_BITS] = {
+    [30] = { .index = FEAT_7_0_ECX, .mask = CPUID_7_0_ECX_PKS },
+    [31] = { .index = FEAT_7_0_ECX, .mask = CPUID_7_0_ECX_KeyLocker},
+};
+
 typedef struct KvmTdxCpuidLookup {
     uint32_t tdx_fixed0;
     uint32_t tdx_fixed1;
@@ -375,6 +382,8 @@ static void update_tdx_cpuid_lookup_by_tdx_caps(void)
     FeatureWordInfo *fi;
     uint32_t config;
     FeatureWord w;
+    FeatureMask *fm;
+    int i;
 
     /*
      * Patch tdx_fixed0/1 by tdx_caps that what TDX module reports as
@@ -396,6 +405,18 @@ static void update_tdx_cpuid_lookup_by_tdx_caps(void)
         entry->tdx_fixed1 &= ~config;
     }
 
+    for (i = 0; i < ARRAY_SIZE(tdx_attrs_ctrl_fields); i++) {
+        fm = &tdx_attrs_ctrl_fields[i];
+
+        if (tdx_caps->attrs_fixed0 & (1ULL << i)) {
+            tdx_cpuid_lookup[fm->index].tdx_fixed0 |= fm->mask;
+        }
+
+        if (tdx_caps->attrs_fixed1 & (1ULL << i)) {
+            tdx_cpuid_lookup[fm->index].tdx_fixed1 |= fm->mask;
+        }
+    }
+
     /*
      * Because KVM gets XFAM settings via CPUID leaves 0xD,  map
      * tdx_caps->xfam_fixed{0, 1} into tdx_cpuid_lookup[].tdx_fixed{0, 1}.
Signature · 1 line
2.34.1
author
#27 lore raw

PATCH v2 12/58 i386/kvm: Move architectural CPUID leaf generation to separate helper

In reply to Xiaoyao Li

From: Sean Christopherson <[email protected]>

Move the architectural (for lack of a better term) CPUID leaf generation to a separate helper so that the generation code can be reused by TDX, which needs to generate a canonical VM-scoped configuration.

---
 target/i386/kvm/kvm.c      | 454 +++++++++++++++++++------------------
 target/i386/kvm/kvm_i386.h |   3 +
 2 files changed, 235 insertions(+), 222 deletions(-)
target/i386/kvm/kvm.c +232 -222
index 46a455a1e331..9ee41fffc445 100644
@@ -1799,6 +1799,236 @@ static void kvm_init_nested_state(CPUX86State *env)
     }
 }
 
+uint32_t kvm_x86_arch_cpuid(CPUX86State *env, struct kvm_cpuid_entry2 *entries,
+                            uint32_t cpuid_i)
+{
+    uint32_t limit, i, j;
+    uint32_t unused;
+    struct kvm_cpuid_entry2 *c;
+
+    cpu_x86_cpuid(env, 0, 0, &limit, &unused, &unused, &unused);
+
+    for (i = 0; i <= limit; i++) {
+        if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
+            fprintf(stderr, "unsupported level value: 0x%x\n", limit);
+            abort();
+        }
+        c = &entries[cpuid_i++];
+
+        switch (i) {
+        case 2: {
+            /* Keep reading function 2 till all the input is received */
+            int times;
+
+            c->function = i;
+            c->flags = KVM_CPUID_FLAG_STATEFUL_FUNC |
+                       KVM_CPUID_FLAG_STATE_READ_NEXT;
+            cpu_x86_cpuid(env, i, 0, &c->eax, &c->ebx, &c->ecx, &c->edx);
+            times = c->eax & 0xff;
+
+            for (j = 1; j < times; ++j) {
+                if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
+                    fprintf(stderr, "cpuid_data is full, no space for "
+                            "cpuid(eax:2):eax & 0xf = 0x%x\n", times);
+                    abort();
+                }
+                c = &entries[cpuid_i++];
+                c->function = i;
+                c->flags = KVM_CPUID_FLAG_STATEFUL_FUNC;
+                cpu_x86_cpuid(env, i, 0, &c->eax, &c->ebx, &c->ecx, &c->edx);
+            }
+            break;
+        }
+        case 0x1f:
+            if (env->nr_dies < 2) {
+                cpuid_i--;
+                break;
+            }
+            /* fallthrough */
+        case 4:
+        case 0xb:
+        case 0xd:
+            for (j = 0; ; j++) {
+                if (i == 0xd && j == 64) {
+                    break;
+                }
+
+                c->function = i;
+                c->flags = KVM_CPUID_FLAG_SIGNIFCANT_INDEX;
+                c->index = j;
+                cpu_x86_cpuid(env, i, j, &c->eax, &c->ebx, &c->ecx, &c->edx);
+
+                if (i == 4 && c->eax == 0) {
+                    break;
+                }
+                if (i == 0xb && !(c->ecx & 0xff00)) {
+                    break;
+                }
+                if (i == 0x1f && !(c->ecx & 0xff00)) {
+                    break;
+                }
+                if (i == 0xd && c->eax == 0) {
+                    continue;
+                }
+                if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
+                    fprintf(stderr, "cpuid_data is full, no space for "
+                            "cpuid(eax:0x%x,ecx:0x%x)\n", i, j);
+                    abort();
+                }
+                c = &entries[cpuid_i++];
+            }
+            break;
+        case 0x7:
+        case 0x12:
+            for (j = 0; ; j++) {
+                c->function = i;
+                c->flags = KVM_CPUID_FLAG_SIGNIFCANT_INDEX;
+                c->index = j;
+                cpu_x86_cpuid(env, i, j, &c->eax, &c->ebx, &c->ecx, &c->edx);
+
+                if (j > 1 && (c->eax & 0xf) != 1) {
+                    break;
+                }
+
+                if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
+                    fprintf(stderr, "cpuid_data is full, no space for "
+                                "cpuid(eax:0x12,ecx:0x%x)\n", j);
+                    abort();
+                }
+                c = &entries[cpuid_i++];
+            }
+            break;
+        case 0x14:
+        case 0x1d:
+        case 0x1e: {
+            uint32_t times;
+
+            c->function = i;
+            c->index = 0;
+            c->flags = KVM_CPUID_FLAG_SIGNIFCANT_INDEX;
+            cpu_x86_cpuid(env, i, 0, &c->eax, &c->ebx, &c->ecx, &c->edx);
+            times = c->eax;
+
+            for (j = 1; j <= times; ++j) {
+                if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
+                    fprintf(stderr, "cpuid_data is full, no space for "
+                                "cpuid(eax:0x%x,ecx:0x%x)\n", i, j);
+                    abort();
+                }
+                c = &entries[cpuid_i++];
+                c->function = i;
+                c->index = j;
+                c->flags = KVM_CPUID_FLAG_SIGNIFCANT_INDEX;
+                cpu_x86_cpuid(env, i, j, &c->eax, &c->ebx, &c->ecx, &c->edx);
+            }
+            break;
+        }
+        default:
+            c->function = i;
+            c->flags = 0;
+            cpu_x86_cpuid(env, i, 0, &c->eax, &c->ebx, &c->ecx, &c->edx);
+            if (!c->eax && !c->ebx && !c->ecx && !c->edx) {
+                /*
+                 * KVM already returns all zeroes if a CPUID entry is missing,
+                 * so we can omit it and avoid hitting KVM's 80-entry limit.
+                 */
+                cpuid_i--;
+            }
+            break;
+        }
+    }
+
+    if (limit >= 0x0a) {
+        uint32_t eax, edx;
+
+        cpu_x86_cpuid(env, 0x0a, 0, &eax, &unused, &unused, &edx);
+
+        has_architectural_pmu_version = eax & 0xff;
+        if (has_architectural_pmu_version > 0) {
+            num_architectural_pmu_gp_counters = (eax & 0xff00) >> 8;
+
+            /* Shouldn't be more than 32, since that's the number of bits
+             * available in EBX to tell us _which_ counters are available.
+             * Play it safe.
+             */
+            if (num_architectural_pmu_gp_counters > MAX_GP_COUNTERS) {
+                num_architectural_pmu_gp_counters = MAX_GP_COUNTERS;
+            }
+
+            if (has_architectural_pmu_version > 1) {
+                num_architectural_pmu_fixed_counters = edx & 0x1f;
+
+                if (num_architectural_pmu_fixed_counters > MAX_FIXED_COUNTERS) {
+                    num_architectural_pmu_fixed_counters = MAX_FIXED_COUNTERS;
+                }
+            }
+        }
+    }
+
+    cpu_x86_cpuid(env, 0x80000000, 0, &limit, &unused, &unused, &unused);
+
+    for (i = 0x80000000; i <= limit; i++) {
+        if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
+            fprintf(stderr, "unsupported xlevel value: 0x%x\n", limit);
+            abort();
+        }
+        c = &entries[cpuid_i++];
+
+        switch (i) {
+        case 0x8000001d:
+            /* Query for all AMD cache information leaves */
+            for (j = 0; ; j++) {
+                c->function = i;
+                c->flags = KVM_CPUID_FLAG_SIGNIFCANT_INDEX;
+                c->index = j;
+                cpu_x86_cpuid(env, i, j, &c->eax, &c->ebx, &c->ecx, &c->edx);
+
+                if (c->eax == 0) {
+                    break;
+                }
+                if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
+                    fprintf(stderr, "cpuid_data is full, no space for "
+                            "cpuid(eax:0x%x,ecx:0x%x)\n", i, j);
+                    abort();
+                }
+                c = &entries[cpuid_i++];
+            }
+            break;
+        default:
+            c->function = i;
+            c->flags = 0;
+            cpu_x86_cpuid(env, i, 0, &c->eax, &c->ebx, &c->ecx, &c->edx);
+            if (!c->eax && !c->ebx && !c->ecx && !c->edx) {
+                /*
+                 * KVM already returns all zeroes if a CPUID entry is missing,
+                 * so we can omit it and avoid hitting KVM's 80-entry limit.
+                 */
+                cpuid_i--;
+            }
+            break;
+        }
+    }
+
+    /* Call Centaur's CPUID instructions they are supported. */
+    if (env->cpuid_xlevel2 > 0) {
+        cpu_x86_cpuid(env, 0xC0000000, 0, &limit, &unused, &unused, &unused);
+
+        for (i = 0xC0000000; i <= limit; i++) {
+            if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
+                fprintf(stderr, "unsupported xlevel2 value: 0x%x\n", limit);
+                abort();
+            }
+            c = &entries[cpuid_i++];
+
+            c->function = i;
+            c->flags = 0;
+            cpu_x86_cpuid(env, i, 0, &c->eax, &c->ebx, &c->ecx, &c->edx);
+        }
+    }
+
+    return cpuid_i;
+}
+
 int kvm_arch_init_vcpu(CPUState *cs)
 {
     struct {
@@ -1815,8 +2045,7 @@ int kvm_arch_init_vcpu(CPUState *cs)
 
     X86CPU *cpu = X86_CPU(cs);
     CPUX86State *env = &cpu->env;
-    uint32_t limit, i, j, cpuid_i;
-    uint32_t unused;
+    uint32_t cpuid_i;
     struct kvm_cpuid_entry2 *c;
     uint32_t signature[3];
     int kvm_base = KVM_CPUID_SIGNATURE;
@@ -1965,8 +2194,6 @@ int kvm_arch_init_vcpu(CPUState *cs)
         c->edx = env->features[FEAT_KVM_HINTS];
     }
 
-    cpu_x86_cpuid(env, 0, 0, &limit, &unused, &unused, &unused);
-
     if (cpu->kvm_pv_enforce_cpuid) {
         r = kvm_vcpu_enable_cap(cs, KVM_CAP_ENFORCE_PV_FEATURE_CPUID, 0, 1);
         if (r < 0) {
@@ -1977,224 +2204,7 @@ int kvm_arch_init_vcpu(CPUState *cs)
         }
     }
 
-    for (i = 0; i <= limit; i++) {
-        if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
-            fprintf(stderr, "unsupported level value: 0x%x\n", limit);
-            abort();
-        }
-        c = &cpuid_data.entries[cpuid_i++];
-
-        switch (i) {
-        case 2: {
-            /* Keep reading function 2 till all the input is received */
-            int times;
-
-            c->function = i;
-            c->flags = KVM_CPUID_FLAG_STATEFUL_FUNC |
-                       KVM_CPUID_FLAG_STATE_READ_NEXT;
-            cpu_x86_cpuid(env, i, 0, &c->eax, &c->ebx, &c->ecx, &c->edx);
-            times = c->eax & 0xff;
-
-            for (j = 1; j < times; ++j) {
-                if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
-                    fprintf(stderr, "cpuid_data is full, no space for "
-                            "cpuid(eax:2):eax & 0xf = 0x%x\n", times);
-                    abort();
-                }
-                c = &cpuid_data.entries[cpuid_i++];
-                c->function = i;
-                c->flags = KVM_CPUID_FLAG_STATEFUL_FUNC;
-                cpu_x86_cpuid(env, i, 0, &c->eax, &c->ebx, &c->ecx, &c->edx);
-            }
-            break;
-        }
-        case 0x1f:
-            if (env->nr_dies < 2) {
-                cpuid_i--;
-                break;
-            }
-            /* fallthrough */
-        case 4:
-        case 0xb:
-        case 0xd:
-            for (j = 0; ; j++) {
-                if (i == 0xd && j == 64) {
-                    break;
-                }
-
-                c->function = i;
-                c->flags = KVM_CPUID_FLAG_SIGNIFCANT_INDEX;
-                c->index = j;
-                cpu_x86_cpuid(env, i, j, &c->eax, &c->ebx, &c->ecx, &c->edx);
-
-                if (i == 4 && c->eax == 0) {
-                    break;
-                }
-                if (i == 0xb && !(c->ecx & 0xff00)) {
-                    break;
-                }
-                if (i == 0x1f && !(c->ecx & 0xff00)) {
-                    break;
-                }
-                if (i == 0xd && c->eax == 0) {
-                    continue;
-                }
-                if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
-                    fprintf(stderr, "cpuid_data is full, no space for "
-                            "cpuid(eax:0x%x,ecx:0x%x)\n", i, j);
-                    abort();
-                }
-                c = &cpuid_data.entries[cpuid_i++];
-            }
-            break;
-        case 0x7:
-        case 0x12:
-            for (j = 0; ; j++) {
-                c->function = i;
-                c->flags = KVM_CPUID_FLAG_SIGNIFCANT_INDEX;
-                c->index = j;
-                cpu_x86_cpuid(env, i, j, &c->eax, &c->ebx, &c->ecx, &c->edx);
-
-                if (j > 1 && (c->eax & 0xf) != 1) {
-                    break;
-                }
-
-                if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
-                    fprintf(stderr, "cpuid_data is full, no space for "
-                                "cpuid(eax:0x12,ecx:0x%x)\n", j);
-                    abort();
-                }
-                c = &cpuid_data.entries[cpuid_i++];
-            }
-            break;
-        case 0x14:
-        case 0x1d:
-        case 0x1e: {
-            uint32_t times;
-
-            c->function = i;
-            c->index = 0;
-            c->flags = KVM_CPUID_FLAG_SIGNIFCANT_INDEX;
-            cpu_x86_cpuid(env, i, 0, &c->eax, &c->ebx, &c->ecx, &c->edx);
-            times = c->eax;
-
-            for (j = 1; j <= times; ++j) {
-                if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
-                    fprintf(stderr, "cpuid_data is full, no space for "
-                                "cpuid(eax:0x%x,ecx:0x%x)\n", i, j);
-                    abort();
-                }
-                c = &cpuid_data.entries[cpuid_i++];
-                c->function = i;
-                c->index = j;
-                c->flags = KVM_CPUID_FLAG_SIGNIFCANT_INDEX;
-                cpu_x86_cpuid(env, i, j, &c->eax, &c->ebx, &c->ecx, &c->edx);
-            }
-            break;
-        }
-        default:
-            c->function = i;
-            c->flags = 0;
-            cpu_x86_cpuid(env, i, 0, &c->eax, &c->ebx, &c->ecx, &c->edx);
-            if (!c->eax && !c->ebx && !c->ecx && !c->edx) {
-                /*
-                 * KVM already returns all zeroes if a CPUID entry is missing,
-                 * so we can omit it and avoid hitting KVM's 80-entry limit.
-                 */
-                cpuid_i--;
-            }
-            break;
-        }
-    }
-
-    if (limit >= 0x0a) {
-        uint32_t eax, edx;
-
-        cpu_x86_cpuid(env, 0x0a, 0, &eax, &unused, &unused, &edx);
-
-        has_architectural_pmu_version = eax & 0xff;
-        if (has_architectural_pmu_version > 0) {
-            num_architectural_pmu_gp_counters = (eax & 0xff00) >> 8;
-
-            /* Shouldn't be more than 32, since that's the number of bits
-             * available in EBX to tell us _which_ counters are available.
-             * Play it safe.
-             */
-            if (num_architectural_pmu_gp_counters > MAX_GP_COUNTERS) {
-                num_architectural_pmu_gp_counters = MAX_GP_COUNTERS;
-            }
-
-            if (has_architectural_pmu_version > 1) {
-                num_architectural_pmu_fixed_counters = edx & 0x1f;
-
-                if (num_architectural_pmu_fixed_counters > MAX_FIXED_COUNTERS) {
-                    num_architectural_pmu_fixed_counters = MAX_FIXED_COUNTERS;
-                }
-            }
-        }
-    }
-
-    cpu_x86_cpuid(env, 0x80000000, 0, &limit, &unused, &unused, &unused);
-
-    for (i = 0x80000000; i <= limit; i++) {
-        if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
-            fprintf(stderr, "unsupported xlevel value: 0x%x\n", limit);
-            abort();
-        }
-        c = &cpuid_data.entries[cpuid_i++];
-
-        switch (i) {
-        case 0x8000001d:
-            /* Query for all AMD cache information leaves */
-            for (j = 0; ; j++) {
-                c->function = i;
-                c->flags = KVM_CPUID_FLAG_SIGNIFCANT_INDEX;
-                c->index = j;
-                cpu_x86_cpuid(env, i, j, &c->eax, &c->ebx, &c->ecx, &c->edx);
-
-                if (c->eax == 0) {
-                    break;
-                }
-                if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
-                    fprintf(stderr, "cpuid_data is full, no space for "
-                            "cpuid(eax:0x%x,ecx:0x%x)\n", i, j);
-                    abort();
-                }
-                c = &cpuid_data.entries[cpuid_i++];
-            }
-            break;
-        default:
-            c->function = i;
-            c->flags = 0;
-            cpu_x86_cpuid(env, i, 0, &c->eax, &c->ebx, &c->ecx, &c->edx);
-            if (!c->eax && !c->ebx && !c->ecx && !c->edx) {
-                /*
-                 * KVM already returns all zeroes if a CPUID entry is missing,
-                 * so we can omit it and avoid hitting KVM's 80-entry limit.
-                 */
-                cpuid_i--;
-            }
-            break;
-        }
-    }
-
-    /* Call Centaur's CPUID instructions they are supported. */
-    if (env->cpuid_xlevel2 > 0) {
-        cpu_x86_cpuid(env, 0xC0000000, 0, &limit, &unused, &unused, &unused);
-
-        for (i = 0xC0000000; i <= limit; i++) {
-            if (cpuid_i == KVM_MAX_CPUID_ENTRIES) {
-                fprintf(stderr, "unsupported xlevel2 value: 0x%x\n", limit);
-                abort();
-            }
-            c = &cpuid_data.entries[cpuid_i++];
-
-            c->function = i;
-            c->flags = 0;
-            cpu_x86_cpuid(env, i, 0, &c->eax, &c->ebx, &c->ecx, &c->edx);
-        }
-    }
-
+    cpuid_i = kvm_x86_arch_cpuid(env, cpuid_data.entries, cpuid_i);
     cpuid_data.cpuid.nent = cpuid_i;
 
     if (((env->cpuid_version >> 8)&0xF) >= 6
target/i386/kvm/kvm_i386.h +3 -0
index 769eadbba56c..fd7e76fcf847 100644
@@ -26,6 +26,9 @@
 #define kvm_ioapic_in_kernel() \
     (kvm_irqchip_in_kernel() && !kvm_irqchip_is_split())
 
+uint32_t kvm_x86_arch_cpuid(CPUX86State *env, struct kvm_cpuid_entry2 *entries,
+                            uint32_t cpuid_i);
+
 #else
 
 #define kvm_pit_in_kernel()      0
Signature · 1 line
2.34.1
author
#28 lore raw

PATCH v2 13/58 kvm: Introduce kvm_arch_pre_create_vcpu()

In reply to Xiaoyao Li

Introduce kvm_arch_pre_create_vcpu(), to perform arch-dependent work prior to create any vcpu. This is for i386 TDX because it needs call TDX_INIT_VM before creating any vcpu.

---
 accel/kvm/kvm-all.c  | 12 ++++++++++++
 include/sysemu/kvm.h |  1 +
 2 files changed, 13 insertions(+)
accel/kvm/kvm-all.c +12 -0
index c9f3aab5e587..5071af917ae0 100644
@@ -422,6 +422,11 @@ static int kvm_get_vcpu(KVMState *s, unsigned long vcpu_id)
     return kvm_vm_ioctl(s, KVM_CREATE_VCPU, (void *)vcpu_id);
 }
 
+int __attribute__ ((weak)) kvm_arch_pre_create_vcpu(CPUState *cpu)
+{
+    return 0;
+}
+
 int kvm_init_vcpu(CPUState *cpu, Error **errp)
 {
     KVMState *s = kvm_state;
@@ -430,6 +435,13 @@ int kvm_init_vcpu(CPUState *cpu, Error **errp)
 
     trace_kvm_init_vcpu(cpu->cpu_index, kvm_arch_vcpu_id(cpu));
 
+    ret = kvm_arch_pre_create_vcpu(cpu);
+    if (ret < 0) {
+        error_setg_errno(errp, -ret, "%s: kvm_arch_pre_create_vcpu() failed",
+                        __func__);
+        goto err;
+    }
+
     ret = kvm_get_vcpu(s, kvm_arch_vcpu_id(cpu));
     if (ret < 0) {
         error_setg_errno(errp, -ret, "kvm_init_vcpu: kvm_get_vcpu failed (%lu)",
include/sysemu/kvm.h +1 -0
index 49c896d8a512..d89ec87072d7 100644
@@ -371,6 +371,7 @@ int kvm_arch_put_registers(CPUState *cpu, int level);
 
 int kvm_arch_init(MachineState *ms, KVMState *s);
 
+int kvm_arch_pre_create_vcpu(CPUState *cpu);
 int kvm_arch_init_vcpu(CPUState *cpu);
 int kvm_arch_destroy_vcpu(CPUState *cpu);
 
Signature · 1 line
2.34.1
#29 lore raw

Re: [PATCH v2 13/58] kvm: Introduce kvm_arch_pre_create_vcpu()

Quoted 28 lines · Xiaoyao Li

Introduce kvm_arch_pre_create_vcpu(), to perform arch-dependent work prior to create any vcpu. This is for i386 TDX because it needs call TDX_INIT_VM before creating any vcpu.

---
 accel/kvm/kvm-all.c  | 12 ++++++++++++
 include/sysemu/kvm.h |  1 +
 2 files changed, 13 insertions(+)
accel/kvm/kvm-all.c +5 -0
index c9f3aab5e587..5071af917ae0 100644
@@ -422,6 +422,11 @@ static int kvm_get_vcpu(KVMState *s, unsigned long vcpu_id)
     return kvm_vm_ioctl(s, KVM_CREATE_VCPU, (void *)vcpu_id);
 }
 
+int __attribute__ ((weak)) kvm_arch_pre_create_vcpu(CPUState *cpu)
+{
+    return 0;
+}
+
 int kvm_init_vcpu(CPUState *cpu, Error **errp)
 {
     KVMState *s = kvm_state;

accel/kvm/kvm-all.c int kvm_init_vcpu(CPUState *cpu, Error **errp) · line 441

Quoted 7 lines
accel/kvm/kvm-all.c +4 -0
@@ -430,6 +435,13 @@ int kvm_init_vcpu(CPUState *cpu, Error **errp)
 
     trace_kvm_init_vcpu(cpu->cpu_index, kvm_arch_vcpu_id(cpu));
 
+    ret = kvm_arch_pre_create_vcpu(cpu);
+    if (ret < 0) {
+        error_setg_errno(errp, -ret, "%s: kvm_arch_pre_create_vcpu() failed",
+                        __func__);

Don't report generic error messages here, when kvm_arch_pre_create_vcpu can provide a better error - pass the 'errp' into the kvm_arch_pre_create_vcpu method.

Quoted 21 lines
accel/kvm/kvm-all.c +3 -0
+        goto err;
+    }
+
     ret = kvm_get_vcpu(s, kvm_arch_vcpu_id(cpu));
     if (ret < 0) {
         error_setg_errno(errp, -ret, "kvm_init_vcpu: kvm_get_vcpu failed (%lu)",
include/sysemu/kvm.h +1 -0
index 49c896d8a512..d89ec87072d7 100644
@@ -371,6 +371,7 @@ int kvm_arch_put_registers(CPUState *cpu, int level);
 
 int kvm_arch_init(MachineState *ms, KVMState *s);
 
+int kvm_arch_pre_create_vcpu(CPUState *cpu);
 int kvm_arch_init_vcpu(CPUState *cpu);
 int kvm_arch_destroy_vcpu(CPUState *cpu);
 
Signature · 1 line
2.34.1

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#30 lore raw

Re: [PATCH v2 13/58] kvm: Introduce kvm_arch_pre_create_vcpu()

In reply to Xiaoyao Li
Quoted 16 lines · Xiaoyao Li

Introduce kvm_arch_pre_create_vcpu(), to perform arch-dependent work prior to create any vcpu. This is for i386 TDX because it needs call TDX_INIT_VM before creating any vcpu.

---
  accel/kvm/kvm-all.c  | 12 ++++++++++++
  include/sysemu/kvm.h |  1 +
  2 files changed, 13 insertions(+)
accel/kvm/kvm-all.c +0 -0
index c9f3aab5e587..5071af917ae0 100644

accel/kvm/kvm-all.c static int kvm_get_vcpu(KVMState *s, unsigned long vcpu_id) · line 428

Quoted 7 lines
accel/kvm/kvm-all.c +4 -0
@@ -422,6 +422,11 @@ static int kvm_get_vcpu(KVMState *s, unsigned long vcpu_id)
      return kvm_vm_ioctl(s, KVM_CREATE_VCPU, (void *)vcpu_id);
  }
 
+int __attribute__ ((weak)) kvm_arch_pre_create_vcpu(CPUState *cpu)
+{
+    return 0;
+}

kvm_arch_init_vcpu() is implemented for each arch. Why not use the same approach here?

Quoted 29 lines
accel/kvm/kvm-all.c +7 -0
  int kvm_init_vcpu(CPUState *cpu, Error **errp)
  {
      KVMState *s = kvm_state;
@@ -430,6 +435,13 @@ int kvm_init_vcpu(CPUState *cpu, Error **errp)
 
      trace_kvm_init_vcpu(cpu->cpu_index, kvm_arch_vcpu_id(cpu));
 
+    ret = kvm_arch_pre_create_vcpu(cpu);
+    if (ret < 0) {
+        error_setg_errno(errp, -ret, "%s: kvm_arch_pre_create_vcpu() failed",
+                        __func__);
+        goto err;
+    }
+
      ret = kvm_get_vcpu(s, kvm_arch_vcpu_id(cpu));
      if (ret < 0) {
          error_setg_errno(errp, -ret, "kvm_init_vcpu: kvm_get_vcpu failed (%lu)",
include/sysemu/kvm.h +1 -0
index 49c896d8a512..d89ec87072d7 100644
@@ -371,6 +371,7 @@ int kvm_arch_put_registers(CPUState *cpu, int level);
 
  int kvm_arch_init(MachineState *ms, KVMState *s);
 
+int kvm_arch_pre_create_vcpu(CPUState *cpu);
  int kvm_arch_init_vcpu(CPUState *cpu);
  int kvm_arch_destroy_vcpu(CPUState *cpu);
 
author
#31 lore raw

Re: [PATCH v2 13/58] kvm: Introduce kvm_arch_pre_create_vcpu()

Quoted 12 lines · Philippe Mathieu-Daudé
Quoted 11 lines · Xiaoyao Li

Introduce kvm_arch_pre_create_vcpu(), to perform arch-dependent work prior to create any vcpu. This is for i386 TDX because it needs call TDX_INIT_VM before creating any vcpu.

---
  accel/kvm/kvm-all.c  | 12 ++++++++++++
  include/sysemu/kvm.h |  1 +

  2 files changed, 13 insertions(+)
Quoted 15 lines

accel/kvm/kvm-all.c static int kvm_get_vcpu(KVMState *s, unsigned · line 428

Quoted 12 lines
accel/kvm/kvm-all.c +4 -0
index c9f3aab5e587..5071af917ae0 100644
@@ -422,6 +422,11 @@ static int kvm_get_vcpu(KVMState *s, unsigned 
 long vcpu_id)
       return kvm_vm_ioctl(s, KVM_CREATE_VCPU, (void *)vcpu_id);
   }
+int __attribute__ ((weak)) kvm_arch_pre_create_vcpu(CPUState *cpu)
+{
+    return 0;
+}

kvm_arch_init_vcpu() is implemented for each arch. Why not use the same approach here?

Because only x86 needs it currently, for TDX. Other arches don't require an implementation.

If don't provide the _weak_ function, it needs to implement the empty function (justing return 0) in all the other arches just as the placeholder. If QEMU community prefers this approach, I can change to it in next version.

Quoted 26 lines
Quoted 25 lines
accel/kvm/kvm-all.c +7 -0
   int kvm_init_vcpu(CPUState *cpu, Error **errp)
   {
       KVMState *s = kvm_state;
@@ -430,6 +435,13 @@ int kvm_init_vcpu(CPUState *cpu, Error **errp)
       trace_kvm_init_vcpu(cpu->cpu_index, kvm_arch_vcpu_id(cpu));
+    ret = kvm_arch_pre_create_vcpu(cpu);
+    if (ret < 0) {
+        error_setg_errno(errp, -ret, "%s: kvm_arch_pre_create_vcpu() 
 failed",
+                        __func__);
+        goto err;
+    }
+
       ret = kvm_get_vcpu(s, kvm_arch_vcpu_id(cpu));
       if (ret < 0) {
           error_setg_errno(errp, -ret, "kvm_init_vcpu: kvm_get_vcpu failed (%lu)",
include/sysemu/kvm.h +1 -0
index 49c896d8a512..d89ec87072d7 100644
@@ -371,6 +371,7 @@ int kvm_arch_put_registers(CPUState *cpu, int level);
   int kvm_arch_init(MachineState *ms, KVMState *s);
+int kvm_arch_pre_create_vcpu(CPUState *cpu);
   int kvm_arch_init_vcpu(CPUState *cpu);
   int kvm_arch_destroy_vcpu(CPUState *cpu);
#32 lore raw

Re: [PATCH v2 13/58] kvm: Introduce kvm_arch_pre_create_vcpu()

Quoted 29 lines · Xiaoyao Li
Quoted 12 lines · Philippe Mathieu-Daudé
Quoted 11 lines · Xiaoyao Li

Introduce kvm_arch_pre_create_vcpu(), to perform arch-dependent work prior to create any vcpu. This is for i386 TDX because it needs call TDX_INIT_VM before creating any vcpu.

---
  accel/kvm/kvm-all.c  | 12 ++++++++++++
  include/sysemu/kvm.h |  1 +

  2 files changed, 13 insertions(+)
Quoted 15 lines

accel/kvm/kvm-all.c static int kvm_get_vcpu(KVMState *s, unsigned · line 428

Quoted 12 lines
accel/kvm/kvm-all.c +4 -0
index c9f3aab5e587..5071af917ae0 100644
@@ -422,6 +422,11 @@ static int kvm_get_vcpu(KVMState *s, unsigned
 long vcpu_id)
       return kvm_vm_ioctl(s, KVM_CREATE_VCPU, (void *)vcpu_id);
   }
+int __attribute__ ((weak)) kvm_arch_pre_create_vcpu(CPUState *cpu)
+{
+    return 0;
+}

kvm_arch_init_vcpu() is implemented for each arch. Why not use the same approach here?

Quoted 6 lines

Because only x86 needs it currently, for TDX. Other arches don't require an implementation.

If don't provide the _weak_ function, it needs to implement the empty function (justing return 0) in all the other arches just as the placeholder. If QEMU community prefers this approach, I can change to it in next version.

Alternative is to move the hook to x86 specific function, not common kvm function. With my quick grepping, x86_cpus_init() or x86_cpu_realizefn().

Signature · 1 line
Isaku Yamahata <[email protected]>
author
#33 lore raw

PATCH v2 14/58 i386/tdx: Initialize TDX before creating TD vcpus

In reply to Xiaoyao Li

Invoke KVM_TDX_INIT in kvm_arch_pre_create_vcpu() that KVM_TDX_INIT configures global TD configurations, e.g. the canonical CPUID config, and must be executed prior to creating vCPUs.

Use kvm_x86_arch_cpuid() to setup the CPUID settings for TDX VM.

Note, this doesn't address the fact that QEMU may change the CPUID configuration when creating vCPUs, i.e. punts on refactoring QEMU to provide a stable CPUID config prior to kvm_arch_init().

---
 accel/kvm/kvm-all.c        |  9 +++++++-
 target/i386/kvm/kvm.c      |  8 +++++++
 target/i386/kvm/tdx-stub.c |  5 +++++
 target/i386/kvm/tdx.c      | 45 ++++++++++++++++++++++++++++++++++++++
 target/i386/kvm/tdx.h      |  4 ++++
 5 files changed, 70 insertions(+), 1 deletion(-)
accel/kvm/kvm-all.c +8 -1
index 5071af917ae0..fceec7f2a83f 100644
@@ -435,10 +435,17 @@ int kvm_init_vcpu(CPUState *cpu, Error **errp)
 
     trace_kvm_init_vcpu(cpu->cpu_index, kvm_arch_vcpu_id(cpu));
 
+    /*
+     * tdx_pre_create_vcpu() may call cpu_x86_cpuid(). It in turn may call
+     * kvm_vm_ioctl(). Set cpu->kvm_state in advance to avoid NULL pointer
+     * dereference.
+     */
+    cpu->kvm_state = s;
     ret = kvm_arch_pre_create_vcpu(cpu);
     if (ret < 0) {
         error_setg_errno(errp, -ret, "%s: kvm_arch_pre_create_vcpu() failed",
                         __func__);
+        cpu->kvm_state = NULL;
         goto err;
     }
 
@@ -446,11 +453,11 @@ int kvm_init_vcpu(CPUState *cpu, Error **errp)
     if (ret < 0) {
         error_setg_errno(errp, -ret, "kvm_init_vcpu: kvm_get_vcpu failed (%lu)",
                          kvm_arch_vcpu_id(cpu));
+        cpu->kvm_state = NULL;
         goto err;
     }
 
     cpu->kvm_fd = ret;
-    cpu->kvm_state = s;
     cpu->vcpu_dirty = true;
     cpu->dirty_pages = 0;
     cpu->throttle_us_per_full = 0;
target/i386/kvm/kvm.c +8 -0
index 9ee41fffc445..d51067fdc12a 100644
@@ -2331,6 +2331,14 @@ int kvm_arch_init_vcpu(CPUState *cs)
     return r;
 }
 
+int kvm_arch_pre_create_vcpu(CPUState *cpu)
+{
+    if (is_tdx_vm())
+        return tdx_pre_create_vcpu(cpu);
+
+    return 0;
+}
+
 int kvm_arch_destroy_vcpu(CPUState *cs)
 {
     X86CPU *cpu = X86_CPU(cs);
target/i386/kvm/tdx-stub.c +5 -0
index 1d866d5496bf..61f70cc0d1d9 100644
@@ -6,3 +6,8 @@ int tdx_kvm_init(MachineState *ms, Error **errp)
 {
     return -EINVAL;
 }
+
+int tdx_pre_create_vcpu(CPUState *cpu)
+{
+    return -EINVAL;
+}
target/i386/kvm/tdx.c +45 -0

57 lines not on this page · raw message

target/i386/kvm/tdx.h +4 -0
index 06599b65b827..46a24ee8c7cc 100644
@@ -17,6 +17,9 @@ typedef struct TdxGuestClass {
 typedef struct TdxGuest {
     ConfidentialGuestSupport parent_obj;
 
+    QemuMutex lock;
+
+    bool initialized;
     uint64_t attributes;    /* TD attributes */
 } TdxGuest;
 
@@ -29,5 +32,6 @@ bool is_tdx_vm(void);
 int tdx_kvm_init(MachineState *ms, Error **errp);
 void tdx_get_supported_cpuid(uint32_t function, uint32_t index, int reg,
                              uint32_t *ret);
+int tdx_pre_create_vcpu(CPUState *cpu);
 
 #endif /* QEMU_I386_TDX_H */
Signature · 1 line
2.34.1
#34 lore raw

Re: [PATCH v2 14/58] i386/tdx: Initialize TDX before creating TD vcpus

Quoted 60 lines · Xiaoyao Li

Invoke KVM_TDX_INIT in kvm_arch_pre_create_vcpu() that KVM_TDX_INIT configures global TD configurations, e.g. the canonical CPUID config, and must be executed prior to creating vCPUs.

Use kvm_x86_arch_cpuid() to setup the CPUID settings for TDX VM.

Note, this doesn't address the fact that QEMU may change the CPUID configuration when creating vCPUs, i.e. punts on refactoring QEMU to provide a stable CPUID config prior to kvm_arch_init().

---
 accel/kvm/kvm-all.c        |  9 +++++++-
 target/i386/kvm/kvm.c      |  8 +++++++
 target/i386/kvm/tdx-stub.c |  5 +++++
 target/i386/kvm/tdx.c      | 45 ++++++++++++++++++++++++++++++++++++++
 target/i386/kvm/tdx.h      |  4 ++++
 5 files changed, 70 insertions(+), 1 deletion(-)
accel/kvm/kvm-all.c +8 -1
index 5071af917ae0..fceec7f2a83f 100644
@@ -435,10 +435,17 @@ int kvm_init_vcpu(CPUState *cpu, Error **errp)
 
     trace_kvm_init_vcpu(cpu->cpu_index, kvm_arch_vcpu_id(cpu));
 
+    /*
+     * tdx_pre_create_vcpu() may call cpu_x86_cpuid(). It in turn may call
+     * kvm_vm_ioctl(). Set cpu->kvm_state in advance to avoid NULL pointer
+     * dereference.
+     */
+    cpu->kvm_state = s;
     ret = kvm_arch_pre_create_vcpu(cpu);
     if (ret < 0) {
         error_setg_errno(errp, -ret, "%s: kvm_arch_pre_create_vcpu() failed",
                         __func__);
+        cpu->kvm_state = NULL;
         goto err;
     }
 
@@ -446,11 +453,11 @@ int kvm_init_vcpu(CPUState *cpu, Error **errp)
     if (ret < 0) {
         error_setg_errno(errp, -ret, "kvm_init_vcpu: kvm_get_vcpu failed (%lu)",
                          kvm_arch_vcpu_id(cpu));
+        cpu->kvm_state = NULL;
         goto err;
     }
 
     cpu->kvm_fd = ret;
-    cpu->kvm_state = s;
     cpu->vcpu_dirty = true;
     cpu->dirty_pages = 0;
     cpu->throttle_us_per_full = 0;
target/i386/kvm/kvm.c +0 -0
index 9ee41fffc445..d51067fdc12a 100644

target/i386/kvm/kvm.c int kvm_arch_init_vcpu(CPUState *cs) · line 2337

Quoted 7 lines
target/i386/kvm/kvm.c +4 -0
@@ -2331,6 +2331,14 @@ int kvm_arch_init_vcpu(CPUState *cs)
     return r;
 }
 
+int kvm_arch_pre_create_vcpu(CPUState *cpu)
+{
+    if (is_tdx_vm())
+        return tdx_pre_create_vcpu(cpu);

Curly braces needed for coding style - run 'scripts/checkpatch.pl' to validate it.

Quoted 25 lines
target/i386/kvm/kvm.c +4 -0
+
+    return 0;
+}
+
 int kvm_arch_destroy_vcpu(CPUState *cs)
 {
     X86CPU *cpu = X86_CPU(cs);
target/i386/kvm/tdx-stub.c +5 -0
index 1d866d5496bf..61f70cc0d1d9 100644
@@ -6,3 +6,8 @@ int tdx_kvm_init(MachineState *ms, Error **errp)
 {
     return -EINVAL;
 }
+
+int tdx_pre_create_vcpu(CPUState *cpu)
+{
+    return -EINVAL;
+}
target/i386/kvm/tdx.c +0 -0
index 29f50fb9529e..3d313ed46bd1 100644

target/i386/kvm/tdx.c int tdx_kvm_init(MachineState *ms, Error **errp) · line 461

Quoted 4 lines
target/i386/kvm/tdx.c +1 -0
@@ -458,6 +458,49 @@ int tdx_kvm_init(MachineState *ms, Error **errp)
     return 0;
 }
 
+int tdx_pre_create_vcpu(CPUState *cpu)

Add 'Error **errp' to this method

Quoted 8 lines
target/i386/kvm/tdx.c +8 -0
+{
+    MachineState *ms = MACHINE(qdev_get_machine());
+    X86CPU *x86cpu = X86_CPU(cpu);
+    CPUX86State *env = &x86cpu->env;
+    struct kvm_tdx_init_vm *init_vm;
+    int r = 0;
+
+    qemu_mutex_lock(&tdx_guest->lock);

target/i386/kvm/tdx.c int tdx_kvm_init(MachineState *ms, Error **errp) · line 479

Quoted 10 lines
target/i386/kvm/tdx.c +10 -0
+    if (tdx_guest->initialized) {
+        goto out;
+    }
+
+    init_vm = g_malloc0(sizeof(struct kvm_tdx_init_vm) +
+                        sizeof(struct kvm_cpuid_entry2) * KVM_MAX_CPUID_ENTRIES);
+
+    r = kvm_vm_enable_cap(kvm_state, KVM_CAP_MAX_VCPUS, 0, ms->smp.cpus);
+    if (r < 0) {
+        error_report("Unable to set MAX VCPUS to %d", ms->smp.cpus);

Use error_setg / error_setg_errno in this method.

Quoted 60 lines
target/i386/kvm/tdx.c +26 -0
+        goto out_free;
+    }
+
+    init_vm->cpuid.nent = kvm_x86_arch_cpuid(env, init_vm->cpuid.entries, 0);
+
+    init_vm->attributes = tdx_guest->attributes;
+
+    do {
+        r = tdx_vm_ioctl(KVM_TDX_INIT_VM, 0, init_vm);
+    } while (r == -EAGAIN);
+    if (r < 0) {
+        error_report("KVM_TDX_INIT_VM failed %s", strerror(-r));
+        goto out_free;
+    }
+
+    tdx_guest->initialized = true;
+
+out_free:
+    g_free(init_vm);
+out:
+    qemu_mutex_unlock(&tdx_guest->lock);
+    return r;
+}
+
 /* tdx guest */
 OBJECT_DEFINE_TYPE_WITH_INTERFACES(TdxGuest,
                                    tdx_guest,
@@ -470,6 +513,8 @@ static void tdx_guest_init(Object *obj)
 {
     TdxGuest *tdx = TDX_GUEST(obj);
 
+    qemu_mutex_init(&tdx->lock);
+
     tdx->attributes = 0;
 }
 
target/i386/kvm/tdx.h +4 -0
index 06599b65b827..46a24ee8c7cc 100644
@@ -17,6 +17,9 @@ typedef struct TdxGuestClass {
 typedef struct TdxGuest {
     ConfidentialGuestSupport parent_obj;
 
+    QemuMutex lock;
+
+    bool initialized;
     uint64_t attributes;    /* TD attributes */
 } TdxGuest;
 
@@ -29,5 +32,6 @@ bool is_tdx_vm(void);
 int tdx_kvm_init(MachineState *ms, Error **errp);
 void tdx_get_supported_cpuid(uint32_t function, uint32_t index, int reg,
                              uint32_t *ret);
+int tdx_pre_create_vcpu(CPUState *cpu);
 
 #endif /* QEMU_I386_TDX_H */
Signature · 1 line
2.34.1

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
author
#35 lore raw

PATCH v2 15/58 i386/tdx: Add property sept-ve-disable for tdx-guest object

In reply to Xiaoyao Li

Bit 28 of TD attribute, named SEPT_VE_DISABLE. When set to 1, it disables EPT violation conversion to #VE on guest TD access of PENDING pages.

Some guest OS (e.g., Linux TD guest) may require this bit as 1. Otherwise refuse to boot.

Add sept-ve-disable property for tdx-guest object, for user to configure this bit.

---
 qapi/qom.json         |  4 +++-
 target/i386/kvm/tdx.c | 24 ++++++++++++++++++++++++
 2 files changed, 27 insertions(+), 1 deletion(-)
qapi/qom.json +3 -1
index 2ca7ce7c0da5..cc08b9a98df9 100644
@@ -871,10 +871,12 @@
 #
 # Properties for tdx-guest objects.
 #
+# @sept-ve-disable: bit 28 of TD attributes (default: 0)
+#
 # Since: 8.2
 ##
 { 'struct': 'TdxGuestProperties',
-  'data': { }}
+  'data': { '*sept-ve-disable': 'bool' } }
 
 ##
 # @ThreadContextProperties:
target/i386/kvm/tdx.c +24 -0
index 3d313ed46bd1..22130382c0c5 100644
@@ -32,6 +32,8 @@
                                      (1U << KVM_FEATURE_PV_SCHED_YIELD) | \
                                      (1U << KVM_FEATURE_MSI_EXT_DEST_ID))
 
+#define TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE   BIT_ULL(28)
+
 #define TDX_ATTRIBUTES_MAX_BITS      64
 
 static FeatureMask tdx_attrs_ctrl_fields[TDX_ATTRIBUTES_MAX_BITS] = {
@@ -501,6 +503,24 @@ out:
     return r;
 }
 
+static bool tdx_guest_get_sept_ve_disable(Object *obj, Error **errp)
+{
+    TdxGuest *tdx = TDX_GUEST(obj);
+
+    return !!(tdx->attributes & TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE);
+}
+
+static void tdx_guest_set_sept_ve_disable(Object *obj, bool value, Error **errp)
+{
+    TdxGuest *tdx = TDX_GUEST(obj);
+
+    if (value) {
+        tdx->attributes |= TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE;
+    } else {
+        tdx->attributes &= ~TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE;
+    }
+}
+
 /* tdx guest */
 OBJECT_DEFINE_TYPE_WITH_INTERFACES(TdxGuest,
                                    tdx_guest,
@@ -516,6 +536,10 @@ static void tdx_guest_init(Object *obj)
     qemu_mutex_init(&tdx->lock);
 
     tdx->attributes = 0;
+
+    object_property_add_bool(obj, "sept-ve-disable",
+                             tdx_guest_get_sept_ve_disable,
+                             tdx_guest_set_sept_ve_disable);
 }
 
 static void tdx_guest_finalize(Object *obj)
Signature · 1 line
2.34.1
#36 lore raw

Re: [PATCH v2 15/58] i386/tdx: Add property sept-ve-disable for tdx-guest object

Quoted 21 lines · Xiaoyao Li

Bit 28 of TD attribute, named SEPT_VE_DISABLE. When set to 1, it disables EPT violation conversion to #VE on guest TD access of PENDING pages.

Some guest OS (e.g., Linux TD guest) may require this bit as 1. Otherwise refuse to boot.

Add sept-ve-disable property for tdx-guest object, for user to configure this bit.

---
 qapi/qom.json         |  4 +++-
 target/i386/kvm/tdx.c | 24 ++++++++++++++++++++++++
 2 files changed, 27 insertions(+), 1 deletion(-)
qapi/qom.json +0 -0
index 2ca7ce7c0da5..cc08b9a98df9 100644

qapi/qom.json line 874

Quoted 4 lines
qapi/qom.json +1 -0
@@ -871,10 +871,12 @@
 #
 # Properties for tdx-guest objects.
 #
+# @sept-ve-disable: bit 28 of TD attributes (default: 0)

This description isn't very useful as it forces the user to go off and read the TDX specification to find out what bit 28 means. You've got a more useful description in the commit message, so please use that in the docs too. eg something like this

@sept-ve-disable: toggle bit 28 of TD attributes to control disabling
of EPT violation conversion to #VE on guest TD access of PENDING pages. Some guest OS (e.g. Linux TD guest) may require this set, otherwise they refuse to boot.

Quoted 61 lines
qapi/qom.json +2 -1
+#
 # Since: 8.2
 ##
 { 'struct': 'TdxGuestProperties',
-  'data': { }}
+  'data': { '*sept-ve-disable': 'bool' } }
 
 ##
 # @ThreadContextProperties:
target/i386/kvm/tdx.c +24 -0
index 3d313ed46bd1..22130382c0c5 100644
@@ -32,6 +32,8 @@
                                      (1U << KVM_FEATURE_PV_SCHED_YIELD) | \
                                      (1U << KVM_FEATURE_MSI_EXT_DEST_ID))
 
+#define TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE   BIT_ULL(28)
+
 #define TDX_ATTRIBUTES_MAX_BITS      64
 
 static FeatureMask tdx_attrs_ctrl_fields[TDX_ATTRIBUTES_MAX_BITS] = {
@@ -501,6 +503,24 @@ out:
     return r;
 }
 
+static bool tdx_guest_get_sept_ve_disable(Object *obj, Error **errp)
+{
+    TdxGuest *tdx = TDX_GUEST(obj);
+
+    return !!(tdx->attributes & TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE);
+}
+
+static void tdx_guest_set_sept_ve_disable(Object *obj, bool value, Error **errp)
+{
+    TdxGuest *tdx = TDX_GUEST(obj);
+
+    if (value) {
+        tdx->attributes |= TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE;
+    } else {
+        tdx->attributes &= ~TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE;
+    }
+}
+
 /* tdx guest */
 OBJECT_DEFINE_TYPE_WITH_INTERFACES(TdxGuest,
                                    tdx_guest,
@@ -516,6 +536,10 @@ static void tdx_guest_init(Object *obj)
     qemu_mutex_init(&tdx->lock);
 
     tdx->attributes = 0;
+
+    object_property_add_bool(obj, "sept-ve-disable",
+                             tdx_guest_get_sept_ve_disable,
+                             tdx_guest_set_sept_ve_disable);
 }
 
 static void tdx_guest_finalize(Object *obj)
Signature · 1 line
2.34.1

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#37 lore raw

Re: [PATCH v2 15/58] i386/tdx: Add property sept-ve-disable for tdx-guest object

Quoted 17 lines · Daniel P. Berrangé
Quoted 16 lines · Xiaoyao Li

Bit 28 of TD attribute, named SEPT_VE_DISABLE. When set to 1, it disables EPT violation conversion to #VE on guest TD access of PENDING pages.

Some guest OS (e.g., Linux TD guest) may require this bit as 1. Otherwise refuse to boot.

Add sept-ve-disable property for tdx-guest object, for user to configure this bit.

---
 qapi/qom.json         |  4 +++-
 target/i386/kvm/tdx.c | 24 ++++++++++++++++++++++++
 2 files changed, 27 insertions(+), 1 deletion(-)
Quoted 12 lines

qapi/qom.json line 874

Quoted 9 lines
qapi/qom.json +1 -0
index 2ca7ce7c0da5..cc08b9a98df9 100644
@@ -871,10 +871,12 @@
 #
 # Properties for tdx-guest objects.
 #
+# @sept-ve-disable: bit 28 of TD attributes (default: 0)

This description isn't very useful as it forces the user to go off and read the TDX specification to find out what bit 28 means. You've got a

Seconded.

Quoted 8 lines

more useful description in the commit message, so please use that in the docs too. eg something like this

@sept-ve-disable: toggle bit 28 of TD attributes to control disabling
of EPT violation conversion to #VE on guest TD access of PENDING pages. Some guest OS (e.g. Linux TD guest) may require this set, otherwise they refuse to boot.

But please format like

# @sept-ve-disable: toggle bit 28 of TD attributes to control disabling

#     of EPT violation conversion to #VE on guest TD access of PENDING
#     pages.  Some guest OS (e.g. Linux TD guest) may require this to
#     be set, otherwise they refuse to boot.

to blend in with recent commit a937b6aa739 (qapi: Reformat doc comments to conform to current conventions).

Quoted 9 lines
Quoted 9 lines
qapi/qom.json +2 -1
+#
 # Since: 8.2
 ##
 { 'struct': 'TdxGuestProperties',
-  'data': { }}
+  'data': { '*sept-ve-disable': 'bool' } }
 
 ##
 # @ThreadContextProperties:

[...]

author
#38 lore raw

Re: [PATCH v2 15/58] i386/tdx: Add property sept-ve-disable for tdx-guest object

Quoted 43 lines · Markus Armbruster
Quoted 17 lines · Daniel P. Berrangé
Quoted 16 lines · Xiaoyao Li

Bit 28 of TD attribute, named SEPT_VE_DISABLE. When set to 1, it disables EPT violation conversion to #VE on guest TD access of PENDING pages.

Some guest OS (e.g., Linux TD guest) may require this bit as 1. Otherwise refuse to boot.

Add sept-ve-disable property for tdx-guest object, for user to configure this bit.

---
  qapi/qom.json         |  4 +++-
  target/i386/kvm/tdx.c | 24 ++++++++++++++++++++++++
  2 files changed, 27 insertions(+), 1 deletion(-)
Quoted 12 lines

qapi/qom.json line 874

Quoted 9 lines
qapi/qom.json +1 -0
index 2ca7ce7c0da5..cc08b9a98df9 100644
@@ -871,10 +871,12 @@
  #
  # Properties for tdx-guest objects.
  #
+# @sept-ve-disable: bit 28 of TD attributes (default: 0)

This description isn't very useful as it forces the user to go off and read the TDX specification to find out what bit 28 means. You've got a

Seconded.

Quoted 8 lines

more useful description in the commit message, so please use that in the docs too. eg something like this

@sept-ve-disable: toggle bit 28 of TD attributes to control disabling
of EPT violation conversion to #VE on guest TD access of PENDING pages. Some guest OS (e.g. Linux TD guest) may require this set, otherwise they refuse to boot.

Quoted 7 lines

But please format like

# @sept-ve-disable: toggle bit 28 of TD attributes to control disabling

#     of EPT violation conversion to #VE on guest TD access of PENDING
#     pages.  Some guest OS (e.g. Linux TD guest) may require this to
#     be set, otherwise they refuse to boot.

Thank you, Daniel and Markus.

Will use above in the next version.

Quoted 15 lines

to blend in with recent commit a937b6aa739 (qapi: Reformat doc comments to conform to current conventions).

Quoted 9 lines
Quoted 9 lines
qapi/qom.json +2 -1
+#
  # Since: 8.2
  ##
  { 'struct': 'TdxGuestProperties',
-  'data': { }}
+  'data': { '*sept-ve-disable': 'bool' } }
 
  ##
  # @ThreadContextProperties:

[...]

author
#39 lore raw

PATCH v2 16/58 i386/tdx: Make sept_ve_disable set by default

In reply to Xiaoyao Li

From: Isaku Yamahata <[email protected]>

For TDX KVM use case, Linux guest is the most major one. It requires sept_ve_disable set. Make it default for the main use case. For other use case, it can be enabled/disabled via qemu command line.

---
 target/i386/kvm/tdx.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
target/i386/kvm/tdx.c +1 -1
index 22130382c0c5..153a75a02599 100644
@@ -535,7 +535,7 @@ static void tdx_guest_init(Object *obj)
 
     qemu_mutex_init(&tdx->lock);
 
-    tdx->attributes = 0;
+    tdx->attributes = TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE;
 
     object_property_add_bool(obj, "sept-ve-disable",
                              tdx_guest_get_sept_ve_disable,
Signature · 1 line
2.34.1
author
#40 lore raw

PATCH v2 17/58 i386/tdx: Wire CPU features up with attributes of TD guest

In reply to Xiaoyao Li

For QEMU VMs, PKS is configured via CPUID_7_0_ECX_PKS and PMU is configured by x86cpu->enable_pmu. Reuse the existing configuration interface for TDX VMs.

---
 target/i386/kvm/tdx.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)
target/i386/kvm/tdx.c +13 -1
index 153a75a02599..629abd267da8 100644
@@ -33,6 +33,8 @@
                                      (1U << KVM_FEATURE_MSI_EXT_DEST_ID))
 
 #define TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE   BIT_ULL(28)
+#define TDX_TD_ATTRIBUTES_PKS               BIT_ULL(30)
+#define TDX_TD_ATTRIBUTES_PERFMON           BIT_ULL(63)
 
 #define TDX_ATTRIBUTES_MAX_BITS      64
 
@@ -460,6 +462,15 @@ int tdx_kvm_init(MachineState *ms, Error **errp)
     return 0;
 }
 
+static void setup_td_guest_attributes(X86CPU *x86cpu)
+{
+    CPUX86State *env = &x86cpu->env;
+
+    tdx_guest->attributes |= (env->features[FEAT_7_0_ECX] & CPUID_7_0_ECX_PKS) ?
+                             TDX_TD_ATTRIBUTES_PKS : 0;
+    tdx_guest->attributes |= x86cpu->enable_pmu ? TDX_TD_ATTRIBUTES_PERFMON : 0;
+}
+
 int tdx_pre_create_vcpu(CPUState *cpu)
 {
     MachineState *ms = MACHINE(qdev_get_machine());
@@ -482,8 +493,9 @@ int tdx_pre_create_vcpu(CPUState *cpu)
         goto out_free;
     }
 
+    setup_td_guest_attributes(x86cpu);
+
     init_vm->cpuid.nent = kvm_x86_arch_cpuid(env, init_vm->cpuid.entries, 0);
-
     init_vm->attributes = tdx_guest->attributes;
 
     do {
Signature · 1 line
2.34.1
author
#41 lore raw

PATCH v2 18/58 i386/tdx: Validate TD attributes

In reply to Xiaoyao Li

Validate TD attributes with tdx_caps that fixed-0 bits must be zero and fixed-1 bits must be set.

Besides, sanity check the attribute bits that have not been supported by QEMU yet. e.g., debug bit, it will be allowed in the future when debug TD support lands in QEMU.

---
 target/i386/kvm/tdx.c | 27 +++++++++++++++++++++++++--
 1 file changed, 25 insertions(+), 2 deletions(-)
target/i386/kvm/tdx.c +25 -2
index 629abd267da8..73da15377ec3 100644
@@ -32,6 +32,7 @@
                                      (1U << KVM_FEATURE_PV_SCHED_YIELD) | \
                                      (1U << KVM_FEATURE_MSI_EXT_DEST_ID))
 
+#define TDX_TD_ATTRIBUTES_DEBUG             BIT_ULL(0)
 #define TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE   BIT_ULL(28)
 #define TDX_TD_ATTRIBUTES_PKS               BIT_ULL(30)
 #define TDX_TD_ATTRIBUTES_PERFMON           BIT_ULL(63)
@@ -462,13 +463,32 @@ int tdx_kvm_init(MachineState *ms, Error **errp)
     return 0;
 }
 
-static void setup_td_guest_attributes(X86CPU *x86cpu)
+static int tdx_validate_attributes(TdxGuest *tdx)
+{
+    if (((tdx->attributes & tdx_caps->attrs_fixed0) | tdx_caps->attrs_fixed1) !=
+        tdx->attributes) {
+            error_report("Invalid attributes 0x%lx for TDX VM (fixed0 0x%llx, fixed1 0x%llx)",
+                          tdx->attributes, tdx_caps->attrs_fixed0, tdx_caps->attrs_fixed1);
+            return -EINVAL;
+    }
+
+    if (tdx->attributes & TDX_TD_ATTRIBUTES_DEBUG) {
+        error_report("Current QEMU doesn't support attributes.debug[bit 0] for TDX VM");
+        return -EINVAL;
+    }
+
+    return 0;
+}
+
+static int setup_td_guest_attributes(X86CPU *x86cpu)
 {
     CPUX86State *env = &x86cpu->env;
 
     tdx_guest->attributes |= (env->features[FEAT_7_0_ECX] & CPUID_7_0_ECX_PKS) ?
                              TDX_TD_ATTRIBUTES_PKS : 0;
     tdx_guest->attributes |= x86cpu->enable_pmu ? TDX_TD_ATTRIBUTES_PERFMON : 0;
+
+    return tdx_validate_attributes(tdx_guest);
 }
 
 int tdx_pre_create_vcpu(CPUState *cpu)
@@ -493,7 +513,10 @@ int tdx_pre_create_vcpu(CPUState *cpu)
         goto out_free;
     }
 
-    setup_td_guest_attributes(x86cpu);
+    r = setup_td_guest_attributes(x86cpu);
+    if (r) {
+        goto out;
+    }
 
     init_vm->cpuid.nent = kvm_x86_arch_cpuid(env, init_vm->cpuid.entries, 0);
     init_vm->attributes = tdx_guest->attributes;
Signature · 1 line
2.34.1
#42 lore raw

Re: [PATCH v2 18/58] i386/tdx: Validate TD attributes

Quoted 33 lines · Xiaoyao Li

Validate TD attributes with tdx_caps that fixed-0 bits must be zero and fixed-1 bits must be set.

Besides, sanity check the attribute bits that have not been supported by QEMU yet. e.g., debug bit, it will be allowed in the future when debug TD support lands in QEMU.

---
 target/i386/kvm/tdx.c | 27 +++++++++++++++++++++++++--
 1 file changed, 25 insertions(+), 2 deletions(-)
target/i386/kvm/tdx.c +4 -1
index 629abd267da8..73da15377ec3 100644
@@ -32,6 +32,7 @@
                                      (1U << KVM_FEATURE_PV_SCHED_YIELD) | \
                                      (1U << KVM_FEATURE_MSI_EXT_DEST_ID))
 
+#define TDX_TD_ATTRIBUTES_DEBUG             BIT_ULL(0)
 #define TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE   BIT_ULL(28)
 #define TDX_TD_ATTRIBUTES_PKS               BIT_ULL(30)
 #define TDX_TD_ATTRIBUTES_PERFMON           BIT_ULL(63)
@@ -462,13 +463,32 @@ int tdx_kvm_init(MachineState *ms, Error **errp)
     return 0;
 }
 
-static void setup_td_guest_attributes(X86CPU *x86cpu)
+static int tdx_validate_attributes(TdxGuest *tdx)
+{
+    if (((tdx->attributes & tdx_caps->attrs_fixed0) | tdx_caps->attrs_fixed1) !=

target/i386/kvm/tdx.c int tdx_kvm_init(MachineState *ms, Error **errp) · line 478

Quoted 10 lines
target/i386/kvm/tdx.c +10 -0
+        tdx->attributes) {
+            error_report("Invalid attributes 0x%lx for TDX VM (fixed0 0x%llx, fixed1 0x%llx)",
+                          tdx->attributes, tdx_caps->attrs_fixed0, tdx_caps->attrs_fixed1);
+            return -EINVAL;
+    }
+
+    if (tdx->attributes & TDX_TD_ATTRIBUTES_DEBUG) {
+        error_report("Current QEMU doesn't support attributes.debug[bit 0] for TDX VM");
+        return -EINVAL;
+    }

Use error_setg() in both cases, passing in a 'Error **errp' object, and 'return -1' instead of returning an errno value.

target/i386/kvm/tdx.c int tdx_kvm_init(MachineState *ms, Error **errp) · line 491

Quoted 13 lines
target/i386/kvm/tdx.c +7 -0
+
+    return 0;
+}
+
+static int setup_td_guest_attributes(X86CPU *x86cpu)
 {
     CPUX86State *env = &x86cpu->env;
 
     tdx_guest->attributes |= (env->features[FEAT_7_0_ECX] & CPUID_7_0_ECX_PKS) ?
                              TDX_TD_ATTRIBUTES_PKS : 0;
     tdx_guest->attributes |= x86cpu->enable_pmu ? TDX_TD_ATTRIBUTES_PERFMON : 0;
+
+    return tdx_validate_attributes(tdx_guest);

Pass along "errp" into this

target/i386/kvm/tdx.c int tdx_pre_create_vcpu(CPUState *cpu)

Quoted 4 lines
target/i386/kvm/tdx.c +0 -0
 }
 
 int tdx_pre_create_vcpu(CPUState *cpu)
@@ -493,7 +513,10 @@ int tdx_pre_create_vcpu(CPUState *cpu)

In an earlier patch I suggested adding 'Error **errp' to this method...

target/i386/kvm/tdx.c int tdx_pre_create_vcpu(CPUState *cpu) · line 516

Quoted 5 lines
target/i386/kvm/tdx.c +1 -1
         goto out_free;
     }
 
-    setup_td_guest_attributes(x86cpu);
+    r = setup_td_guest_attributes(x86cpu);

...it can also be passed into this method

Quoted 9 lines
target/i386/kvm/tdx.c +3 -0
+    if (r) {
+        goto out;
+    }
 
     init_vm->cpuid.nent = kvm_x86_arch_cpuid(env, init_vm->cpuid.entries, 0);
     init_vm->attributes = tdx_guest->attributes;
Signature · 1 line
2.34.1

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
author
#43 lore raw

Re: [PATCH v2 18/58] i386/tdx: Validate TD attributes

Quoted 39 lines · Daniel P. Berrangé
Quoted 38 lines · Xiaoyao Li

Validate TD attributes with tdx_caps that fixed-0 bits must be zero and fixed-1 bits must be set.

Besides, sanity check the attribute bits that have not been supported by QEMU yet. e.g., debug bit, it will be allowed in the future when debug TD support lands in QEMU.

---
  target/i386/kvm/tdx.c | 27 +++++++++++++++++++++++++--
  1 file changed, 25 insertions(+), 2 deletions(-)
target/i386/kvm/tdx.c +9 -1
index 629abd267da8..73da15377ec3 100644
@@ -32,6 +32,7 @@
                                       (1U << KVM_FEATURE_PV_SCHED_YIELD) | \
                                       (1U << KVM_FEATURE_MSI_EXT_DEST_ID))
 
+#define TDX_TD_ATTRIBUTES_DEBUG             BIT_ULL(0)
  #define TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE   BIT_ULL(28)
  #define TDX_TD_ATTRIBUTES_PKS               BIT_ULL(30)
  #define TDX_TD_ATTRIBUTES_PERFMON           BIT_ULL(63)
@@ -462,13 +463,32 @@ int tdx_kvm_init(MachineState *ms, Error **errp)
      return 0;
  }
 
-static void setup_td_guest_attributes(X86CPU *x86cpu)
+static int tdx_validate_attributes(TdxGuest *tdx)
+{
+    if (((tdx->attributes & tdx_caps->attrs_fixed0) | tdx_caps->attrs_fixed1) !=
+        tdx->attributes) {
+            error_report("Invalid attributes 0x%lx for TDX VM (fixed0 0x%llx, fixed1 0x%llx)",
+                          tdx->attributes, tdx_caps->attrs_fixed0, tdx_caps->attrs_fixed1);
+            return -EINVAL;
+    }
Quoted 8 lines

target/i386/kvm/tdx.c int tdx_kvm_init(MachineState *ms, Error **errp) · line 478

Quoted 5 lines
target/i386/kvm/tdx.c +5 -0
+
+    if (tdx->attributes & TDX_TD_ATTRIBUTES_DEBUG) {
+        error_report("Current QEMU doesn't support attributes.debug[bit 0] for TDX VM");
+        return -EINVAL;
+    }

Use error_setg() in both cases, passing in a 'Error **errp' object, and 'return -1' instead of returning an errno value.

Will do it in next version.

thanks!

Quoted 43 lines

target/i386/kvm/tdx.c int tdx_kvm_init(MachineState *ms, Error **errp) · line 491

Quoted 13 lines
target/i386/kvm/tdx.c +7 -0
+
+    return 0;
+}
+
+static int setup_td_guest_attributes(X86CPU *x86cpu)
  {
      CPUX86State *env = &x86cpu->env;
 
      tdx_guest->attributes |= (env->features[FEAT_7_0_ECX] & CPUID_7_0_ECX_PKS) ?
                               TDX_TD_ATTRIBUTES_PKS : 0;
      tdx_guest->attributes |= x86cpu->enable_pmu ? TDX_TD_ATTRIBUTES_PERFMON : 0;
+
+    return tdx_validate_attributes(tdx_guest);

Pass along "errp" into this

target/i386/kvm/tdx.c int tdx_pre_create_vcpu(CPUState *cpu)

Quoted 4 lines
target/i386/kvm/tdx.c +0 -0
  }
 
  int tdx_pre_create_vcpu(CPUState *cpu)
@@ -493,7 +513,10 @@ int tdx_pre_create_vcpu(CPUState *cpu)

In an earlier patch I suggested adding 'Error **errp' to this method...

target/i386/kvm/tdx.c int tdx_pre_create_vcpu(CPUState *cpu) · line 516

Quoted 5 lines
target/i386/kvm/tdx.c +1 -1
          goto out_free;
      }
 
-    setup_td_guest_attributes(x86cpu);
+    r = setup_td_guest_attributes(x86cpu);

...it can also be passed into this method

Quoted 9 lines
target/i386/kvm/tdx.c +3 -0
+    if (r) {
+        goto out;
+    }
 
      init_vm->cpuid.nent = kvm_x86_arch_cpuid(env, init_vm->cpuid.entries, 0);
      init_vm->attributes = tdx_guest->attributes;
Signature · 1 line
2.34.1

With regards,
Daniel

author
#44 lore raw

Re: [PATCH v2 18/58] i386/tdx: Validate TD attributes

In reply to Daniel P. Berrangé
Quoted 39 lines · Daniel P. Berrangé
Quoted 38 lines · Xiaoyao Li

Validate TD attributes with tdx_caps that fixed-0 bits must be zero and fixed-1 bits must be set.

Besides, sanity check the attribute bits that have not been supported by QEMU yet. e.g., debug bit, it will be allowed in the future when debug TD support lands in QEMU.

---
  target/i386/kvm/tdx.c | 27 +++++++++++++++++++++++++--
  1 file changed, 25 insertions(+), 2 deletions(-)
target/i386/kvm/tdx.c +9 -1
index 629abd267da8..73da15377ec3 100644
@@ -32,6 +32,7 @@
                                       (1U << KVM_FEATURE_PV_SCHED_YIELD) | \
                                       (1U << KVM_FEATURE_MSI_EXT_DEST_ID))
 
+#define TDX_TD_ATTRIBUTES_DEBUG             BIT_ULL(0)
  #define TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE   BIT_ULL(28)
  #define TDX_TD_ATTRIBUTES_PKS               BIT_ULL(30)
  #define TDX_TD_ATTRIBUTES_PERFMON           BIT_ULL(63)
@@ -462,13 +463,32 @@ int tdx_kvm_init(MachineState *ms, Error **errp)
      return 0;
  }
 
-static void setup_td_guest_attributes(X86CPU *x86cpu)
+static int tdx_validate_attributes(TdxGuest *tdx)
+{
+    if (((tdx->attributes & tdx_caps->attrs_fixed0) | tdx_caps->attrs_fixed1) !=
+        tdx->attributes) {
+            error_report("Invalid attributes 0x%lx for TDX VM (fixed0 0x%llx, fixed1 0x%llx)",
+                          tdx->attributes, tdx_caps->attrs_fixed0, tdx_caps->attrs_fixed1);
+            return -EINVAL;
+    }
Quoted 9 lines

target/i386/kvm/tdx.c int tdx_kvm_init(MachineState *ms, Error **errp) · line 478

Quoted 5 lines
target/i386/kvm/tdx.c +5 -0
+
+    if (tdx->attributes & TDX_TD_ATTRIBUTES_DEBUG) {
+        error_report("Current QEMU doesn't support attributes.debug[bit 0] for TDX VM");
+        return -EINVAL;
+    }

Use error_setg() in both cases, passing in a 'Error **errp' object, and 'return -1' instead of returning an errno value.

why return -1 instead of -EINVAL?

#45 lore raw

Re: [PATCH v2 18/58] i386/tdx: Validate TD attributes

Quoted 48 lines · Xiaoyao Li
Quoted 37 lines · Daniel P. Berrangé
Quoted 36 lines · Xiaoyao Li

Validate TD attributes with tdx_caps that fixed-0 bits must be zero and fixed-1 bits must be set.

Besides, sanity check the attribute bits that have not been supported by QEMU yet. e.g., debug bit, it will be allowed in the future when debug TD support lands in QEMU.

---
  target/i386/kvm/tdx.c | 27 +++++++++++++++++++++++++--
  1 file changed, 25 insertions(+), 2 deletions(-)
target/i386/kvm/tdx.c +9 -1
index 629abd267da8..73da15377ec3 100644
@@ -32,6 +32,7 @@
                                       (1U << KVM_FEATURE_PV_SCHED_YIELD) | \
                                       (1U << KVM_FEATURE_MSI_EXT_DEST_ID))
+#define TDX_TD_ATTRIBUTES_DEBUG             BIT_ULL(0)
  #define TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE   BIT_ULL(28)
  #define TDX_TD_ATTRIBUTES_PKS               BIT_ULL(30)
  #define TDX_TD_ATTRIBUTES_PERFMON           BIT_ULL(63)
@@ -462,13 +463,32 @@ int tdx_kvm_init(MachineState *ms, Error **errp)
      return 0;
  }
-static void setup_td_guest_attributes(X86CPU *x86cpu)
+static int tdx_validate_attributes(TdxGuest *tdx)
+{
+    if (((tdx->attributes & tdx_caps->attrs_fixed0) | tdx_caps->attrs_fixed1) !=
+        tdx->attributes) {
+            error_report("Invalid attributes 0x%lx for TDX VM (fixed0 0x%llx, fixed1 0x%llx)",
+                          tdx->attributes, tdx_caps->attrs_fixed0, tdx_caps->attrs_fixed1);
+            return -EINVAL;
+    }
Quoted 9 lines

target/i386/kvm/tdx.c int tdx_kvm_init(MachineState *ms, Error **errp) · line 478

Quoted 5 lines
target/i386/kvm/tdx.c +5 -0
+
+    if (tdx->attributes & TDX_TD_ATTRIBUTES_DEBUG) {
+        error_report("Current QEMU doesn't support attributes.debug[bit 0] for TDX VM");
+        return -EINVAL;
+    }

Use error_setg() in both cases, passing in a 'Error **errp' object, and 'return -1' instead of returning an errno value.

Quoted 1 line

why return -1 instead of -EINVAL?

Returning errno values is useful if the method isn't providing an "Error **errp" parameter, because it lets the caller report a more detailed error message via strerror(). Once you add a Error ** parameter though, there is almost never any reason for the caller to care about the original errno value, and so we use 0 / -1 as success/fail indicators.

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
author
#46 lore raw

Re: [PATCH v2 18/58] i386/tdx: Validate TD attributes

Quoted 49 lines · Daniel P. Berrangé
Quoted 48 lines · Xiaoyao Li
Quoted 46 lines · Daniel P. Berrangé
Quoted 31 lines · Xiaoyao Li

Validate TD attributes with tdx_caps that fixed-0 bits must be zero and fixed-1 bits must be set.

Besides, sanity check the attribute bits that have not been supported by QEMU yet. e.g., debug bit, it will be allowed in the future when debug TD support lands in QEMU.

---
   target/i386/kvm/tdx.c | 27 +++++++++++++++++++++++++--
   1 file changed, 25 insertions(+), 2 deletions(-)
target/i386/kvm/tdx.c +4 -1
index 629abd267da8..73da15377ec3 100644
@@ -32,6 +32,7 @@
                                        (1U << KVM_FEATURE_PV_SCHED_YIELD) | \
                                        (1U << KVM_FEATURE_MSI_EXT_DEST_ID))
+#define TDX_TD_ATTRIBUTES_DEBUG             BIT_ULL(0)
   #define TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE   BIT_ULL(28)
   #define TDX_TD_ATTRIBUTES_PKS               BIT_ULL(30)
   #define TDX_TD_ATTRIBUTES_PERFMON           BIT_ULL(63)
@@ -462,13 +463,32 @@ int tdx_kvm_init(MachineState *ms, Error **errp)
       return 0;
   }
-static void setup_td_guest_attributes(X86CPU *x86cpu)
+static int tdx_validate_attributes(TdxGuest *tdx)
+{
+    if (((tdx->attributes & tdx_caps->attrs_fixed0) | tdx_caps->attrs_fixed1) !=

target/i386/kvm/tdx.c int tdx_kvm_init(MachineState *ms, Error **errp) · line 477

Quoted 10 lines
target/i386/kvm/tdx.c +10 -0
+        tdx->attributes) {
+            error_report("Invalid attributes 0x%lx for TDX VM (fixed0 0x%llx, fixed1 0x%llx)",
+                          tdx->attributes, tdx_caps->attrs_fixed0, tdx_caps->attrs_fixed1);
+            return -EINVAL;
+    }
+
+    if (tdx->attributes & TDX_TD_ATTRIBUTES_DEBUG) {
+        error_report("Current QEMU doesn't support attributes.debug[bit 0] for TDX VM");
+        return -EINVAL;
+    }

Use error_setg() in both cases, passing in a 'Error **errp' object, and 'return -1' instead of returning an errno value.

Quoted 8 lines
Quoted 1 line

why return -1 instead of -EINVAL?

Returning errno values is useful if the method isn't providing an "Error **errp" parameter, because it lets the caller report a more detailed error message via strerror(). Once you add a Error ** parameter though, there is almost never any reason for the caller to care about the original errno value, and so we use 0 / -1 as success/fail indicators.

I see.

Thanks,
-Xiaoyao

Quoted 2 lines

With regards,
Daniel

author
#47 lore raw

PATCH v2 19/58 qom: implement property helper for sha384

In reply to Xiaoyao Li

From: Isaku Yamahata <[email protected]>

Implement property_add_sha384() which converts hex string <-> uint8_t[48] It will be used for TDX which uses sha384 for measurement.

---
 include/qom/object.h | 17 ++++++++++
 qom/object.c         | 76 ++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 93 insertions(+)
include/qom/object.h +17 -0
index ef7258a5e149..70399a5b1940 100644
@@ -1887,6 +1887,23 @@ ObjectProperty *object_property_add_alias(Object *obj, const char *name,
 ObjectProperty *object_property_add_const_link(Object *obj, const char *name,
                                                Object *target);
 
+
+/**
+ * object_property_add_sha384:
+ * @obj: the object to add a property to
+ * @name: the name of the property
+ * @v: pointer to value
+ * @flags: bitwise-or'd ObjectPropertyFlags
+ *
+ * Add an sha384 property in memory.  This function will add a
+ * property of type 'sha384'.
+ *
+ * Returns: The newly added property on success, or %NULL on failure.
+ */
+ObjectProperty * object_property_add_sha384(Object *obj, const char *name,
+                                            const uint8_t *v,
+                                            ObjectPropertyFlags flags);
+
 /**
  * object_property_set_description:
  * @obj: the object owning the property
qom/object.c +76 -0

88 lines not on this page · raw message

Signature · 1 line
2.34.1
#48 lore raw

Re: [PATCH v2 19/58] qom: implement property helper for sha384

Quoted 4 lines · Xiaoyao Li

From: Isaku Yamahata <[email protected]>

Implement property_add_sha384() which converts hex string <-> uint8_t[48] It will be used for TDX which uses sha384 for measurement.

I think it is likely a better idea to use base64 for the encoding the binary hash - we use base64 for all the sev-guest properties that were binary data.

At which points the property set/get logic is much simpler as it is just needing a call to g_base64_encode / g_base64_decode and length validation for the decode case.

Quoted 133 lines
---
 include/qom/object.h | 17 ++++++++++
 qom/object.c         | 76 ++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 93 insertions(+)
include/qom/object.h +17 -0
index ef7258a5e149..70399a5b1940 100644
@@ -1887,6 +1887,23 @@ ObjectProperty *object_property_add_alias(Object *obj, const char *name,
 ObjectProperty *object_property_add_const_link(Object *obj, const char *name,
                                                Object *target);
 
+
+/**
+ * object_property_add_sha384:
+ * @obj: the object to add a property to
+ * @name: the name of the property
+ * @v: pointer to value
+ * @flags: bitwise-or'd ObjectPropertyFlags
+ *
+ * Add an sha384 property in memory.  This function will add a
+ * property of type 'sha384'.
+ *
+ * Returns: The newly added property on success, or %NULL on failure.
+ */
+ObjectProperty * object_property_add_sha384(Object *obj, const char *name,
+                                            const uint8_t *v,
+                                            ObjectPropertyFlags flags);
+
 /**
  * object_property_set_description:
  * @obj: the object owning the property
qom/object.c +76 -0
index e25f1e96db1e..e71ce46ed576 100644
@@ -15,6 +15,7 @@
 #include "qapi/error.h"
 #include "qom/object.h"
 #include "qom/object_interfaces.h"
+#include "qemu/ctype.h"
 #include "qemu/cutils.h"
 #include "qemu/memalign.h"
 #include "qapi/visitor.h"
@@ -2781,6 +2782,81 @@ object_property_add_alias(Object *obj, const char *name,
     return op;
 }
 
+#define SHA384_DIGEST_SIZE      48
+static void property_get_sha384(Object *obj, Visitor *v, const char *name,
+                                void *opaque, Error **errp)
+{
+    uint8_t *value = (uint8_t *)opaque;
+    char str[SHA384_DIGEST_SIZE * 2 + 1];
+    char *str_ = (char*)str;
+    size_t i;
+
+    for (i = 0; i < SHA384_DIGEST_SIZE; i++) {
+        char *buf;
+        buf = &str[i * 2];
+
+        sprintf(buf, "%02hhx", value[i]);
+    }
+    str[SHA384_DIGEST_SIZE * 2] = '\0';
+
+    visit_type_str(v, name, &str_, errp);
+}
+
+static void property_set_sha384(Object *obj, Visitor *v, const char *name,
+                                    void *opaque, Error **errp)
+{
+    uint8_t *value = (uint8_t *)opaque;
+    char* str;
+    size_t len;
+    size_t i;
+
+    if (!visit_type_str(v, name, &str, errp)) {
+        goto err;
+    }
+
+    len = strlen(str);
+    if (len != SHA384_DIGEST_SIZE * 2) {
+        error_setg(errp, "invalid length for sha348 hex string %s. "
+                   "it must be 48 * 2 hex", name);
+        goto err;
+    }
+
+    for (i = 0; i < SHA384_DIGEST_SIZE; i++) {
+        if (!qemu_isxdigit(str[i * 2]) || !qemu_isxdigit(str[i * 2 + 1])) {
+            error_setg(errp, "invalid char for sha318 hex string %s at %c%c",
+                       name, str[i * 2], str[i * 2 + 1]);
+            goto err;
+        }
+
+        if (sscanf(str + i * 2, "%02hhx", &value[i]) != 1) {
+            error_setg(errp, "invalid format for sha318 hex string %s", name);
+            goto err;
+        }
+    }
+
+err:
+    g_free(str);
+}
+
+ObjectProperty *
+object_property_add_sha384(Object *obj, const char *name,
+                           const uint8_t *v, ObjectPropertyFlags flags)
+{
+    ObjectPropertyAccessor *getter = NULL;
+    ObjectPropertyAccessor *setter = NULL;
+
+    if ((flags & OBJ_PROP_FLAG_READ) == OBJ_PROP_FLAG_READ) {
+        getter = property_get_sha384;
+    }
+
+    if ((flags & OBJ_PROP_FLAG_WRITE) == OBJ_PROP_FLAG_WRITE) {
+        setter = property_set_sha384;
+    }
+
+    return object_property_add(obj, name, "sha384",
+                               getter, setter, NULL, (void *)v);
+}
+
 void object_property_set_description(Object *obj, const char *name,
                                      const char *description)
 {
Signature · 1 line
2.34.1

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#49 lore raw

Re: [PATCH v2 19/58] qom: implement property helper for sha384

Quoted 6 lines · Daniel P. Berrangé
Quoted 4 lines · Xiaoyao Li

From: Isaku Yamahata <[email protected]>

Implement property_add_sha384() which converts hex string <-> uint8_t[48] It will be used for TDX which uses sha384 for measurement.

Quoted 7 lines

I think it is likely a better idea to use base64 for the encoding the binary hash - we use base64 for all the sev-guest properties that were binary data.

At which points the property set/get logic is much simpler as it is just needing a call to g_base64_encode / g_base64_decode and length validation for the decode case.

Hex string is poplar to show hash value, isn't it? Anyway it's easy for human operator, shell scripts, libvirt or whatever to convert those representations with utility commands like base64 or xxd, or library call. Either way would work.

Signature · 1 line
Isaku Yamahata <[email protected]>
author
#50 lore raw

PATCH v2 20/58 i386/tdx: Allows mrconfigid/mrowner/mrownerconfig for TDX_INIT_VM

In reply to Xiaoyao Li

From: Isaku Yamahata <[email protected]>

When creating TDX vm, three sha384 hash values can be provided for TDX attestation.

So far they were hard coded as 0. Now allow user to specify those values via property mrconfigid, mrowner and mrownerconfig. Choose hex-encoded string as format since it's friendly for user to input.

example
-object tdx-guest, \
mrconfigid=0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef, \
mrowner=fedcba9876543210fedcba9876543210fedcba9876543210fedcba9876543210fedcba9876543210fedcba9876543210, \
mrownerconfig=0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef

---
TODO:
- community requests to use base64 encoding if no special reason

---
 qapi/qom.json         | 11 ++++++++++-
 target/i386/kvm/tdx.c | 13 +++++++++++++
 target/i386/kvm/tdx.h |  3 +++
 3 files changed, 26 insertions(+), 1 deletion(-)
qapi/qom.json +10 -1
index cc08b9a98df9..87c1d440f331 100644
@@ -873,10 +873,19 @@
 #
 # @sept-ve-disable: bit 28 of TD attributes (default: 0)
 #
+# @mrconfigid: MRCONFIGID SHA384 hex string of 48 * 2 length (default: 0)
+#
+# @mrowner: MROWNER SHA384 hex string of 48 * 2 length (default: 0)
+#
+# @mrownerconfig: MROWNERCONFIG SHA384 hex string of 48 * 2 length (default: 0)
+#
 # Since: 8.2
 ##
 { 'struct': 'TdxGuestProperties',
-  'data': { '*sept-ve-disable': 'bool' } }
+  'data': { '*sept-ve-disable': 'bool',
+            '*mrconfigid': 'str',
+            '*mrowner': 'str',
+            '*mrownerconfig': 'str' } }
 
 ##
 # @ThreadContextProperties:
target/i386/kvm/tdx.c +13 -0
index 73da15377ec3..33d015a08c34 100644
@@ -521,6 +521,13 @@ int tdx_pre_create_vcpu(CPUState *cpu)
     init_vm->cpuid.nent = kvm_x86_arch_cpuid(env, init_vm->cpuid.entries, 0);
     init_vm->attributes = tdx_guest->attributes;
 
+    QEMU_BUILD_BUG_ON(sizeof(init_vm->mrconfigid) != sizeof(tdx_guest->mrconfigid));
+    QEMU_BUILD_BUG_ON(sizeof(init_vm->mrowner) != sizeof(tdx_guest->mrowner));
+    QEMU_BUILD_BUG_ON(sizeof(init_vm->mrownerconfig) != sizeof(tdx_guest->mrownerconfig));
+    memcpy(init_vm->mrconfigid, tdx_guest->mrconfigid, sizeof(tdx_guest->mrconfigid));
+    memcpy(init_vm->mrowner, tdx_guest->mrowner, sizeof(tdx_guest->mrowner));
+    memcpy(init_vm->mrownerconfig, tdx_guest->mrownerconfig, sizeof(tdx_guest->mrownerconfig));
+
     do {
         r = tdx_vm_ioctl(KVM_TDX_INIT_VM, 0, init_vm);
     } while (r == -EAGAIN);
@@ -575,6 +582,12 @@ static void tdx_guest_init(Object *obj)
     object_property_add_bool(obj, "sept-ve-disable",
                              tdx_guest_get_sept_ve_disable,
                              tdx_guest_set_sept_ve_disable);
+    object_property_add_sha384(obj, "mrconfigid", tdx->mrconfigid,
+                               OBJ_PROP_FLAG_READWRITE);
+    object_property_add_sha384(obj, "mrowner", tdx->mrowner,
+                               OBJ_PROP_FLAG_READWRITE);
+    object_property_add_sha384(obj, "mrownerconfig", tdx->mrownerconfig,
+                               OBJ_PROP_FLAG_READWRITE);
 }
 
 static void tdx_guest_finalize(Object *obj)
target/i386/kvm/tdx.h +3 -0
index 46a24ee8c7cc..68f8327f2231 100644
@@ -21,6 +21,9 @@ typedef struct TdxGuest {
 
     bool initialized;
     uint64_t attributes;    /* TD attributes */
+    uint8_t mrconfigid[48];     /* sha348 digest */
+    uint8_t mrowner[48];        /* sha348 digest */
+    uint8_t mrownerconfig[48];  /* sha348 digest */
 } TdxGuest;
 
 #ifdef CONFIG_TDX
Signature · 1 line
2.34.1
#51 lore raw

Re: [PATCH v2 20/58] i386/tdx: Allows mrconfigid/mrowner/mrownerconfig for TDX_INIT_VM

Quoted 31 lines · Xiaoyao Li

From: Isaku Yamahata <[email protected]>

When creating TDX vm, three sha384 hash values can be provided for TDX attestation.

So far they were hard coded as 0. Now allow user to specify those values via property mrconfigid, mrowner and mrownerconfig. Choose hex-encoded string as format since it's friendly for user to input.

example
-object tdx-guest, \
mrconfigid=0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef, \
mrowner=fedcba9876543210fedcba9876543210fedcba9876543210fedcba9876543210fedcba9876543210fedcba9876543210, \
mrownerconfig=0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef

---
TODO:
- community requests to use base64 encoding if no special reason

---
 qapi/qom.json         | 11 ++++++++++-
 target/i386/kvm/tdx.c | 13 +++++++++++++
 target/i386/kvm/tdx.h |  3 +++
 3 files changed, 26 insertions(+), 1 deletion(-)
qapi/qom.json +0 -0
index cc08b9a98df9..87c1d440f331 100644

qapi/qom.json line 880

Quoted 8 lines
qapi/qom.json +5 -0
@@ -873,10 +873,19 @@
 #
 # @sept-ve-disable: bit 28 of TD attributes (default: 0)
 #
+# @mrconfigid: MRCONFIGID SHA384 hex string of 48 * 2 length (default: 0)
+#
+# @mrowner: MROWNER SHA384 hex string of 48 * 2 length (default: 0)
+#
+# @mrownerconfig: MROWNERCONFIG SHA384 hex string of 48 * 2 length (default: 0)

Per previous patch, I suggest these should all be passed in base64 instead of hex. Also 'default: 0' makes no sense for a string, which would be 'default: nil', and no need to document that as the default is implicit from the fact that its an optional string field. So eg

@mrconfigid: base64 encoded MRCONFIGID SHA384 digest

Quoted 60 lines
qapi/qom.json +5 -1
+#
 # Since: 8.2
 ##
 { 'struct': 'TdxGuestProperties',
-  'data': { '*sept-ve-disable': 'bool' } }
+  'data': { '*sept-ve-disable': 'bool',
+            '*mrconfigid': 'str',
+            '*mrowner': 'str',
+            '*mrownerconfig': 'str' } }
 
 ##
 # @ThreadContextProperties:
target/i386/kvm/tdx.c +13 -0
index 73da15377ec3..33d015a08c34 100644
@@ -521,6 +521,13 @@ int tdx_pre_create_vcpu(CPUState *cpu)
     init_vm->cpuid.nent = kvm_x86_arch_cpuid(env, init_vm->cpuid.entries, 0);
     init_vm->attributes = tdx_guest->attributes;
 
+    QEMU_BUILD_BUG_ON(sizeof(init_vm->mrconfigid) != sizeof(tdx_guest->mrconfigid));
+    QEMU_BUILD_BUG_ON(sizeof(init_vm->mrowner) != sizeof(tdx_guest->mrowner));
+    QEMU_BUILD_BUG_ON(sizeof(init_vm->mrownerconfig) != sizeof(tdx_guest->mrownerconfig));
+    memcpy(init_vm->mrconfigid, tdx_guest->mrconfigid, sizeof(tdx_guest->mrconfigid));
+    memcpy(init_vm->mrowner, tdx_guest->mrowner, sizeof(tdx_guest->mrowner));
+    memcpy(init_vm->mrownerconfig, tdx_guest->mrownerconfig, sizeof(tdx_guest->mrownerconfig));
+
     do {
         r = tdx_vm_ioctl(KVM_TDX_INIT_VM, 0, init_vm);
     } while (r == -EAGAIN);
@@ -575,6 +582,12 @@ static void tdx_guest_init(Object *obj)
     object_property_add_bool(obj, "sept-ve-disable",
                              tdx_guest_get_sept_ve_disable,
                              tdx_guest_set_sept_ve_disable);
+    object_property_add_sha384(obj, "mrconfigid", tdx->mrconfigid,
+                               OBJ_PROP_FLAG_READWRITE);
+    object_property_add_sha384(obj, "mrowner", tdx->mrowner,
+                               OBJ_PROP_FLAG_READWRITE);
+    object_property_add_sha384(obj, "mrownerconfig", tdx->mrownerconfig,
+                               OBJ_PROP_FLAG_READWRITE);
 }
 
 static void tdx_guest_finalize(Object *obj)
target/i386/kvm/tdx.h +3 -0
index 46a24ee8c7cc..68f8327f2231 100644
@@ -21,6 +21,9 @@ typedef struct TdxGuest {
 
     bool initialized;
     uint64_t attributes;    /* TD attributes */
+    uint8_t mrconfigid[48];     /* sha348 digest */
+    uint8_t mrowner[48];        /* sha348 digest */
+    uint8_t mrownerconfig[48];  /* sha348 digest */
 } TdxGuest;
 
 #ifdef CONFIG_TDX
Signature · 1 line
2.34.1

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#52 lore raw

Re: [PATCH v2 20/58] i386/tdx: Allows mrconfigid/mrowner/mrownerconfig for TDX_INIT_VM

Quoted 27 lines · Daniel P. Berrangé
Quoted 26 lines · Xiaoyao Li

From: Isaku Yamahata <[email protected]>

When creating TDX vm, three sha384 hash values can be provided for TDX attestation.

So far they were hard coded as 0. Now allow user to specify those values via property mrconfigid, mrowner and mrownerconfig. Choose hex-encoded string as format since it's friendly for user to input.

example
-object tdx-guest, \
mrconfigid=0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef, \
mrowner=fedcba9876543210fedcba9876543210fedcba9876543210fedcba9876543210fedcba9876543210fedcba9876543210, \
mrownerconfig=0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef

---
TODO:
- community requests to use base64 encoding if no special reason

---
 qapi/qom.json         | 11 ++++++++++-
 target/i386/kvm/tdx.c | 13 +++++++++++++
 target/i386/kvm/tdx.h |  3 +++
 3 files changed, 26 insertions(+), 1 deletion(-)
Quoted 16 lines

qapi/qom.json line 880

Quoted 13 lines
qapi/qom.json +5 -0
index cc08b9a98df9..87c1d440f331 100644
@@ -873,10 +873,19 @@
 #
 # @sept-ve-disable: bit 28 of TD attributes (default: 0)
 #
+# @mrconfigid: MRCONFIGID SHA384 hex string of 48 * 2 length (default: 0)
+#
+# @mrowner: MROWNER SHA384 hex string of 48 * 2 length (default: 0)
+#
+# @mrownerconfig: MROWNERCONFIG SHA384 hex string of 48 * 2 length (default: 0)

Per previous patch, I suggest these should all be passed in base64 instead of hex.

I'm upgrading this suggestion to a demand: we use base64 for encoding binary data everywhere in QAPI/QMP. Consistency matters.

Quoted 6 lines

Also 'default: 0' makes no sense for a string,
which would be 'default: nil', and no need to document that as the default is implicit from the fact that its an optional string field. So eg

@mrconfigid: base64 encoded MRCONFIGID SHA384 digest

Agree.

The member names are abbreviations all run together, wheras QAPI/QMP favors words-separated-with-dashes. If you invented them, please change them to QAPI/QMP style. If they are established TDX terminology, keep them as they are, but please point to your evidence.

Quoted 12 lines
Quoted 12 lines
qapi/qom.json +5 -1
+#
 # Since: 8.2
 ##
 { 'struct': 'TdxGuestProperties',
-  'data': { '*sept-ve-disable': 'bool' } }
+  'data': { '*sept-ve-disable': 'bool',
+            '*mrconfigid': 'str',
+            '*mrowner': 'str',
+            '*mrownerconfig': 'str' } }
 
 ##
 # @ThreadContextProperties:

[...]

author
#53 lore raw

PATCH v2 21/58 i386/tdx: Implement user specified tsc frequency

In reply to Xiaoyao Li

Reuse "-cpu,tsc-frequency=" to get user wanted tsc frequency and call VM scope VM_SET_TSC_KHZ to set the tsc frequency of TD before KVM_TDX_INIT_VM.

Besides, sanity check the tsc frequency to be in the legal range and legal granularity (required by TDX module).

---
Changes from RFC v4:
- Use VM scope VM_SET_TSC_KHZ to set the TSC frequency of TD since KVM side drop the @tsc_khz field in struct kvm_tdx_init_vm

---
 target/i386/kvm/kvm.c |  9 +++++++++
 target/i386/kvm/tdx.c | 24 ++++++++++++++++++++++++
 2 files changed, 33 insertions(+)
target/i386/kvm/kvm.c +9 -0
index d51067fdc12a..4a146bc42f63 100644
@@ -859,6 +859,15 @@ static int kvm_arch_set_tsc_khz(CPUState *cs)
     int r, cur_freq;
     bool set_ioctl = false;
 
+    /*
+     * TSC of TD vcpu is immutable, it cannot be set/changed via vcpu scope
+     * VM_SET_TSC_KHZ, but only be initialized via VM scope VM_SET_TSC_KHZ
+     * before ioctl KVM_TDX_INIT_VM in tdx_pre_create_vcpu()
+     */
+    if (is_tdx_vm()) {
+        return 0;
+    }
+
     if (!env->tsc_khz) {
         return 0;
     }
target/i386/kvm/tdx.c +24 -0
index 33d015a08c34..a72badfbfd65 100644
@@ -32,6 +32,9 @@
                                      (1U << KVM_FEATURE_PV_SCHED_YIELD) | \
                                      (1U << KVM_FEATURE_MSI_EXT_DEST_ID))
 
+#define TDX_MIN_TSC_FREQUENCY_KHZ   (100 * 1000)
+#define TDX_MAX_TSC_FREQUENCY_KHZ   (10 * 1000 * 1000)
+
 #define TDX_TD_ATTRIBUTES_DEBUG             BIT_ULL(0)
 #define TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE   BIT_ULL(28)
 #define TDX_TD_ATTRIBUTES_PKS               BIT_ULL(30)
@@ -513,6 +516,27 @@ int tdx_pre_create_vcpu(CPUState *cpu)
         goto out_free;
     }
 
+    r = -EINVAL;
+    if (env->tsc_khz && (env->tsc_khz < TDX_MIN_TSC_FREQUENCY_KHZ ||
+                         env->tsc_khz > TDX_MAX_TSC_FREQUENCY_KHZ)) {
+        error_report("Invalid TSC %ld KHz, must specify cpu_frequency between [%d, %d] kHz",
+                      env->tsc_khz, TDX_MIN_TSC_FREQUENCY_KHZ,
+                      TDX_MAX_TSC_FREQUENCY_KHZ);
+        goto out;
+    }
+
+    if (env->tsc_khz % (25 * 1000)) {
+        error_report("Invalid TSC %ld KHz, it must be multiple of 25MHz", env->tsc_khz);
+        goto out;
+    }
+
+    /* it's safe even env->tsc_khz is 0. KVM uses host's tsc_khz in this case */
+    r = kvm_vm_ioctl(kvm_state, KVM_SET_TSC_KHZ, env->tsc_khz);
+    if (r < 0) {
+        error_report("Unable to set TSC frequency to %" PRId64 " kHz", env->tsc_khz);
+        goto out;
+    }
+
     r = setup_td_guest_attributes(x86cpu);
     if (r) {
         goto out;
Signature · 1 line
2.34.1
#54 lore raw

Re: [PATCH v2 21/58] i386/tdx: Implement user specified tsc frequency

Quoted 65 lines · Xiaoyao Li

Reuse "-cpu,tsc-frequency=" to get user wanted tsc frequency and call VM scope VM_SET_TSC_KHZ to set the tsc frequency of TD before KVM_TDX_INIT_VM.

Besides, sanity check the tsc frequency to be in the legal range and legal granularity (required by TDX module).

---
Changes from RFC v4:
- Use VM scope VM_SET_TSC_KHZ to set the TSC frequency of TD since KVM side drop the @tsc_khz field in struct kvm_tdx_init_vm

---
 target/i386/kvm/kvm.c |  9 +++++++++
 target/i386/kvm/tdx.c | 24 ++++++++++++++++++++++++
 2 files changed, 33 insertions(+)
target/i386/kvm/kvm.c +9 -0
index d51067fdc12a..4a146bc42f63 100644
@@ -859,6 +859,15 @@ static int kvm_arch_set_tsc_khz(CPUState *cs)
     int r, cur_freq;
     bool set_ioctl = false;
 
+    /*
+     * TSC of TD vcpu is immutable, it cannot be set/changed via vcpu scope
+     * VM_SET_TSC_KHZ, but only be initialized via VM scope VM_SET_TSC_KHZ
+     * before ioctl KVM_TDX_INIT_VM in tdx_pre_create_vcpu()
+     */
+    if (is_tdx_vm()) {
+        return 0;
+    }
+
     if (!env->tsc_khz) {
         return 0;
     }
target/i386/kvm/tdx.c +13 -0
index 33d015a08c34..a72badfbfd65 100644
@@ -32,6 +32,9 @@
                                      (1U << KVM_FEATURE_PV_SCHED_YIELD) | \
                                      (1U << KVM_FEATURE_MSI_EXT_DEST_ID))
 
+#define TDX_MIN_TSC_FREQUENCY_KHZ   (100 * 1000)
+#define TDX_MAX_TSC_FREQUENCY_KHZ   (10 * 1000 * 1000)
+
 #define TDX_TD_ATTRIBUTES_DEBUG             BIT_ULL(0)
 #define TDX_TD_ATTRIBUTES_SEPT_VE_DISABLE   BIT_ULL(28)
 #define TDX_TD_ATTRIBUTES_PKS               BIT_ULL(30)
@@ -513,6 +516,27 @@ int tdx_pre_create_vcpu(CPUState *cpu)
         goto out_free;
     }
 
+    r = -EINVAL;
+    if (env->tsc_khz && (env->tsc_khz < TDX_MIN_TSC_FREQUENCY_KHZ ||
+                         env->tsc_khz > TDX_MAX_TSC_FREQUENCY_KHZ)) {
+        error_report("Invalid TSC %ld KHz, must specify cpu_frequency between [%d, %d] kHz",
+                      env->tsc_khz, TDX_MIN_TSC_FREQUENCY_KHZ,
+                      TDX_MAX_TSC_FREQUENCY_KHZ);
+        goto out;
+    }
+
+    if (env->tsc_khz % (25 * 1000)) {

target/i386/kvm/tdx.c int tdx_pre_create_vcpu(CPUState *cpu) · line 538

Quoted 10 lines
target/i386/kvm/tdx.c +10 -0
+        error_report("Invalid TSC %ld KHz, it must be multiple of 25MHz", env->tsc_khz);
+        goto out;
+    }
+
+    /* it's safe even env->tsc_khz is 0. KVM uses host's tsc_khz in this case */
+    r = kvm_vm_ioctl(kvm_state, KVM_SET_TSC_KHZ, env->tsc_khz);
+    if (r < 0) {
+        error_report("Unable to set TSC frequency to %" PRId64 " kHz", env->tsc_khz);
+        goto out;
+    }

error_setg(errp, ....) in all of these cases.

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
author
#55 lore raw

PATCH v2 22/58 i386/tdx: Set kvm_readonly_mem_enabled to false for TDX VM

In reply to Xiaoyao Li

TDX only supports readonly for shared memory but not for private memory.

In the view of QEMU, it has no idea whether a memslot is used as shared memory of private. Thus just mark kvm_readonly_mem_enabled to false to TDX VM for simplicity.

---
 target/i386/kvm/tdx.c | 9 +++++++++
 1 file changed, 9 insertions(+)
target/i386/kvm/tdx.c +9 -0
index a72badfbfd65..8a2491ed03c2 100644
@@ -461,6 +461,15 @@ int tdx_kvm_init(MachineState *ms, Error **errp)
 
     update_tdx_cpuid_lookup_by_tdx_caps();
 
+    /*
+     * Set kvm_readonly_mem_allowed to false, because TDX only supports readonly
+     * memory for shared memory but not for private memory. Besides, whether a
+     * memslot is private or shared is not determined by QEMU.
+     *
+     * Thus, just mark readonly memory not supported for simplicity.
+     */
+    kvm_readonly_mem_allowed = false;
+
     tdx_guest = tdx;
 
     return 0;
Signature · 1 line
2.34.1
author
#56 lore raw

PATCH v2 23/58 i386/tdx: Make memory type private by default

In reply to Xiaoyao Li

By default (due to the recent UPM change), restricted memory attribute is shared. Convert the memory region from shared to private at the memory slot creation time.

add kvm region registering function to check the flag
and convert the region, and add memory listener to TDX guest code to set the flag to the possible memory region.

Without this patch
- Secure-EPT violation on private area
- KVM_MEMORY_FAULT EXIT (kvm -> qemu)
- qemu converts the 4K page from shared to private
- Resume VCPU execution
- Secure-EPT violation again
- KVM resolves EPT Violation
This also prevents huge page because page conversion is done at 4K granularity. Although it's possible to merge 4K private mapping into 2M large page, it slows guest boot.

With this patch
- After memory slot creation, convert the region from private to shared
- Secure-EPT violation on private area.
- KVM resolves EPT Violation

Originated-from: Isaku Yamahata <[email protected]>

---
 target/i386/kvm/tdx.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)
target/i386/kvm/tdx.c +22 -0
index 8a2491ed03c2..775110f8bd02 100644
@@ -18,6 +18,7 @@
 #include "standard-headers/asm-x86/kvm_para.h"
 #include "sysemu/kvm.h"
 #include "sysemu/sysemu.h"
+#include "exec/address-spaces.h"
 
 #include "hw/i386/x86.h"
 #include "kvm_i386.h"
@@ -578,6 +579,21 @@ out:
     return r;
 }
 
+static void tdx_guest_region_add(MemoryListener *listener,
+                                 MemoryRegionSection *section)
+{
+    if (memory_region_can_be_private(section->mr)) {
+        memory_region_set_default_private(section->mr);
+    }
+}
+
+static MemoryListener tdx_memory_listener = {
+    .name = TYPE_TDX_GUEST,
+    .region_add = tdx_guest_region_add,
+    /* Higher than KVM memory listener = 10. */
+    .priority = MEMORY_LISTENER_PRIORITY_ACCEL_HIGH,
+};
+
 static bool tdx_guest_get_sept_ve_disable(Object *obj, Error **errp)
 {
     TdxGuest *tdx = TDX_GUEST(obj);
@@ -607,6 +623,12 @@ OBJECT_DEFINE_TYPE_WITH_INTERFACES(TdxGuest,
 static void tdx_guest_init(Object *obj)
 {
     TdxGuest *tdx = TDX_GUEST(obj);
+    static bool memory_listener_registered = false;
+
+    if (!memory_listener_registered) {
+        memory_listener_register(&tdx_memory_listener, &address_space_memory);
+        memory_listener_registered = true;
+    }
 
     qemu_mutex_init(&tdx->lock);
 
Signature · 1 line
2.34.1
author
#57 lore raw

PATCH v2 24/58 i386/tdx: Create kvm gmem for TD

In reply to Xiaoyao Li

From: Isaku Yamahata <[email protected]>

Allocate private gmem for TD guest, if the MemoryRegion is memory backend and has private property on.

---
 target/i386/kvm/tdx.c | 27 +++++++++++++++++++++++++--
 1 file changed, 25 insertions(+), 2 deletions(-)
target/i386/kvm/tdx.c +25 -2
index 775110f8bd02..f1305191e939 100644
@@ -19,6 +19,7 @@
 #include "sysemu/kvm.h"
 #include "sysemu/sysemu.h"
 #include "exec/address-spaces.h"
+#include "exec/ramblock.h"
 
 #include "hw/i386/x86.h"
 #include "kvm_i386.h"
@@ -582,8 +583,30 @@ out:
 static void tdx_guest_region_add(MemoryListener *listener,
                                  MemoryRegionSection *section)
 {
-    if (memory_region_can_be_private(section->mr)) {
-        memory_region_set_default_private(section->mr);
+    MemoryRegion *mr = section->mr;
+    Object *owner = memory_region_owner(mr);
+
+    if (owner && object_dynamic_cast(owner, TYPE_MEMORY_BACKEND) &&
+        object_property_get_bool(owner, "private", NULL) &&
+        mr->ram_block && mr->ram_block->gmem_fd < 0) {
+        struct kvm_create_guest_memfd gmem = {
+            .size = memory_region_size(mr),
+            /* TODO: add property to hostmem backend for huge pmd */
+            .flags = KVM_GUEST_MEMFD_ALLOW_HUGEPAGE,
+        };
+        int fd;
+
+        fd = kvm_vm_ioctl(kvm_state, KVM_CREATE_GUEST_MEMFD, &gmem);
+        if (fd < 0) {
+            fprintf(stderr, "%s: error creating gmem: %s\n", __func__,
+                    strerror(-fd));
+            abort();
+        }
+        memory_region_set_gmem_fd(mr, fd);
+    }
+
+    if (memory_region_can_be_private(mr)) {
+        memory_region_set_default_private(mr);
     }
 }
 
Signature · 1 line
2.34.1
author
#58 lore raw

PATCH v2 25/58 kvm/tdx: Don't complain when converting vMMIO region to shared

In reply to Xiaoyao Li

From: Isaku Yamahata <[email protected]>

Because vMMIO region needs to be shared region, guest TD may explicitly convert such region from private to shared. Don't complain such conversion.

---
 accel/kvm/kvm-all.c | 20 ++++++++++++++++++--
 1 file changed, 18 insertions(+), 2 deletions(-)
accel/kvm/kvm-all.c +18 -2
index fceec7f2a83f..9d0aa8c97feb 100644
@@ -3094,8 +3094,24 @@ static int kvm_convert_memory(hwaddr start, hwaddr size, bool to_private)
          */
         ram_block_convert_range(rb, offset, size, to_private);
     } else {
-        warn_report("Convert non guest-memfd backed memory region (0x%"HWADDR_PRIx" ,+ 0x%"HWADDR_PRIx") to %s",
-                    start, size, to_private ? "private" : "shared");
+        MemoryRegion *mr = section.mr;
+
+        /*
+         * Because vMMIO region must be shared, guest TD may convert vMMIO
+         * region to shared explicitly.  Don't complain such case.  See
+         * memory_region_type() for checking if the region is MMIO region.
+         */
+        if (to_private ||
+            memory_region_is_ram(mr) ||
+            memory_region_is_ram_device(mr) ||
+            memory_region_is_rom(mr) ||
+            memory_region_is_romd(mr)) {
+            warn_report("Convert non guest-memfd backed memory region (0x%"HWADDR_PRIx" ,+ 0x%"HWADDR_PRIx") of %s to %s",
+                        start, size, mr->name, to_private ? "private" : "shared");
+	    } else {
+		    ret = 0;
+	    }
+
     }
 
     memory_region_unref(section.mr);
Signature · 1 line
2.34.1
#59 lore raw

Re: [PATCH v2 25/58] kvm/tdx: Don't complain when converting vMMIO region to shared

Quoted 24 lines · Xiaoyao Li

From: Isaku Yamahata <[email protected]>

Because vMMIO region needs to be shared region, guest TD may explicitly convert such region from private to shared. Don't complain such conversion.

---
 accel/kvm/kvm-all.c | 20 ++++++++++++++++++--
 1 file changed, 18 insertions(+), 2 deletions(-)
accel/kvm/kvm-all.c +2 -2
index fceec7f2a83f..9d0aa8c97feb 100644
@@ -3094,8 +3094,24 @@ static int kvm_convert_memory(hwaddr start, hwaddr size, bool to_private)
          */
         ram_block_convert_range(rb, offset, size, to_private);
     } else {
-        warn_report("Convert non guest-memfd backed memory region (0x%"HWADDR_PRIx" ,+ 0x%"HWADDR_PRIx") to %s",
-                    start, size, to_private ? "private" : "shared");
+        MemoryRegion *mr = section.mr;
+

accel/kvm/kvm-all.c static int kvm_convert_memory(hwaddr start, hwaddr size, bool to_private) · line 3108

Quoted 10 lines
accel/kvm/kvm-all.c +10 -0
+        /*
+         * Because vMMIO region must be shared, guest TD may convert vMMIO
+         * region to shared explicitly.  Don't complain such case.  See
+         * memory_region_type() for checking if the region is MMIO region.
+         */
+        if (to_private ||
+            memory_region_is_ram(mr) ||
+            memory_region_is_ram_device(mr) ||
+            memory_region_is_rom(mr) ||
+            memory_region_is_romd(mr)) {

Should we also have !is_tdx_vm() as the first clause to check, as IIUC you only need the special logic for TDX VMs.

accel/kvm/kvm-all.c static int kvm_convert_memory(hwaddr start, hwaddr size, bool to_private) · line 3113

Quoted 5 lines
accel/kvm/kvm-all.c +5 -0
+            warn_report("Convert non guest-memfd backed memory region (0x%"HWADDR_PRIx" ,+ 0x%"HWADDR_PRIx") of %s to %s",
+                        start, size, mr->name, to_private ? "private" : "shared");
+	    } else {
+		    ret = 0;
+	    }

Inconsistent indentation here due to use of tabs

Quoted 7 lines
accel/kvm/kvm-all.c +1 -0
+
     }
 
     memory_region_unref(section.mr);
Signature · 1 line
2.34.1

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
author
#60 lore raw

PATCH v2 26/58 kvm/tdx: Ignore memory conversion to shared of unassigned region

In reply to Xiaoyao Li

From: Isaku Yamahata <[email protected]>

TDX requires vMMIO region to be shared. For KVM, MMIO region is the region which kvm memslot isn't assigned to (except in-kernel emulation). qemu has the memory region for vMMIO at each device level.

While OVMF issues MapGPA(to-shared) conservatively on 32bit PCI MMIO region, qemu doesn't find corresponding vMMIO region because it's before PCI device allocation and memory_region_find() finds the device region, not PCI bus region. It's safe to ignore MapGPA(to-shared) because when guest accesses those region they use GPA with shared bit set for vMMIO. Ignore memory conversion request of non-assigned region to shared and return success. Otherwise OVMF is confused and panics there.

---
 accel/kvm/kvm-all.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)
accel/kvm/kvm-all.c +12 -0
index 9d0aa8c97feb..8d53c89e9dbf 100644
@@ -3070,6 +3070,18 @@ static int kvm_convert_memory(hwaddr start, hwaddr size, bool to_private)
     trace_kvm_convert_memory(start, size, to_private ? "shared_to_private" : "private_to_shared");
     section = memory_region_find(get_system_memory(), start, size);
     if (!section.mr) {
+        /*
+         * Ignore converting non-assigned region to shared.
+         *
+         * TDX requires vMMIO region to be shared to inject #VE to guest.
+         * OVMF issues conservatively MapGPA(shared) on 32bit PCI MMIO region,
+         * and vIO-APIC 0xFEC00000 4K page.
+         * OVMF assigns 32bit PCI MMIO region to
+         * [top of low memory: typically 2GB=0xC000000,  0xFC00000)
+         */
+        if (!to_private) {
+            ret = 0;
+        }
         return ret;
     }
 
Signature · 1 line
2.34.1
author
#61 lore raw

PATCH v2 27/58 i386/tdvf: Introduce function to parse TDVF metadata

In reply to Xiaoyao Li

From: Isaku Yamahata <[email protected]>

TDX VM needs to boot with its specialized firmware, Trusted Domain Virtual Firmware (TDVF). QEMU needs to parse TDVF and map it in TD guest memory prior to running the TDX VM.

A TDVF Metadata in TDVF image describes the structure of firmware. QEMU refers to it to setup memory for TDVF. Introduce function tdvf_parse_metadata() to parse the metadata from TDVF image and store the info of each TDVF section.

TDX metadata is located by a TDX metadata offset block, which is a GUID-ed structure. The data portion of the GUID structure contains only an 4-byte field that is the offset of TDX metadata to the end of firmware file.

Select X86_FW_OVMF when TDX is enable to leverage existing functions to parse and search OVMF's GUID-ed structures.

---
Changes from RFC v4:
- rename tdvf_parse_section_entry() to tdvf_parse_and_check_section_entry()
Changes in v4:
- rename TDX_METADATA_GUID to TDX_METADATA_OFFSET_GUID

---
 hw/i386/Kconfig        |   1 +
 hw/i386/meson.build    |   1 +
 hw/i386/tdvf.c         | 199 +++++++++++++++++++++++++++++++++++++++++
 include/hw/i386/tdvf.h |  51 +++++++++++
 4 files changed, 252 insertions(+)
 create mode 100644 hw/i386/tdvf.c
 create mode 100644 include/hw/i386/tdvf.h
hw/i386/Kconfig +1 -0
index 929f6c3f0e85..6007bdef184d 100644
@@ -12,6 +12,7 @@ config SGX
 
 config TDX
     bool
+    select X86_FW_OVMF
     depends on KVM
 
 config PC
hw/i386/meson.build +1 -0
index cfdbfdcbcb2d..45d90bb2af52 100644
@@ -28,6 +28,7 @@ i386_ss.add(when: 'CONFIG_PC', if_true: files(
   'port92.c'))
 i386_ss.add(when: 'CONFIG_X86_FW_OVMF', if_true: files('pc_sysfw_ovmf.c'),
                                         if_false: files('pc_sysfw_ovmf-stubs.c'))
+i386_ss.add(when: 'CONFIG_TDX', if_true: files('tdvf.c'))
 
 subdir('kvm')
 subdir('xen')
hw/i386/tdvf.c new file +199 -0

199 lines not on this page · raw message

include/hw/i386/tdvf.h new file +51 -0

51 lines not on this page · raw message

Signature · 1 line
2.34.1
author
#62 lore raw

PATCH v2 28/58 i386/tdx: Parse TDVF metadata for TDX VM

In reply to Xiaoyao Li

TDX cannot support pflash device since it doesn't support read-only memslot and doesn't support emulation. Load TDVF(OVMF) with -bios option for TDs.

When boot a TD, besides load TDVF to the address below 4G, it needs parse TDVF metadata.

---
 hw/i386/pc_sysfw.c         | 7 +++++++
 hw/i386/x86.c              | 3 ++-
 target/i386/kvm/tdx-stub.c | 5 +++++
 target/i386/kvm/tdx.c      | 5 +++++
 target/i386/kvm/tdx.h      | 4 ++++
 5 files changed, 23 insertions(+), 1 deletion(-)
hw/i386/pc_sysfw.c +7 -0
index c8d9e71b889b..cf63434ba89d 100644
@@ -37,6 +37,7 @@
 #include "hw/block/flash.h"
 #include "sysemu/kvm.h"
 #include "sev.h"
+#include "kvm/tdx.h"
 
 #define FLASH_SECTOR_SIZE 4096
 
@@ -265,5 +266,11 @@ void x86_firmware_configure(void *ptr, int size)
         }
 
         sev_encrypt_flash(ptr, size, &error_fatal);
+    } else if (is_tdx_vm()) {
+        ret = tdx_parse_tdvf(ptr, size);
+        if (ret) {
+            error_report("failed to parse TDVF for TDX VM");
+            exit(1);
+        }
     }
 }
hw/i386/x86.c +2 -1
index 3ccd06154249..dabd33cb830b 100644
@@ -47,6 +47,7 @@
 #include "hw/intc/i8259.h"
 #include "hw/rtc/mc146818rtc.h"
 #include "target/i386/sev.h"
+#include "kvm/tdx.h"
 
 #include "hw/acpi/cpu_hotplug.h"
 #include "hw/irq.h"
@@ -1152,7 +1153,7 @@ void x86_bios_rom_init(MachineState *ms, const char *default_firmware,
     }
     bios = g_malloc(sizeof(*bios));
     memory_region_init_ram(bios, NULL, "pc.bios", bios_size, &error_fatal);
-    if (sev_enabled()) {
+    if (sev_enabled() || is_tdx_vm()) {
         /*
          * The concept of a "reset" simply doesn't exist for
          * confidential computing guests, we have to destroy and
target/i386/kvm/tdx-stub.c +5 -0
index 61f70cc0d1d9..3cd114476d78 100644
@@ -11,3 +11,8 @@ int tdx_pre_create_vcpu(CPUState *cpu)
 {
     return -EINVAL;
 }
+
+int tdx_parse_tdvf(void *flash_ptr, int size)
+{
+    return -EINVAL;
+}
target/i386/kvm/tdx.c +5 -0
index f1305191e939..e41ea18c04db 100644
@@ -580,6 +580,11 @@ out:
     return r;
 }
 
+int tdx_parse_tdvf(void *flash_ptr, int size)
+{
+    return tdvf_parse_metadata(&tdx_guest->tdvf, flash_ptr, size);
+}
+
 static void tdx_guest_region_add(MemoryListener *listener,
                                  MemoryRegionSection *section)
 {
target/i386/kvm/tdx.h +4 -0
index 68f8327f2231..e9d2888162ce 100644
@@ -6,6 +6,7 @@
 #endif
 
 #include "exec/confidential-guest-support.h"
+#include "hw/i386/tdvf.h"
 
 #define TYPE_TDX_GUEST "tdx-guest"
 #define TDX_GUEST(obj)  OBJECT_CHECK(TdxGuest, (obj), TYPE_TDX_GUEST)
@@ -24,6 +25,8 @@ typedef struct TdxGuest {
     uint8_t mrconfigid[48];     /* sha348 digest */
     uint8_t mrowner[48];        /* sha348 digest */
     uint8_t mrownerconfig[48];  /* sha348 digest */
+
+    TdxFirmware tdvf;
 } TdxGuest;
 
 #ifdef CONFIG_TDX
@@ -36,5 +39,6 @@ int tdx_kvm_init(MachineState *ms, Error **errp);
 void tdx_get_supported_cpuid(uint32_t function, uint32_t index, int reg,
                              uint32_t *ret);
 int tdx_pre_create_vcpu(CPUState *cpu);
+int tdx_parse_tdvf(void *flash_ptr, int size);
 
 #endif /* QEMU_I386_TDX_H */
Signature · 1 line
2.34.1
author
#63 lore raw

PATCH v2 29/58 i386/tdx: Skip BIOS shadowing setup

In reply to Xiaoyao Li

TDX doesn't support map different GPAs to same private memory. Thus, aliasing top 128KB of BIOS as isa-bios is not supported.

On the other hand, TDX guest cannot go to real mode, it can work fine without isa-bios.

---
Changes from RFC v4:
- update commit message and comment to clarify

---
 hw/i386/x86.c | 25 ++++++++++++++-----------
 1 file changed, 14 insertions(+), 11 deletions(-)
hw/i386/x86.c +14 -11
index dabd33cb830b..e2a1369e5dfc 100644
@@ -1175,17 +1175,20 @@ void x86_bios_rom_init(MachineState *ms, const char *default_firmware,
     }
     g_free(filename);
 
-    /* map the last 128KB of the BIOS in ISA space */
-    isa_bios_size = MIN(bios_size, 128 * KiB);
-    isa_bios = g_malloc(sizeof(*isa_bios));
-    memory_region_init_alias(isa_bios, NULL, "isa-bios", bios,
-                             bios_size - isa_bios_size, isa_bios_size);
-    memory_region_add_subregion_overlap(rom_memory,
-                                        0x100000 - isa_bios_size,
-                                        isa_bios,
-                                        1);
-    if (!isapc_ram_fw) {
-        memory_region_set_readonly(isa_bios, true);
+    /* For TDX, alias different GPAs to same private memory is not supported */
+    if (!is_tdx_vm()) {
+        /* map the last 128KB of the BIOS in ISA space */
+        isa_bios_size = MIN(bios_size, 128 * KiB);
+        isa_bios = g_malloc(sizeof(*isa_bios));
+        memory_region_init_alias(isa_bios, NULL, "isa-bios", bios,
+                                bios_size - isa_bios_size, isa_bios_size);
+        memory_region_add_subregion_overlap(rom_memory,
+                                            0x100000 - isa_bios_size,
+                                            isa_bios,
+                                            1);
+        if (!isapc_ram_fw) {
+            memory_region_set_readonly(isa_bios, true);
+        }
     }
 
     /* map all the bios at the top of memory */
Signature · 1 line
2.34.1
author
#64 lore raw

PATCH v2 30/58 i386/tdx: Don't initialize pc.rom for TDX VMs

In reply to Xiaoyao Li

For TDX, the address below 1MB are entirely general RAM. No need to initialize pc.rom memory region for TDs.

---
This is more as a workaround of the issue that for q35 machine type, the real memslot update (which requires memslot deletion )for pc.rom happens after tdx_init_memory_region. It leads to the private memory ADD'ed before get lost. I haven't work out a good solution to resolve the order issue. So just skip the pc.rom setup to avoid memslot deletion.

---
 hw/i386/pc.c | 21 ++++++++++++---------
 1 file changed, 12 insertions(+), 9 deletions(-)
hw/i386/pc.c +12 -9
index abeadd903827..bc307fed0f44 100644
@@ -62,6 +62,7 @@
 #include "sysemu/reset.h"
 #include "sysemu/runstate.h"
 #include "kvm/kvm_i386.h"
+#include "kvm/tdx.h"
 #include "hw/xen/xen.h"
 #include "hw/xen/start_info.h"
 #include "ui/qemu-spice.h"
@@ -1095,16 +1096,18 @@ void pc_memory_init(PCMachineState *pcms,
     /* Initialize PC system firmware */
     pc_system_firmware_init(pcms, rom_memory);
 
-    option_rom_mr = g_malloc(sizeof(*option_rom_mr));
-    memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
-                           &error_fatal);
-    if (pcmc->pci_enabled) {
-        memory_region_set_readonly(option_rom_mr, true);
+    if (!is_tdx_vm()) {
+        option_rom_mr = g_malloc(sizeof(*option_rom_mr));
+        memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
+                            &error_fatal);
+        if (pcmc->pci_enabled) {
+            memory_region_set_readonly(option_rom_mr, true);
+        }
+        memory_region_add_subregion_overlap(rom_memory,
+                                            PC_ROM_MIN_VGA,
+                                            option_rom_mr,
+                                            1);
     }
-    memory_region_add_subregion_overlap(rom_memory,
-                                        PC_ROM_MIN_VGA,
-                                        option_rom_mr,
-                                        1);
 
     fw_cfg = fw_cfg_arch_create(machine,
                                 x86ms->boot_cpus, x86ms->apic_id_limit);
Signature · 1 line
2.34.1
author
#65 lore raw

PATCH v2 31/58 i386/tdx: Track mem_ptr for each firmware entry of TDVF

In reply to Xiaoyao Li

For each TDVF sections, QEMU needs to copy the content to guest private memory via KVM API (KVM_TDX_INIT_MEM_REGION).

Introduce a field @mem_ptr for TdxFirmwareEntry to track the memory pointer of each TDVF sections. So that QEMU can add/copy them to guest private memory later.

TDVF sections can be classified into two groups:
- Firmware itself, e.g., TDVF BFV and CFV, that located separately from guest RAM. Its memory pointer is the bios pointer.

- Sections located at guest RAM, e.g., TEMP_MEM and TD_HOB. mmap a new memory range for them.

Register a machine_init_done callback to do the stuff.

---
 hw/i386/tdvf.c         |  1 +
 include/hw/i386/tdvf.h |  7 +++++++
 target/i386/kvm/tdx.c  | 31 +++++++++++++++++++++++++++++++
 3 files changed, 39 insertions(+)
hw/i386/tdvf.c +1 -0
index ff51f40088f0..0a6445705160 100644
@@ -189,6 +189,7 @@ int tdvf_parse_metadata(TdxFirmware *fw, void *flash_ptr, int size)
     }
     g_free(sections);
 
+    fw->mem_ptr = flash_ptr;
     return 0;
 
 err:
include/hw/i386/tdvf.h +7 -0
index 593341eb2e93..d880af245a73 100644
@@ -39,13 +39,20 @@ typedef struct TdxFirmwareEntry {
     uint64_t size;
     uint32_t type;
     uint32_t attributes;
+
+    void *mem_ptr;
 } TdxFirmwareEntry;
 
 typedef struct TdxFirmware {
+    void *mem_ptr;
+
     uint32_t nr_entries;
     TdxFirmwareEntry *entries;
 } TdxFirmware;
 
+#define for_each_tdx_fw_entry(fw, e)    \
+    for (e = (fw)->entries; e != (fw)->entries + (fw)->nr_entries; e++)
+
 int tdvf_parse_metadata(TdxFirmware *fw, void *flash_ptr, int size);
 
 #endif /* HW_I386_TDVF_H */
target/i386/kvm/tdx.c +31 -0
index e41ea18c04db..bb806736b4ff 100644
@@ -13,6 +13,7 @@
 
 #include "qemu/osdep.h"
 #include "qemu/error-report.h"
+#include "qemu/mmap-alloc.h"
 #include "qapi/error.h"
 #include "qom/object_interfaces.h"
 #include "standard-headers/asm-x86/kvm_para.h"
@@ -22,6 +23,7 @@
 #include "exec/ramblock.h"
 
 #include "hw/i386/x86.h"
+#include "hw/i386/tdvf.h"
 #include "kvm_i386.h"
 #include "tdx.h"
 #include "../cpu-internal.h"
@@ -452,6 +454,33 @@ static void update_tdx_cpuid_lookup_by_tdx_caps(void)
             (tdx_caps->xfam_fixed1 & CPUID_XSTATE_XSS_MASK) >> 32;
 }
 
+static void tdx_finalize_vm(Notifier *notifier, void *unused)
+{
+    TdxFirmware *tdvf = &tdx_guest->tdvf;
+    TdxFirmwareEntry *entry;
+
+    for_each_tdx_fw_entry(tdvf, entry) {
+        switch (entry->type) {
+        case TDVF_SECTION_TYPE_BFV:
+        case TDVF_SECTION_TYPE_CFV:
+            entry->mem_ptr = tdvf->mem_ptr + entry->data_offset;
+            break;
+        case TDVF_SECTION_TYPE_TD_HOB:
+        case TDVF_SECTION_TYPE_TEMP_MEM:
+            entry->mem_ptr = qemu_ram_mmap(-1, entry->size,
+                                           qemu_real_host_page_size(), 0, 0);
+            break;
+        default:
+            error_report("Unsupported TDVF section %d", entry->type);
+            exit(1);
+        }
+    }
+}
+
+static Notifier tdx_machine_done_notify = {
+    .notify = tdx_finalize_vm,
+};
+
 int tdx_kvm_init(MachineState *ms, Error **errp)
 {
     TdxGuest *tdx = (TdxGuest *)object_dynamic_cast(OBJECT(ms->cgs),
@@ -472,6 +501,8 @@ int tdx_kvm_init(MachineState *ms, Error **errp)
      */
     kvm_readonly_mem_allowed = false;
 
+    qemu_add_machine_init_done_notifier(&tdx_machine_done_notify);
+
     tdx_guest = tdx;
 
     return 0;
Signature · 1 line
2.34.1
author
#66 lore raw

PATCH v2 32/58 i386/tdx: Track RAM entries for TDX VM

In reply to Xiaoyao Li

The RAM of TDX VM can be classified into two types:

- TDX_RAM_UNACCEPTED: default type of TDX memory, which needs to be accepted by TDX guest before it can be used and will be all-zeros after being accepted.

- TDX_RAM_ADDED: the RAM that is ADD'ed to TD guest before running, and can be used directly. E.g., TD HOB and TEMP MEM that needed by TDVF.

Maintain TdxRamEntries[] which grabs the initial RAM info from e820 table and mark each RAM range as default type TDX_RAM_UNACCEPTED.

Then turn the range of TD HOB and TEMP MEM to TDX_RAM_ADDED since these ranges will be ADD'ed before TD runs and no need to be accepted runtime.

The TdxRamEntries[] are later used to setup the memory TD resource HOB that passes memory info from QEMU to TDVF.

---
Changes from RFC v4:
- simplify the algorithm of tdx_accept_ram_range() (Suggested-by: Gerd Hoffman)
(1) Change the existing entry to cover the accepted ram range.
(2) If there is room before the accepted ram range add a TDX_RAM_UNACCEPTED entry for that.
(3) If there is room after the accepted ram range add a TDX_RAM_UNACCEPTED entry for that.

---
 target/i386/kvm/tdx.c | 110 ++++++++++++++++++++++++++++++++++++++++++
 target/i386/kvm/tdx.h |  14 ++++++
 2 files changed, 124 insertions(+)
target/i386/kvm/tdx.c +110 -0
index bb806736b4ff..ed617ebab266 100644
@@ -22,6 +22,7 @@
 #include "exec/address-spaces.h"
 #include "exec/ramblock.h"
 
+#include "hw/i386/e820_memory_layout.h"
 #include "hw/i386/x86.h"
 #include "hw/i386/tdvf.h"
 #include "kvm_i386.h"
@@ -454,11 +455,116 @@ static void update_tdx_cpuid_lookup_by_tdx_caps(void)
             (tdx_caps->xfam_fixed1 & CPUID_XSTATE_XSS_MASK) >> 32;
 }
 
+static void tdx_add_ram_entry(uint64_t address, uint64_t length, uint32_t type)
+{
+    uint32_t nr_entries = tdx_guest->nr_ram_entries;
+    tdx_guest->ram_entries = g_renew(TdxRamEntry, tdx_guest->ram_entries,
+                                     nr_entries + 1);
+
+    tdx_guest->ram_entries[nr_entries].address = address;
+    tdx_guest->ram_entries[nr_entries].length = length;
+    tdx_guest->ram_entries[nr_entries].type = type;
+    tdx_guest->nr_ram_entries++;
+}
+
+static int tdx_accept_ram_range(uint64_t address, uint64_t length)
+{
+    uint64_t head_start, tail_start, head_length, tail_length;
+    uint64_t tmp_address, tmp_length;
+    TdxRamEntry *e;
+    int i;
+
+    for (i = 0; i < tdx_guest->nr_ram_entries; i++) {
+        e = &tdx_guest->ram_entries[i];
+
+        if (address + length <= e->address ||
+            e->address + e->length <= address) {
+                continue;
+        }
+
+        /*
+         * The to-be-accepted ram range must be fully contained by one
+         * RAM entry.
+         */
+        if (e->address > address ||
+            e->address + e->length < address + length) {
+            return -EINVAL;
+        }
+
+        if (e->type == TDX_RAM_ADDED) {
+            return -EINVAL;
+        }
+
+        break;
+    }
+
+    if (i == tdx_guest->nr_ram_entries) {
+        return -1;
+    }
+
+    tmp_address = e->address;
+    tmp_length = e->length;
+
+    e->address = address;
+    e->length = length;
+    e->type = TDX_RAM_ADDED;
+
+    head_length = address - tmp_address;
+    if (head_length > 0) {
+        head_start = tmp_address;
+        tdx_add_ram_entry(head_start, head_length, TDX_RAM_UNACCEPTED);
+    }
+
+    tail_start = address + length;
+    if (tail_start < tmp_address + tmp_length) {
+        tail_length = tmp_address + tmp_length - tail_start;
+        tdx_add_ram_entry(tail_start, tail_length, TDX_RAM_UNACCEPTED);
+    }
+
+    return 0;
+}
+
+static int tdx_ram_entry_compare(const void *lhs_, const void* rhs_)
+{
+    const TdxRamEntry *lhs = lhs_;
+    const TdxRamEntry *rhs = rhs_;
+
+    if (lhs->address == rhs->address) {
+        return 0;
+    }
+    if (le64_to_cpu(lhs->address) > le64_to_cpu(rhs->address)) {
+        return 1;
+    }
+    return -1;
+}
+
+static void tdx_init_ram_entries(void)
+{
+    unsigned i, j, nr_e820_entries;
+
+    nr_e820_entries = e820_get_num_entries();
+    tdx_guest->ram_entries = g_new(TdxRamEntry, nr_e820_entries);
+
+    for (i = 0, j = 0; i < nr_e820_entries; i++) {
+        uint64_t addr, len;
+
+        if (e820_get_entry(i, E820_RAM, &addr, &len)) {
+            tdx_guest->ram_entries[j].address = addr;
+            tdx_guest->ram_entries[j].length = len;
+            tdx_guest->ram_entries[j].type = TDX_RAM_UNACCEPTED;
+            j++;
+        }
+    }
+    tdx_guest->nr_ram_entries = j;
+}
+
 static void tdx_finalize_vm(Notifier *notifier, void *unused)
 {
     TdxFirmware *tdvf = &tdx_guest->tdvf;
     TdxFirmwareEntry *entry;
 
+    tdx_init_ram_entries();
+
     for_each_tdx_fw_entry(tdvf, entry) {
         switch (entry->type) {
         case TDVF_SECTION_TYPE_BFV:
@@ -469,12 +575,16 @@ static void tdx_finalize_vm(Notifier *notifier, void *unused)
         case TDVF_SECTION_TYPE_TEMP_MEM:
             entry->mem_ptr = qemu_ram_mmap(-1, entry->size,
                                            qemu_real_host_page_size(), 0, 0);
+            tdx_accept_ram_range(entry->address, entry->size);
             break;
         default:
             error_report("Unsupported TDVF section %d", entry->type);
             exit(1);
         }
     }
+
+    qsort(tdx_guest->ram_entries, tdx_guest->nr_ram_entries,
+          sizeof(TdxRamEntry), &tdx_ram_entry_compare);
 }
 
 static Notifier tdx_machine_done_notify = {
target/i386/kvm/tdx.h +14 -0
index e9d2888162ce..9b3c427766ef 100644
@@ -15,6 +15,17 @@ typedef struct TdxGuestClass {
     ConfidentialGuestSupportClass parent_class;
 } TdxGuestClass;
 
+enum TdxRamType{
+    TDX_RAM_UNACCEPTED,
+    TDX_RAM_ADDED,
+};
+
+typedef struct TdxRamEntry {
+    uint64_t address;
+    uint64_t length;
+    uint32_t type;
+} TdxRamEntry;
+
 typedef struct TdxGuest {
     ConfidentialGuestSupport parent_obj;
 
@@ -27,6 +38,9 @@ typedef struct TdxGuest {
     uint8_t mrownerconfig[48];  /* sha348 digest */
 
     TdxFirmware tdvf;
+
+    uint32_t nr_ram_entries;
+    TdxRamEntry *ram_entries;
 } TdxGuest;
 
 #ifdef CONFIG_TDX
Signature · 1 line
2.34.1
#67 lore raw

Re: [PATCH v2 32/58] i386/tdx: Track RAM entries for TDX VM

Quoted 41 lines · Xiaoyao Li

The RAM of TDX VM can be classified into two types:

- TDX_RAM_UNACCEPTED: default type of TDX memory, which needs to be accepted by TDX guest before it can be used and will be all-zeros after being accepted.

- TDX_RAM_ADDED: the RAM that is ADD'ed to TD guest before running, and can be used directly. E.g., TD HOB and TEMP MEM that needed by TDVF.

Maintain TdxRamEntries[] which grabs the initial RAM info from e820 table and mark each RAM range as default type TDX_RAM_UNACCEPTED.

Then turn the range of TD HOB and TEMP MEM to TDX_RAM_ADDED since these ranges will be ADD'ed before TD runs and no need to be accepted runtime.

The TdxRamEntries[] are later used to setup the memory TD resource HOB that passes memory info from QEMU to TDVF.

---
Changes from RFC v4:
- simplify the algorithm of tdx_accept_ram_range() (Suggested-by: Gerd Hoffman)
(1) Change the existing entry to cover the accepted ram range.
(2) If there is room before the accepted ram range add a TDX_RAM_UNACCEPTED entry for that.
(3) If there is room after the accepted ram range add a TDX_RAM_UNACCEPTED entry for that.

---
 target/i386/kvm/tdx.c | 110 ++++++++++++++++++++++++++++++++++++++++++
 target/i386/kvm/tdx.h |  14 ++++++
 2 files changed, 124 insertions(+)
target/i386/kvm/tdx.c +3 -0
index bb806736b4ff..ed617ebab266 100644
+static int tdx_accept_ram_range(uint64_t address, uint64_t length)
+{
+    uint64_t head_start, tail_start, head_length, tail_length;

target/i386/kvm/tdx.c static void update_tdx_cpuid_lookup_by_tdx_caps(void) · line 482

Quoted 10 lines
target/i386/kvm/tdx.c +10 -0
+    uint64_t tmp_address, tmp_length;
+    TdxRamEntry *e;
+    int i;
+
+    for (i = 0; i < tdx_guest->nr_ram_entries; i++) {
+        e = &tdx_guest->ram_entries[i];
+
+        if (address + length <= e->address ||
+            e->address + e->length <= address) {
+                continue;

Indented too far

Quoted 33 lines
target/i386/kvm/tdx.c +33 -0
+        }
+
+        /*
+         * The to-be-accepted ram range must be fully contained by one
+         * RAM entry.
+         */
+        if (e->address > address ||
+            e->address + e->length < address + length) {
+            return -EINVAL;
+        }
+
+        if (e->type == TDX_RAM_ADDED) {
+            return -EINVAL;
+        }
+
+        break;
+    }
+
+    if (i == tdx_guest->nr_ram_entries) {
+        return -1;
+    }
+
+    tmp_address = e->address;
+    tmp_length = e->length;
+
+    e->address = address;
+    e->length = length;
+    e->type = TDX_RAM_ADDED;
+
+    head_length = address - tmp_address;
+    if (head_length > 0) {
+        head_start = tmp_address;
+        tdx_add_ram_entry(head_start, head_length, TDX_RAM_UNACCEPTED);

target/i386/kvm/tdx.c static void update_tdx_cpuid_lookup_by_tdx_caps(void) · line 525

Quoted 10 lines
target/i386/kvm/tdx.c +10 -0
+    }
+
+    tail_start = address + length;
+    if (tail_start < tmp_address + tmp_length) {
+        tail_length = tmp_address + tmp_length - tail_start;
+        tdx_add_ram_entry(tail_start, tail_length, TDX_RAM_UNACCEPTED);
+    }
+
+    return 0;
+}

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
author
#68 lore raw

Re: [PATCH v2 32/58] i386/tdx: Track RAM entries for TDX VM

Quoted 10 lines · Daniel P. Berrangé
Quoted 10 lines
target/i386/kvm/tdx.c +6 -0
index bb806736b4ff..ed617ebab266 100644
+static int tdx_accept_ram_range(uint64_t address, uint64_t length)
+{
+    uint64_t head_start, tail_start, head_length, tail_length;
+    uint64_t tmp_address, tmp_length;
+    TdxRamEntry *e;
+    int i;
Quoted 9 lines

target/i386/kvm/tdx.c static void update_tdx_cpuid_lookup_by_tdx_caps(void) · line 482

Quoted 7 lines
target/i386/kvm/tdx.c +7 -0
+
+    for (i = 0; i < tdx_guest->nr_ram_entries; i++) {
+        e = &tdx_guest->ram_entries[i];
+
+        if (address + length <= e->address ||
+            e->address + e->length <= address) {
+                continue;

Indented too far

Fixed.

Thanks!
-Xiaoyao

#69 lore raw

Re: [PATCH v2 32/58] i386/tdx: Track RAM entries for TDX VM

In reply to Xiaoyao Li
Quoted 7 lines · Xiaoyao Li
target/i386/kvm/tdx.h +0 -0
index e9d2888162ce..9b3c427766ef 100644
@@ -15,6 +15,17 @@ typedef struct TdxGuestClass {
     ConfidentialGuestSupportClass parent_class;
 } TdxGuestClass;

target/i386/kvm/tdx.h typedef struct TdxGuestClass { · line 26

Quoted 10 lines
target/i386/kvm/tdx.h +9 -0
 
+enum TdxRamType{
+    TDX_RAM_UNACCEPTED,
+    TDX_RAM_ADDED,
+};
+
+typedef struct TdxRamEntry {
+    uint64_t address;
+    uint64_t length;
+    uint32_t type;

nitpick: enum TdxRamType. and related function arguments.

Signature · 1 line
Isaku Yamahata <[email protected]>
author
#70 lore raw

Re: [PATCH v2 32/58] i386/tdx: Track RAM entries for TDX VM

Quoted 11 lines · Isaku Yamahata
Quoted 8 lines · Xiaoyao Li
target/i386/kvm/tdx.h +0 -0
index e9d2888162ce..9b3c427766ef 100644
@@ -15,6 +15,17 @@ typedef struct TdxGuestClass {
      ConfidentialGuestSupportClass parent_class;
  } TdxGuestClass;
 
Quoted 12 lines

target/i386/kvm/tdx.h typedef struct TdxGuestClass { · line 26

Quoted 9 lines
target/i386/kvm/tdx.h +9 -0
+enum TdxRamType{
+    TDX_RAM_UNACCEPTED,
+    TDX_RAM_ADDED,
+};
+
+typedef struct TdxRamEntry {
+    uint64_t address;
+    uint64_t length;
+    uint32_t type;

nitpick: enum TdxRamType. and related function arguments.

Will do it.

Thanks!

author
#71 lore raw

PATCH v2 33/58 headers: Add definitions from UEFI spec for volumes, resources, etc...

In reply to Xiaoyao Li

Add UEFI definitions for literals, enums, structs, GUIDs, etc... that will be used by TDX to build the UEFI Hand-Off Block (HOB) that is passed to the Trusted Domain Virtual Firmware (TDVF).

All values come from the UEFI specification and TDVF design guide. [1]

Note, EFI_RESOURCE_MEMORY_UNACCEPTED will be added in future UEFI spec.

[1] https://software.intel.com/content/dam/develop/external/us/en/documents/tdx-virtual-firmware-design-guide-rev-1.pdf

---
 include/standard-headers/uefi/uefi.h | 198 +++++++++++++++++++++++++++
 1 file changed, 198 insertions(+)
 create mode 100644 include/standard-headers/uefi/uefi.h
include/standard-headers/uefi/uefi.h new file +198 -0
new file mode 100644
index 000000000000..b15aba796156
@@ -0,0 +1,198 @@
+/*
+ * Copyright (C) 2020 Intel Corporation
+ *
+ * Author: Isaku Yamahata <isaku.yamahata at gmail.com>
+ *                        <isaku.yamahata at intel.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, see <http://www.gnu.org/licenses/>.
+ *
+ */
+
+#ifndef HW_I386_UEFI_H
+#define HW_I386_UEFI_H
+
+/***************************************************************************/
+/*
+ * basic EFI definitions
+ * supplemented with UEFI Specification Version 2.8 (Errata A)
+ * released February 2020
+ */
+/* UEFI integer is little endian */
+
+typedef struct {
+    uint32_t Data1;
+    uint16_t Data2;
+    uint16_t Data3;
+    uint8_t Data4[8];
+} EFI_GUID;
+
+typedef enum {
+    EfiReservedMemoryType,
+    EfiLoaderCode,
+    EfiLoaderData,
+    EfiBootServicesCode,
+    EfiBootServicesData,
+    EfiRuntimeServicesCode,
+    EfiRuntimeServicesData,
+    EfiConventionalMemory,
+    EfiUnusableMemory,
+    EfiACPIReclaimMemory,
+    EfiACPIMemoryNVS,
+    EfiMemoryMappedIO,
+    EfiMemoryMappedIOPortSpace,
+    EfiPalCode,
+    EfiPersistentMemory,
+    EfiUnacceptedMemoryType,
+    EfiMaxMemoryType
+} EFI_MEMORY_TYPE;
+
+#define EFI_HOB_HANDOFF_TABLE_VERSION 0x0009
+
+#define EFI_HOB_TYPE_HANDOFF              0x0001
+#define EFI_HOB_TYPE_MEMORY_ALLOCATION    0x0002
+#define EFI_HOB_TYPE_RESOURCE_DESCRIPTOR  0x0003
+#define EFI_HOB_TYPE_GUID_EXTENSION       0x0004
+#define EFI_HOB_TYPE_FV                   0x0005
+#define EFI_HOB_TYPE_CPU                  0x0006
+#define EFI_HOB_TYPE_MEMORY_POOL          0x0007
+#define EFI_HOB_TYPE_FV2                  0x0009
+#define EFI_HOB_TYPE_LOAD_PEIM_UNUSED     0x000A
+#define EFI_HOB_TYPE_UEFI_CAPSULE         0x000B
+#define EFI_HOB_TYPE_FV3                  0x000C
+#define EFI_HOB_TYPE_UNUSED               0xFFFE
+#define EFI_HOB_TYPE_END_OF_HOB_LIST      0xFFFF
+
+typedef struct {
+    uint16_t HobType;
+    uint16_t HobLength;
+    uint32_t Reserved;
+} EFI_HOB_GENERIC_HEADER;
+
+typedef uint64_t EFI_PHYSICAL_ADDRESS;
+typedef uint32_t EFI_BOOT_MODE;
+
+typedef struct {
+    EFI_HOB_GENERIC_HEADER Header;
+    uint32_t Version;
+    EFI_BOOT_MODE BootMode;
+    EFI_PHYSICAL_ADDRESS EfiMemoryTop;
+    EFI_PHYSICAL_ADDRESS EfiMemoryBottom;
+    EFI_PHYSICAL_ADDRESS EfiFreeMemoryTop;
+    EFI_PHYSICAL_ADDRESS EfiFreeMemoryBottom;
+    EFI_PHYSICAL_ADDRESS EfiEndOfHobList;
+} EFI_HOB_HANDOFF_INFO_TABLE;
+
+#define EFI_RESOURCE_SYSTEM_MEMORY          0x00000000
+#define EFI_RESOURCE_MEMORY_MAPPED_IO       0x00000001
+#define EFI_RESOURCE_IO                     0x00000002
+#define EFI_RESOURCE_FIRMWARE_DEVICE        0x00000003
+#define EFI_RESOURCE_MEMORY_MAPPED_IO_PORT  0x00000004
+#define EFI_RESOURCE_MEMORY_RESERVED        0x00000005
+#define EFI_RESOURCE_IO_RESERVED            0x00000006
+#define EFI_RESOURCE_MEMORY_UNACCEPTED      0x00000007
+#define EFI_RESOURCE_MAX_MEMORY_TYPE        0x00000008
+
+#define EFI_RESOURCE_ATTRIBUTE_PRESENT                  0x00000001
+#define EFI_RESOURCE_ATTRIBUTE_INITIALIZED              0x00000002
+#define EFI_RESOURCE_ATTRIBUTE_TESTED                   0x00000004
+#define EFI_RESOURCE_ATTRIBUTE_SINGLE_BIT_ECC           0x00000008
+#define EFI_RESOURCE_ATTRIBUTE_MULTIPLE_BIT_ECC         0x00000010
+#define EFI_RESOURCE_ATTRIBUTE_ECC_RESERVED_1           0x00000020
+#define EFI_RESOURCE_ATTRIBUTE_ECC_RESERVED_2           0x00000040
+#define EFI_RESOURCE_ATTRIBUTE_READ_PROTECTED           0x00000080
+#define EFI_RESOURCE_ATTRIBUTE_WRITE_PROTECTED          0x00000100
+#define EFI_RESOURCE_ATTRIBUTE_EXECUTION_PROTECTED      0x00000200
+#define EFI_RESOURCE_ATTRIBUTE_UNCACHEABLE              0x00000400
+#define EFI_RESOURCE_ATTRIBUTE_WRITE_COMBINEABLE        0x00000800
+#define EFI_RESOURCE_ATTRIBUTE_WRITE_THROUGH_CACHEABLE  0x00001000
+#define EFI_RESOURCE_ATTRIBUTE_WRITE_BACK_CACHEABLE     0x00002000
+#define EFI_RESOURCE_ATTRIBUTE_16_BIT_IO                0x00004000
+#define EFI_RESOURCE_ATTRIBUTE_32_BIT_IO                0x00008000
+#define EFI_RESOURCE_ATTRIBUTE_64_BIT_IO                0x00010000
+#define EFI_RESOURCE_ATTRIBUTE_UNCACHED_EXPORTED        0x00020000
+#define EFI_RESOURCE_ATTRIBUTE_READ_ONLY_PROTECTED      0x00040000
+#define EFI_RESOURCE_ATTRIBUTE_READ_ONLY_PROTECTABLE    0x00080000
+#define EFI_RESOURCE_ATTRIBUTE_READ_PROTECTABLE         0x00100000
+#define EFI_RESOURCE_ATTRIBUTE_WRITE_PROTECTABLE        0x00200000
+#define EFI_RESOURCE_ATTRIBUTE_EXECUTION_PROTECTABLE    0x00400000
+#define EFI_RESOURCE_ATTRIBUTE_PERSISTENT               0x00800000
+#define EFI_RESOURCE_ATTRIBUTE_PERSISTABLE              0x01000000
+#define EFI_RESOURCE_ATTRIBUTE_MORE_RELIABLE            0x02000000
+
+typedef uint32_t EFI_RESOURCE_TYPE;
+typedef uint32_t EFI_RESOURCE_ATTRIBUTE_TYPE;
+
+typedef struct {
+    EFI_HOB_GENERIC_HEADER Header;
+    EFI_GUID Owner;
+    EFI_RESOURCE_TYPE ResourceType;
+    EFI_RESOURCE_ATTRIBUTE_TYPE ResourceAttribute;
+    EFI_PHYSICAL_ADDRESS PhysicalStart;
+    uint64_t ResourceLength;
+} EFI_HOB_RESOURCE_DESCRIPTOR;
+
+typedef struct {
+    EFI_HOB_GENERIC_HEADER Header;
+    EFI_GUID Name;
+
+    /* guid specific data follows */
+} EFI_HOB_GUID_TYPE;
+
+typedef struct {
+    EFI_HOB_GENERIC_HEADER Header;
+    EFI_PHYSICAL_ADDRESS BaseAddress;
+    uint64_t Length;
+} EFI_HOB_FIRMWARE_VOLUME;
+
+typedef struct {
+    EFI_HOB_GENERIC_HEADER Header;
+    EFI_PHYSICAL_ADDRESS BaseAddress;
+    uint64_t Length;
+    EFI_GUID FvName;
+    EFI_GUID FileName;
+} EFI_HOB_FIRMWARE_VOLUME2;
+
+typedef struct {
+    EFI_HOB_GENERIC_HEADER Header;
+    EFI_PHYSICAL_ADDRESS BaseAddress;
+    uint64_t Length;
+    uint32_t AuthenticationStatus;
+    bool ExtractedFv;
+    EFI_GUID FvName;
+    EFI_GUID FileName;
+} EFI_HOB_FIRMWARE_VOLUME3;
+
+typedef struct {
+    EFI_HOB_GENERIC_HEADER Header;
+    uint8_t SizeOfMemorySpace;
+    uint8_t SizeOfIoSpace;
+    uint8_t Reserved[6];
+} EFI_HOB_CPU;
+
+typedef struct {
+    EFI_HOB_GENERIC_HEADER Header;
+} EFI_HOB_MEMORY_POOL;
+
+typedef struct {
+    EFI_HOB_GENERIC_HEADER Header;
+
+    EFI_PHYSICAL_ADDRESS BaseAddress;
+    uint64_t Length;
+} EFI_HOB_UEFI_CAPSULE;
+
+#define EFI_HOB_OWNER_ZERO                                      \
+    ((EFI_GUID){ 0x00000000, 0x0000, 0x0000,                    \
+        { 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00 } })
+
+#endif
Signature · 1 line
2.34.1
#72 lore raw

Re: [PATCH v2 33/58] headers: Add definitions from UEFI spec for volumes, resources, etc...

Quoted 9 lines · Xiaoyao Li

Add UEFI definitions for literals, enums, structs, GUIDs, etc... that will be used by TDX to build the UEFI Hand-Off Block (HOB) that is passed to the Trusted Domain Virtual Firmware (TDVF).

All values come from the UEFI specification and TDVF design guide. [1]

Note, EFI_RESOURCE_MEMORY_UNACCEPTED will be added in future UEFI spec.

[1] https://software.intel.com/content/dam/develop/external/us/en/documents/tdx-virtual-firmware-design-guide-rev-1.pdf

Nitpick: The specs [1] [2] include unaccepted memory.

[1] UEFI Specification Version 2.10 (released August 2022)
[2] UEFI Platform Initialization Distribution Packaging Specification Version 1.1)

Signature · 1 line
Isaku Yamahata <[email protected]>
author
#73 lore raw

Re: [PATCH v2 33/58] headers: Add definitions from UEFI spec for volumes, resources, etc...

Quoted 7 lines · Isaku Yamahata
Quoted 4 lines · Xiaoyao Li

Add UEFI definitions for literals, enums, structs, GUIDs, etc... that will be used by TDX to build the UEFI Hand-Off Block (HOB) that is passed to the Trusted Domain Virtual Firmware (TDVF).

Quoted 7 lines
Quoted 5 lines

All values come from the UEFI specification and TDVF design guide. [1]

Note, EFI_RESOURCE_MEMORY_UNACCEPTED will be added in future UEFI spec.

[1] https://software.intel.com/content/dam/develop/external/us/en/documents/tdx-virtual-firmware-design-guide-rev-1.pdf

Nitpick: The specs [1] [2] include unaccepted memory.

EfiUnacceptedMemoryType shows in UEFI spec while EFI_RESOURCE_MEMORY_UNACCEPTED is still missing in PI spec.

https://github.com/tianocore/edk2/commit/00bbb1e584ec05547159f405cca383e8ba5e4ddb

Quoted 2 lines

[1] UEFI Specification Version 2.10 (released August 2022)
[2] UEFI Platform Initialization Distribution Packaging Specification Version 1.1)

author
#74 lore raw

Re: [PATCH v2 33/58] headers: Add definitions from UEFI spec for volumes, resources, etc...

Quoted 12 lines · Xiaoyao Li
Quoted 11 lines · Isaku Yamahata
Quoted 8 lines · Xiaoyao Li

Add UEFI definitions for literals, enums, structs, GUIDs, etc... that will be used by TDX to build the UEFI Hand-Off Block (HOB) that is passed to the Trusted Domain Virtual Firmware (TDVF).

All values come from the UEFI specification and TDVF design guide. [1]

Note, EFI_RESOURCE_MEMORY_UNACCEPTED will be added in future UEFI spec.

Quoted 7 lines
Quoted 3 lines
Quoted 1 line

[1] https://software.intel.com/content/dam/develop/external/us/en/documents/tdx-virtual-firmware-design-guide-rev-1.pdf

Nitpick: The specs [1] [2] include unaccepted memory.

EfiUnacceptedMemoryType shows in UEFI spec while EFI_RESOURCE_MEMORY_UNACCEPTED is still missing in PI spec.

https://github.com/tianocore/edk2/commit/00bbb1e584ec05547159f405cca383e8ba5e4ddb

Sorry, I just find it shows in latest PI spec.

https://uefi.org/sites/default/files/resources/UEFI_PI_Spec_1_8_March3.pdf

Quoted 4 lines
Quoted 2 lines

[1] UEFI Specification Version 2.10 (released August 2022)
[2] UEFI Platform Initialization Distribution Packaging Specification Version 1.1)

author
#75 lore raw

PATCH v2 34/58 i386/tdx: Setup the TD HOB list

In reply to Xiaoyao Li

The TD HOB list is used to pass the information from VMM to TDVF. The TD HOB must include PHIT HOB and Resource Descriptor HOB. More details can be found in TDVF specification and PI specification.

Build the TD HOB in TDX's machine_init_done callback.

---
Changes from RFC v4:
- drop the code of adding mmio resources since OVMF prepares all the MMIO hob itself.

---
 hw/i386/meson.build   |   2 +-
 hw/i386/tdvf-hob.c    | 147 ++++++++++++++++++++++++++++++++++++++++++
 hw/i386/tdvf-hob.h    |  24 +++++++
 target/i386/kvm/tdx.c |  16 +++++
 4 files changed, 188 insertions(+), 1 deletion(-)
 create mode 100644 hw/i386/tdvf-hob.c
 create mode 100644 hw/i386/tdvf-hob.h
hw/i386/meson.build +1 -1
index 45d90bb2af52..b38ea89665f0 100644
@@ -28,7 +28,7 @@ i386_ss.add(when: 'CONFIG_PC', if_true: files(
   'port92.c'))
 i386_ss.add(when: 'CONFIG_X86_FW_OVMF', if_true: files('pc_sysfw_ovmf.c'),
                                         if_false: files('pc_sysfw_ovmf-stubs.c'))
-i386_ss.add(when: 'CONFIG_TDX', if_true: files('tdvf.c'))
+i386_ss.add(when: 'CONFIG_TDX', if_true: files('tdvf.c', 'tdvf-hob.c'))
 
 subdir('kvm')
 subdir('xen')
hw/i386/tdvf-hob.c new file +147 -0

147 lines not on this page · raw message

hw/i386/tdvf-hob.h new file +24 -0
new file mode 100644
index 000000000000..1b737e946a8d
@@ -0,0 +1,24 @@
+#ifndef HW_I386_TD_HOB_H
+#define HW_I386_TD_HOB_H
+
+#include "hw/i386/tdvf.h"
+#include "target/i386/kvm/tdx.h"
+
+void tdvf_hob_create(TdxGuest *tdx, TdxFirmwareEntry *td_hob);
+
+#define EFI_RESOURCE_ATTRIBUTE_TDVF_PRIVATE     \
+    (EFI_RESOURCE_ATTRIBUTE_PRESENT |           \
+     EFI_RESOURCE_ATTRIBUTE_INITIALIZED |       \
+     EFI_RESOURCE_ATTRIBUTE_TESTED)
+
+#define EFI_RESOURCE_ATTRIBUTE_TDVF_UNACCEPTED  \
+    (EFI_RESOURCE_ATTRIBUTE_PRESENT |           \
+     EFI_RESOURCE_ATTRIBUTE_INITIALIZED |       \
+     EFI_RESOURCE_ATTRIBUTE_TESTED)
+
+#define EFI_RESOURCE_ATTRIBUTE_TDVF_MMIO        \
+    (EFI_RESOURCE_ATTRIBUTE_PRESENT     |       \
+     EFI_RESOURCE_ATTRIBUTE_INITIALIZED |       \
+     EFI_RESOURCE_ATTRIBUTE_UNCACHEABLE)
+
+#endif
target/i386/kvm/tdx.c +16 -0
index ed617ebab266..3a93ad293129 100644
@@ -25,6 +25,7 @@
 #include "hw/i386/e820_memory_layout.h"
 #include "hw/i386/x86.h"
 #include "hw/i386/tdvf.h"
+#include "hw/i386/tdvf-hob.h"
 #include "kvm_i386.h"
 #include "tdx.h"
 #include "../cpu-internal.h"
@@ -455,6 +456,19 @@ static void update_tdx_cpuid_lookup_by_tdx_caps(void)
             (tdx_caps->xfam_fixed1 & CPUID_XSTATE_XSS_MASK) >> 32;
 }
 
+static TdxFirmwareEntry *tdx_get_hob_entry(TdxGuest *tdx)
+{
+    TdxFirmwareEntry *entry;
+
+    for_each_tdx_fw_entry(&tdx->tdvf, entry) {
+        if (entry->type == TDVF_SECTION_TYPE_TD_HOB) {
+            return entry;
+        }
+    }
+    error_report("TDVF metadata doesn't specify TD_HOB location.");
+    exit(1);
+}
+
 static void tdx_add_ram_entry(uint64_t address, uint64_t length, uint32_t type)
 {
     uint32_t nr_entries = tdx_guest->nr_ram_entries;
@@ -585,6 +599,8 @@ static void tdx_finalize_vm(Notifier *notifier, void *unused)
 
     qsort(tdx_guest->ram_entries, tdx_guest->nr_ram_entries,
           sizeof(TdxRamEntry), &tdx_ram_entry_compare);
+
+    tdvf_hob_create(tdx_guest, tdx_get_hob_entry(tdx_guest));
 }
 
 static Notifier tdx_machine_done_notify = {
Signature · 1 line
2.34.1
author
#76 lore raw

PATCH v2 35/58 i386/tdx: Add TDVF memory via KVM_TDX_INIT_MEM_REGION

In reply to Xiaoyao Li

From: Isaku Yamahata <[email protected]>

TDVF firmware (CODE and VARS) needs to be added/copied to TD's private memory via KVM_TDX_INIT_MEM_REGION, as well as TD HOB and TEMP memory.

---
Changes from RFC v4:
- rename variable @metadata to @flags

---
 target/i386/kvm/tdx.c | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)
target/i386/kvm/tdx.c +24 -0
index 3a93ad293129..37ff0f4eea11 100644
@@ -576,6 +576,7 @@ static void tdx_finalize_vm(Notifier *notifier, void *unused)
 {
     TdxFirmware *tdvf = &tdx_guest->tdvf;
     TdxFirmwareEntry *entry;
+    int r;
 
     tdx_init_ram_entries();
 
@@ -601,6 +602,29 @@ static void tdx_finalize_vm(Notifier *notifier, void *unused)
           sizeof(TdxRamEntry), &tdx_ram_entry_compare);
 
     tdvf_hob_create(tdx_guest, tdx_get_hob_entry(tdx_guest));
+
+    for_each_tdx_fw_entry(tdvf, entry) {
+        struct kvm_tdx_init_mem_region mem_region = {
+            .source_addr = (__u64)entry->mem_ptr,
+            .gpa = entry->address,
+            .nr_pages = entry->size / 4096,
+        };
+
+        __u32 flags = entry->attributes & TDVF_SECTION_ATTRIBUTES_MR_EXTEND ?
+                      KVM_TDX_MEASURE_MEMORY_REGION : 0;
+
+        r = tdx_vm_ioctl(KVM_TDX_INIT_MEM_REGION, flags, &mem_region);
+        if (r < 0) {
+             error_report("KVM_TDX_INIT_MEM_REGION failed %s", strerror(-r));
+             exit(1);
+        }
+
+        if (entry->type == TDVF_SECTION_TYPE_TD_HOB ||
+            entry->type == TDVF_SECTION_TYPE_TEMP_MEM) {
+            qemu_ram_munmap(-1, entry->mem_ptr, entry->size);
+            entry->mem_ptr = NULL;
+        }
+    }
 }
 
 static Notifier tdx_machine_done_notify = {
Signature · 1 line
2.34.1
author
#77 lore raw

PATCH v2 36/58 memory: Introduce memory_region_init_ram_gmem()

In reply to Xiaoyao Li

Introduce memory_region_init_ram_gmem() to allocate private gmem on the MemoryRegion initialization. It's for the usercase of TDVF, which must be private on TDX case.

---
 include/exec/memory.h |  6 +++++
 softmmu/memory.c      | 52 +++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 58 insertions(+)
include/exec/memory.h +6 -0
index 759f797b6acd..127ffb6556b9 100644
@@ -1564,6 +1564,12 @@ void memory_region_init_ram(MemoryRegion *mr,
                             uint64_t size,
                             Error **errp);
 
+void memory_region_init_ram_gmem(MemoryRegion *mr,
+                                 Object *owner,
+                                 const char *name,
+                                 uint64_t size,
+                                 Error **errp);
+
 /**
  * memory_region_init_rom: Initialize a ROM memory region.
  *
softmmu/memory.c +52 -0

64 lines not on this page · raw message

Signature · 1 line
2.34.1
#78 lore raw

Re: [PATCH v2 36/58] memory: Introduce memory_region_init_ram_gmem()

Quoted 23 lines · Xiaoyao Li

Introduce memory_region_init_ram_gmem() to allocate private gmem on the MemoryRegion initialization. It's for the usercase of TDVF, which must be private on TDX case.

---
 include/exec/memory.h |  6 +++++
 softmmu/memory.c      | 52 +++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 58 insertions(+)
softmmu/memory.c +1 -0
index af6aa3c1e3c9..ded44dcef1aa 100644
@@ -25,6 +25,7 @@
 #include "qom/object.h"
 #include "trace.h"
 
+#include <linux/kvm.h>
 #include "exec/memory-internal.h"
 #include "exec/ram_addr.h"
 #include "sysemu/kvm.h"

softmmu/memory.c void memory_region_init_ram(MemoryRegion *mr, · line 3611

Quoted 9 lines
softmmu/memory.c +6 -0
@@ -3602,6 +3603,57 @@ void memory_region_init_ram(MemoryRegion *mr,
     vmstate_register_ram(mr, owner_dev);
 }
 
+#ifdef CONFIG_KVM
+void memory_region_init_ram_gmem(MemoryRegion *mr,
+                                 Object *owner,
+                                 const char *name,
+                                 uint64_t size,
+                                 Error **errp)

Since you have an 'errp' parameter here....

Quoted 14 lines
softmmu/memory.c +14 -0
+{
+    DeviceState *owner_dev;
+    Error *err = NULL;
+    int priv_fd;
+
+    memory_region_init_ram_nomigrate(mr, owner, name, size, &err);
+    if (err) {
+        error_propagate(errp, err);
+        return;
+    }
+
+    if (object_dynamic_cast(OBJECT(current_accel()), TYPE_KVM_ACCEL)) {
+        KVMState *s = KVM_STATE(current_accel());
+        struct kvm_create_guest_memfd gmem = {

softmmu/memory.c void memory_region_init_ram(MemoryRegion *mr, · line 3635

Quoted 10 lines
softmmu/memory.c +10 -0
+            .size = size,
+            /* TODO: add property to hostmem backend for huge pmd */
+            .flags = KVM_GUEST_MEMFD_ALLOW_HUGEPAGE,
+        };
+
+        priv_fd = kvm_vm_ioctl(s, KVM_CREATE_GUEST_MEMFD, &gmem);
+        if (priv_fd < 0) {
+            fprintf(stderr, "%s: error creating gmem: %s\n", __func__,
+                    strerror(-priv_fd));
+            abort();

It should be using error_setg_errno() here and return not abort

softmmu/memory.c void memory_region_init_ram(MemoryRegion *mr, · line 3639

Quoted 4 lines
softmmu/memory.c +4 -0
+        }
+    } else {
+        fprintf(stderr, "%s: gmem unsupported accel: %s\n", __func__,
+                current_accel_name());

and error_setg() here and return.

Quoted 23 lines
softmmu/memory.c +17 -0
+        abort();
+    }
+
+    memory_region_set_gmem_fd(mr, priv_fd);
+    memory_region_set_default_private(mr);
+
+    /* This will assert if owner is neither NULL nor a DeviceState.
+     * We only want the owner here for the purposes of defining a
+     * unique name for migration. TODO: Ideally we should implement
+     * a naming scheme for Objects which are not DeviceStates, in
+     * which case we can relax this restriction.
+     */
+    owner_dev = DEVICE(owner);
+    vmstate_register_ram(mr, owner_dev);
+}
+#endif
+
 void memory_region_init_rom(MemoryRegion *mr,
                             Object *owner,
                             const char *name,
Signature · 1 line
2.34.1

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#79 lore raw

Re: [PATCH v2 36/58] memory: Introduce memory_region_init_ram_gmem()

In reply to Xiaoyao Li
Quoted 16 lines · Xiaoyao Li

Introduce memory_region_init_ram_gmem() to allocate private gmem on the MemoryRegion initialization. It's for the usercase of TDVF, which must be private on TDX case.

---
  include/exec/memory.h |  6 +++++
  softmmu/memory.c      | 52 +++++++++++++++++++++++++++++++++++++++++++
  2 files changed, 58 insertions(+)
softmmu/memory.c +0 -0
index af6aa3c1e3c9..ded44dcef1aa 100644

softmmu/memory.c line 28

Quoted 4 lines
softmmu/memory.c +1 -0
@@ -25,6 +25,7 @@
  #include "qom/object.h"
  #include "trace.h"
 
+#include <linux/kvm.h>

Unlikely to build on non-Linux hosts.

Quoted 3 lines
softmmu/memory.c +0 -0
  #include "exec/memory-internal.h"
  #include "exec/ram_addr.h"
  #include "sysemu/kvm.h"
author
#80 lore raw

Re: [PATCH v2 36/58] memory: Introduce memory_region_init_ram_gmem()

Quoted 12 lines · Philippe Mathieu-Daudé
Quoted 9 lines · Xiaoyao Li

Introduce memory_region_init_ram_gmem() to allocate private gmem on the MemoryRegion initialization. It's for the usercase of TDVF, which must be private on TDX case.

---
  include/exec/memory.h |  6 +++++
  softmmu/memory.c      | 52 +++++++++++++++++++++++++++++++++++++++++++

  2 files changed, 58 insertions(+)
Quoted 10 lines

softmmu/memory.c line 27

Quoted 8 lines
softmmu/memory.c +1 -0
index af6aa3c1e3c9..ded44dcef1aa 100644
@@ -25,6 +25,7 @@
   #include "qom/object.h"
   #include "trace.h"
+#include <linux/kvm.h>

Unlikely to build on non-Linux hosts.

Thanks for catching it!

Will warp it with CONFIG_KVM.

Anyway, it's the main open of how to integrating KVM gmem into QEMU's memory system, in QMEU gmem series[*]. I'm still working on it.

[*] https://lore.kernel.org/qemu-devel/[email protected]/

Quoted 4 lines
Quoted 3 lines
softmmu/memory.c +0 -0
   #include "exec/memory-internal.h"
   #include "exec/ram_addr.h"
   #include "sysemu/kvm.h"
author
#81 lore raw

PATCH v2 37/58 i386/tdx: register TDVF as private memory

In reply to Xiaoyao Li

From: Chao Peng <[email protected]>

Allocate private gmem memory for BIOS if it's TD VM.

---
 hw/i386/x86.c         |  9 ++++++++-
 target/i386/kvm/tdx.c | 17 +++++++++++++++++
 target/i386/kvm/tdx.h |  2 ++
 3 files changed, 27 insertions(+), 1 deletion(-)
hw/i386/x86.c +8 -1
index e2a1369e5dfc..a0c9f4d646e2 100644
@@ -1151,8 +1151,15 @@ void x86_bios_rom_init(MachineState *ms, const char *default_firmware,
         (bios_size % 65536) != 0) {
         goto bios_error;
     }
+
     bios = g_malloc(sizeof(*bios));
-    memory_region_init_ram(bios, NULL, "pc.bios", bios_size, &error_fatal);
+    if (is_tdx_vm()) {
+        memory_region_init_ram_gmem(bios, NULL, "pc.bios", bios_size, &error_fatal);
+        tdx_set_tdvf_region(bios);
+    } else {
+        memory_region_init_ram(bios, NULL, "pc.bios", bios_size, &error_fatal);
+    }
+
     if (sev_enabled() || is_tdx_vm()) {
         /*
          * The concept of a "reset" simply doesn't exist for
target/i386/kvm/tdx.c +17 -0
index 37ff0f4eea11..5b688eb39327 100644
@@ -456,6 +456,12 @@ static void update_tdx_cpuid_lookup_by_tdx_caps(void)
             (tdx_caps->xfam_fixed1 & CPUID_XSTATE_XSS_MASK) >> 32;
 }
 
+void tdx_set_tdvf_region(MemoryRegion *tdvf_region)
+{
+    assert(!tdx_guest->tdvf_region);
+    tdx_guest->tdvf_region = tdvf_region;
+}
+
 static TdxFirmwareEntry *tdx_get_hob_entry(TdxGuest *tdx)
 {
     TdxFirmwareEntry *entry;
@@ -576,6 +582,7 @@ static void tdx_finalize_vm(Notifier *notifier, void *unused)
 {
     TdxFirmware *tdvf = &tdx_guest->tdvf;
     TdxFirmwareEntry *entry;
+    RAMBlock *ram_block;
     int r;
 
     tdx_init_ram_entries();
@@ -610,6 +617,12 @@ static void tdx_finalize_vm(Notifier *notifier, void *unused)
             .nr_pages = entry->size / 4096,
         };
 
+        r = kvm_set_memory_attributes_private(entry->address, entry->size);
+        if (r < 0) {
+             error_report("Reserve initial private memory failed %s", strerror(-r));
+             exit(1);
+        }
+
         __u32 flags = entry->attributes & TDVF_SECTION_ATTRIBUTES_MR_EXTEND ?
                       KVM_TDX_MEASURE_MEMORY_REGION : 0;
 
@@ -625,6 +638,10 @@ static void tdx_finalize_vm(Notifier *notifier, void *unused)
             entry->mem_ptr = NULL;
         }
     }
+
+    /* Tdvf image was copied into private region above. It becomes unnecessary. */
+    ram_block = tdx_guest->tdvf_region->ram_block;
+    ram_block_discard_range(ram_block, 0, ram_block->max_length);
 }
 
 static Notifier tdx_machine_done_notify = {
target/i386/kvm/tdx.h +2 -0
index 9b3c427766ef..1c444b6cdb3f 100644
@@ -38,6 +38,7 @@ typedef struct TdxGuest {
     uint8_t mrownerconfig[48];  /* sha348 digest */
 
     TdxFirmware tdvf;
+    MemoryRegion *tdvf_region;
 
     uint32_t nr_ram_entries;
     TdxRamEntry *ram_entries;
@@ -53,6 +54,7 @@ int tdx_kvm_init(MachineState *ms, Error **errp);
 void tdx_get_supported_cpuid(uint32_t function, uint32_t index, int reg,
                              uint32_t *ret);
 int tdx_pre_create_vcpu(CPUState *cpu);
+void tdx_set_tdvf_region(MemoryRegion *tdvf_region);
 int tdx_parse_tdvf(void *flash_ptr, int size);
 
 #endif /* QEMU_I386_TDX_H */
Signature · 1 line
2.34.1
author
#82 lore raw

PATCH v2 38/58 i386/tdx: Call KVM_TDX_INIT_VCPU to initialize TDX vcpu

In reply to Xiaoyao Li

TDX vcpu needs to be initialized by SEAMCALL(TDH.VP.INIT) and KVM provides vcpu level IOCTL KVM_TDX_INIT_VCPU for it.

KVM_TDX_INIT_VCPU needs the address of the HOB as input. Invoke it for each vcpu after HOB list is created.

---
 target/i386/kvm/tdx.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
target/i386/kvm/tdx.c +18 -0
index 5b688eb39327..2d0efca6787b 100644
@@ -578,6 +578,22 @@ static void tdx_init_ram_entries(void)
     tdx_guest->nr_ram_entries = j;
 }
 
+static void tdx_post_init_vcpus(void)
+{
+    TdxFirmwareEntry *hob;
+    CPUState *cpu;
+    int r;
+
+    hob = tdx_get_hob_entry(tdx_guest);
+    CPU_FOREACH(cpu) {
+        r = tdx_vcpu_ioctl(cpu, KVM_TDX_INIT_VCPU, 0, (void *)hob->address);
+        if (r < 0) {
+            error_report("KVM_TDX_INIT_VCPU failed %s", strerror(-r));
+            exit(1);
+        }
+    }
+}
+
 static void tdx_finalize_vm(Notifier *notifier, void *unused)
 {
     TdxFirmware *tdvf = &tdx_guest->tdvf;
@@ -610,6 +626,8 @@ static void tdx_finalize_vm(Notifier *notifier, void *unused)
 
     tdvf_hob_create(tdx_guest, tdx_get_hob_entry(tdx_guest));
 
+    tdx_post_init_vcpus();
+
     for_each_tdx_fw_entry(tdvf, entry) {
         struct kvm_tdx_init_mem_region mem_region = {
             .source_addr = (__u64)entry->mem_ptr,
Signature · 1 line
2.34.1
author
#83 lore raw

PATCH v2 39/58 i386/tdx: Finalize TDX VM

In reply to Xiaoyao Li

Invoke KVM_TDX_FINALIZE_VM to finalize the TD's measurement and make the TD vCPUs runnable once machine initialization is complete.

---
 target/i386/kvm/tdx.c | 7 +++++++
 1 file changed, 7 insertions(+)
target/i386/kvm/tdx.c +7 -0
index 2d0efca6787b..5eabdafbe95c 100644
@@ -660,6 +660,13 @@ static void tdx_finalize_vm(Notifier *notifier, void *unused)
     /* Tdvf image was copied into private region above. It becomes unnecessary. */
     ram_block = tdx_guest->tdvf_region->ram_block;
     ram_block_discard_range(ram_block, 0, ram_block->max_length);
+
+    r = tdx_vm_ioctl(KVM_TDX_FINALIZE_VM, 0, NULL);
+    if (r < 0) {
+        error_report("KVM_TDX_FINALIZE_VM failed %s", strerror(-r));
+        exit(0);
+    }
+    tdx_guest->parent_obj.ready = true;
 }
 
 static Notifier tdx_machine_done_notify = {
Signature · 1 line
2.34.1
author
#84 lore raw

PATCH v2 40/58 i386/tdx: handle TDG.VP.VMCALL<SetupEventNotifyInterrupt>

In reply to Xiaoyao Li

From: Isaku Yamahata <[email protected]>

For SetupEventNotifyInterrupt, record interrupt vector and the apic id of the vcpu that received this TDVMCALL.

Later it can inject interrupt with given vector to the specific vcpu that received SetupEventNotifyInterrupt.

---
 target/i386/kvm/kvm.c      |  9 ++++++
 target/i386/kvm/tdx-stub.c |  4 +++
 target/i386/kvm/tdx.c      | 61 ++++++++++++++++++++++++++++++++++++++
 target/i386/kvm/tdx.h      |  6 ++++
 4 files changed, 80 insertions(+)
target/i386/kvm/kvm.c +9 -0
index 4a146bc42f63..601683d836c8 100644
@@ -5657,6 +5657,15 @@ int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
         ret = kvm_xen_handle_exit(cpu, &run->xen);
         break;
 #endif
+    case KVM_EXIT_TDX:
+        if (!is_tdx_vm()) {
+            error_report("KVM: get KVM_EXIT_TDX for a non-TDX VM.");
+            ret = -1;
+            break;
+        }
+        tdx_handle_exit(cpu, &run->tdx);
+        ret = 0;
+        break;
     default:
         fprintf(stderr, "KVM: unknown exit reason %d\n", run->exit_reason);
         ret = -1;
target/i386/kvm/tdx-stub.c +4 -0
index 3cd114476d78..e26e2111f606 100644
@@ -16,3 +16,7 @@ int tdx_parse_tdvf(void *flash_ptr, int size)
 {
     return -EINVAL;
 }
+
+void tdx_handle_exit(X86CPU *cpu, struct kvm_tdx_exit *tdx_exit)
+{
+}
target/i386/kvm/tdx.c +61 -0

70 lines not on this page · raw message

target/i386/kvm/tdx.h +6 -0
index 1c444b6cdb3f..50a151fc79c2 100644
@@ -7,6 +7,7 @@
 
 #include "exec/confidential-guest-support.h"
 #include "hw/i386/tdvf.h"
+#include "sysemu/kvm.h"
 
 #define TYPE_TDX_GUEST "tdx-guest"
 #define TDX_GUEST(obj)  OBJECT_CHECK(TdxGuest, (obj), TYPE_TDX_GUEST)
@@ -42,6 +43,10 @@ typedef struct TdxGuest {
 
     uint32_t nr_ram_entries;
     TdxRamEntry *ram_entries;
+
+    /* runtime state */
+    int event_notify_interrupt;
+    uint32_t event_notify_apic_id;
 } TdxGuest;
 
 #ifdef CONFIG_TDX
@@ -56,5 +61,6 @@ void tdx_get_supported_cpuid(uint32_t function, uint32_t index, int reg,
 int tdx_pre_create_vcpu(CPUState *cpu);
 void tdx_set_tdvf_region(MemoryRegion *tdvf_region);
 int tdx_parse_tdvf(void *flash_ptr, int size);
+void tdx_handle_exit(X86CPU *cpu, struct kvm_tdx_exit *tdx_exit);
 
 #endif /* QEMU_I386_TDX_H */
Signature · 1 line
2.34.1
author
#85 lore raw

PATCH v2 41/58 i386/tdx: handle TDG.VP.VMCALL<GetQuote>

In reply to Xiaoyao Li

From: Isaku Yamahata <[email protected]>

For GetQuote, delegate a request to Quote Generation Service. Add property of address of quote generation server and On request, connect to the server, read request buffer from shared guest memory, send the request buffer to the server and store the response into shared guest memory and notify TD guest by interrupt.

"quote-generation-service" is a property to specify Quote Generation Service(QGS) in qemu socket address format. The examples of the supported format are "vsock:2:1234", "unix:/run/qgs", "localhost:1234".

command line example:

  qemu-system-x86_64 \
    -object 'tdx-guest,id=tdx0,quote-generation-service=localhost:1234' \
    -machine confidential-guest-support=tdx0
---
 qapi/qom.json         |   5 +-
 target/i386/kvm/tdx.c | 380 ++++++++++++++++++++++++++++++++++++++++++
 target/i386/kvm/tdx.h |   7 +
 3 files changed, 391 insertions(+), 1 deletion(-)
qapi/qom.json +4 -1
index 87c1d440f331..37139949d761 100644
@@ -879,13 +879,16 @@
 #
 # @mrownerconfig: MROWNERCONFIG SHA384 hex string of 48 * 2 length (default: 0)
 #
+# @quote-generation-service: socket address for Quote Generation Service(QGS)
+#
 # Since: 8.2
 ##
 { 'struct': 'TdxGuestProperties',
   'data': { '*sept-ve-disable': 'bool',
             '*mrconfigid': 'str',
             '*mrowner': 'str',
-            '*mrownerconfig': 'str' } }
+            '*mrownerconfig': 'str',
+            '*quote-generation-service': 'str' } }
 
 ##
 # @ThreadContextProperties:
target/i386/kvm/tdx.c +380 -0

416 lines not on this page · raw message

target/i386/kvm/tdx.h +7 -0
index 50a151fc79c2..d861d8516668 100644
@@ -5,8 +5,10 @@
 #include CONFIG_DEVICES /* CONFIG_TDX */
 #endif
 
+#include <linux/kvm.h>
 #include "exec/confidential-guest-support.h"
 #include "hw/i386/tdvf.h"
+#include "io/channel-socket.h"
 #include "sysemu/kvm.h"
 
 #define TYPE_TDX_GUEST "tdx-guest"
@@ -47,6 +49,11 @@ typedef struct TdxGuest {
     /* runtime state */
     int event_notify_interrupt;
     uint32_t event_notify_apic_id;
+
+    /* GetQuote */
+    int quote_generation_num;
+    char *quote_generation_str;
+    SocketAddress *quote_generation;
 } TdxGuest;
 
 #ifdef CONFIG_TDX
Signature · 1 line
2.34.1
#86 lore raw

Re: [PATCH v2 41/58] i386/tdx: handle TDG.VP.VMCALL<GetQuote>

Quoted 34 lines · Xiaoyao Li

From: Isaku Yamahata <[email protected]>

For GetQuote, delegate a request to Quote Generation Service. Add property of address of quote generation server and On request, connect to the server, read request buffer from shared guest memory, send the request buffer to the server and store the response into shared guest memory and notify TD guest by interrupt.

"quote-generation-service" is a property to specify Quote Generation Service(QGS) in qemu socket address format. The examples of the supported format are "vsock:2:1234", "unix:/run/qgs", "localhost:1234".

command line example:

  qemu-system-x86_64 \
    -object 'tdx-guest,id=tdx0,quote-generation-service=localhost:1234' \
    -machine confidential-guest-support=tdx0
---
 qapi/qom.json         |   5 +-
 target/i386/kvm/tdx.c | 380 ++++++++++++++++++++++++++++++++++++++++++
 target/i386/kvm/tdx.h |   7 +
 3 files changed, 391 insertions(+), 1 deletion(-)
qapi/qom.json +1 -0
index 87c1d440f331..37139949d761 100644
@@ -879,13 +879,16 @@
 #
 # @mrownerconfig: MROWNERCONFIG SHA384 hex string of 48 * 2 length (default: 0)
 #
+# @quote-generation-service: socket address for Quote Generation Service(QGS)

qapi/qom.json line 891

Quoted 10 lines
qapi/qom.json +3 -1
+#
 # Since: 8.2
 ##
 { 'struct': 'TdxGuestProperties',
   'data': { '*sept-ve-disable': 'bool',
             '*mrconfigid': 'str',
             '*mrowner': 'str',
-            '*mrownerconfig': 'str' } }
+            '*mrownerconfig': 'str',
+            '*quote-generation-service': 'str' } }

Why not type SocketAddress?

Quoted 3 lines
qapi/qom.json +0 -0
 ##
 # @ThreadContextProperties:

[...]

#87 lore raw

Re: [PATCH v2 41/58] i386/tdx: handle TDG.VP.VMCALL<GetQuote>

Quoted 40 lines · Markus Armbruster
Quoted 38 lines · Xiaoyao Li

From: Isaku Yamahata <[email protected]>

For GetQuote, delegate a request to Quote Generation Service. Add property of address of quote generation server and On request, connect to the server, read request buffer from shared guest memory, send the request buffer to the server and store the response into shared guest memory and notify TD guest by interrupt.

"quote-generation-service" is a property to specify Quote Generation Service(QGS) in qemu socket address format. The examples of the supported format are "vsock:2:1234", "unix:/run/qgs", "localhost:1234".

command line example:

  qemu-system-x86_64 \
    -object 'tdx-guest,id=tdx0,quote-generation-service=localhost:1234' \
    -machine confidential-guest-support=tdx0
---
 qapi/qom.json         |   5 +-
 target/i386/kvm/tdx.c | 380 ++++++++++++++++++++++++++++++++++++++++++
 target/i386/kvm/tdx.h |   7 +
 3 files changed, 391 insertions(+), 1 deletion(-)
qapi/qom.json +2 -0
index 87c1d440f331..37139949d761 100644
@@ -879,13 +879,16 @@
 #
 # @mrownerconfig: MROWNERCONFIG SHA384 hex string of 48 * 2 length (default: 0)
 #
+# @quote-generation-service: socket address for Quote Generation Service(QGS)
+#
 # Since: 8.2
 ##
 { 'struct': 'TdxGuestProperties',
Quoted 8 lines

qapi/qom.json line 891

Quoted 6 lines
qapi/qom.json +2 -1
   'data': { '*sept-ve-disable': 'bool',
             '*mrconfigid': 'str',
             '*mrowner': 'str',
-            '*mrownerconfig': 'str' } }
+            '*mrownerconfig': 'str',
+            '*quote-generation-service': 'str' } }

Why not type SocketAddress?

Yes, the code uses SocketAddress internally when it eventually calls qio_channel_socket_connect_async(), so we should directly use SocketAddress in the QAPI from the start.

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#88 lore raw

Re: [PATCH v2 41/58] i386/tdx: handle TDG.VP.VMCALL<GetQuote>

Quoted 50 lines · Daniel P. Berrangé
Quoted 40 lines · Markus Armbruster
Quoted 38 lines · Xiaoyao Li

From: Isaku Yamahata <[email protected]>

For GetQuote, delegate a request to Quote Generation Service. Add property of address of quote generation server and On request, connect to the server, read request buffer from shared guest memory, send the request buffer to the server and store the response into shared guest memory and notify TD guest by interrupt.

"quote-generation-service" is a property to specify Quote Generation Service(QGS) in qemu socket address format. The examples of the supported format are "vsock:2:1234", "unix:/run/qgs", "localhost:1234".

command line example:

  qemu-system-x86_64 \
    -object 'tdx-guest,id=tdx0,quote-generation-service=localhost:1234' \
    -machine confidential-guest-support=tdx0
---
 qapi/qom.json         |   5 +-
 target/i386/kvm/tdx.c | 380 ++++++++++++++++++++++++++++++++++++++++++
 target/i386/kvm/tdx.h |   7 +
 3 files changed, 391 insertions(+), 1 deletion(-)
qapi/qom.json +2 -0
index 87c1d440f331..37139949d761 100644
@@ -879,13 +879,16 @@
 #
 # @mrownerconfig: MROWNERCONFIG SHA384 hex string of 48 * 2 length (default: 0)
 #
+# @quote-generation-service: socket address for Quote Generation Service(QGS)
+#
 # Since: 8.2
 ##
 { 'struct': 'TdxGuestProperties',
Quoted 8 lines

qapi/qom.json line 891

Quoted 6 lines
qapi/qom.json +2 -1
   'data': { '*sept-ve-disable': 'bool',
             '*mrconfigid': 'str',
             '*mrowner': 'str',
-            '*mrownerconfig': 'str' } }
+            '*mrownerconfig': 'str',
+            '*quote-generation-service': 'str' } }

Why not type SocketAddress?

Quoted 3 lines

Yes, the code uses SocketAddress internally when it eventually calls qio_channel_socket_connect_async(), so we should directly use SocketAddress in the QAPI from the start.

Any benefit to directly use SocketAddress?

"quote-generation-service" here is optional, it seems not trivial to add and parse the SocketAddress type in QEMU command. After I change 'str' to 'SocketAddress' and specify the command like "-object tdx-guest,type=vsock,cid=2,port=1234...", it will report "invalid parameter cid".

Quoted 3 lines

With regards,
Daniel

#89 lore raw

Re: [PATCH v2 41/58] i386/tdx: handle TDG.VP.VMCALL<GetQuote>

Quoted 53 lines · Chenyi Qiang
Quoted 50 lines · Daniel P. Berrangé
Quoted 48 lines · Markus Armbruster
Quoted 34 lines · Xiaoyao Li

From: Isaku Yamahata <[email protected]>

For GetQuote, delegate a request to Quote Generation Service. Add property of address of quote generation server and On request, connect to the server, read request buffer from shared guest memory, send the request buffer to the server and store the response into shared guest memory and notify TD guest by interrupt.

"quote-generation-service" is a property to specify Quote Generation Service(QGS) in qemu socket address format. The examples of the supported format are "vsock:2:1234", "unix:/run/qgs", "localhost:1234".

command line example:

  qemu-system-x86_64 \
    -object 'tdx-guest,id=tdx0,quote-generation-service=localhost:1234' \
    -machine confidential-guest-support=tdx0
---
 qapi/qom.json         |   5 +-
 target/i386/kvm/tdx.c | 380 ++++++++++++++++++++++++++++++++++++++++++
 target/i386/kvm/tdx.h |   7 +
 3 files changed, 391 insertions(+), 1 deletion(-)
qapi/qom.json +1 -0
index 87c1d440f331..37139949d761 100644
@@ -879,13 +879,16 @@
 #
 # @mrownerconfig: MROWNERCONFIG SHA384 hex string of 48 * 2 length (default: 0)
 #
+# @quote-generation-service: socket address for Quote Generation Service(QGS)

qapi/qom.json line 891

Quoted 10 lines
qapi/qom.json +3 -1
+#
 # Since: 8.2
 ##
 { 'struct': 'TdxGuestProperties',
   'data': { '*sept-ve-disable': 'bool',
             '*mrconfigid': 'str',
             '*mrowner': 'str',
-            '*mrownerconfig': 'str' } }
+            '*mrownerconfig': 'str',
+            '*quote-generation-service': 'str' } }

Why not type SocketAddress?

Quoted 5 lines
Quoted 3 lines

Yes, the code uses SocketAddress internally when it eventually calls qio_channel_socket_connect_async(), so we should directly use SocketAddress in the QAPI from the start.

Any benefit to directly use SocketAddress?

We don't want whatever code consumes the configuration to do a second level of parsing to convert the 'str' value into the 'SocketAddress' object it actually needs.

QEMU has a long history of having a second round of ad-hoc parsing of configuration and we've found it to be a serious maintenence burden. Thus we strive to have everything represented in QAPI using the desired final type, and avoid the second round of parsing.

Quoted 5 lines

"quote-generation-service" here is optional, it seems not trivial to add and parse the SocketAddress type in QEMU command. After I change 'str' to 'SocketAddress' and specify the command like "-object tdx-guest,type=vsock,cid=2,port=1234...", it will report "invalid parameter cid".

The -object parameter supports JSON syntax for this reason

-object '{"qom-type":"tdx-guest","quote-generation-service":{"type": "vsock", "cid":"2","port":"1234"}}'

libvirt will always use the JSON syntax for -object with a new enough QEMU.

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
#90 lore raw

Re: [PATCH v2 41/58] i386/tdx: handle TDG.VP.VMCALL<GetQuote>

Quoted 64 lines · Daniel P. Berrangé
Quoted 53 lines · Chenyi Qiang
Quoted 50 lines · Daniel P. Berrangé
Quoted 48 lines · Markus Armbruster
Quoted 34 lines · Xiaoyao Li

From: Isaku Yamahata <[email protected]>

For GetQuote, delegate a request to Quote Generation Service. Add property of address of quote generation server and On request, connect to the server, read request buffer from shared guest memory, send the request buffer to the server and store the response into shared guest memory and notify TD guest by interrupt.

"quote-generation-service" is a property to specify Quote Generation Service(QGS) in qemu socket address format. The examples of the supported format are "vsock:2:1234", "unix:/run/qgs", "localhost:1234".

command line example:

  qemu-system-x86_64 \
    -object 'tdx-guest,id=tdx0,quote-generation-service=localhost:1234' \
    -machine confidential-guest-support=tdx0
---
 qapi/qom.json         |   5 +-
 target/i386/kvm/tdx.c | 380 ++++++++++++++++++++++++++++++++++++++++++
 target/i386/kvm/tdx.h |   7 +
 3 files changed, 391 insertions(+), 1 deletion(-)
qapi/qom.json +1 -0
index 87c1d440f331..37139949d761 100644
@@ -879,13 +879,16 @@
 #
 # @mrownerconfig: MROWNERCONFIG SHA384 hex string of 48 * 2 length (default: 0)
 #
+# @quote-generation-service: socket address for Quote Generation Service(QGS)

qapi/qom.json line 891

Quoted 10 lines
qapi/qom.json +3 -1
+#
 # Since: 8.2
 ##
 { 'struct': 'TdxGuestProperties',
   'data': { '*sept-ve-disable': 'bool',
             '*mrconfigid': 'str',
             '*mrowner': 'str',
-            '*mrownerconfig': 'str' } }
+            '*mrownerconfig': 'str',
+            '*quote-generation-service': 'str' } }

Why not type SocketAddress?

Quoted 5 lines
Quoted 3 lines

Yes, the code uses SocketAddress internally when it eventually calls qio_channel_socket_connect_async(), so we should directly use SocketAddress in the QAPI from the start.

Any benefit to directly use SocketAddress?

We don't want whatever code consumes the configuration to do a second level of parsing to convert the 'str' value into the 'SocketAddress' object it actually needs.

Quoted 5 lines

QEMU has a long history of having a second round of ad-hoc parsing of configuration and we've found it to be a serious maintenence burden. Thus we strive to have everything represented in QAPI using the desired final type, and avoid the second round of parsing.

Thanks for your explanation.

Quoted 7 lines
Quoted 6 lines

"quote-generation-service" here is optional, it seems not trivial to add and parse the SocketAddress type in QEMU command. After I change 'str' to 'SocketAddress' and specify the command like "-object tdx-guest,type=vsock,cid=2,port=1234...", it will report "invalid parameter cid".

Quoted 6 lines

The -object parameter supports JSON syntax for this reason

   -object '{"qom-type":"tdx-guest","quote-generation-service":{"type": "vsock", "cid":"2","port":"1234"}}'

libvirt will always use the JSON syntax for -object with a new enough QEMU.

The JSON syntax works for me. Then, we need to add some doc about using JSON syntax when quote-generation-service is required.

Quoted 3 lines

With regards,
Daniel

author
#91 lore raw

Re: [PATCH v2 41/58] i386/tdx: handle TDG.VP.VMCALL<GetQuote>

Quoted 84 lines · Chenyi Qiang
Quoted 64 lines · Daniel P. Berrangé
Quoted 53 lines · Chenyi Qiang
Quoted 50 lines · Daniel P. Berrangé
Quoted 48 lines · Markus Armbruster
Quoted 34 lines · Xiaoyao Li

From: Isaku Yamahata <[email protected]>

For GetQuote, delegate a request to Quote Generation Service. Add property of address of quote generation server and On request, connect to the server, read request buffer from shared guest memory, send the request buffer to the server and store the response into shared guest memory and notify TD guest by interrupt.

"quote-generation-service" is a property to specify Quote Generation Service(QGS) in qemu socket address format. The examples of the supported format are "vsock:2:1234", "unix:/run/qgs", "localhost:1234".

command line example:

   qemu-system-x86_64 \
     -object 'tdx-guest,id=tdx0,quote-generation-service=localhost:1234' \
     -machine confidential-guest-support=tdx0
---
  qapi/qom.json         |   5 +-
  target/i386/kvm/tdx.c | 380 ++++++++++++++++++++++++++++++++++++++++++
  target/i386/kvm/tdx.h |   7 +
  3 files changed, 391 insertions(+), 1 deletion(-)
qapi/qom.json +1 -0
index 87c1d440f331..37139949d761 100644
@@ -879,13 +879,16 @@
  #
  # @mrownerconfig: MROWNERCONFIG SHA384 hex string of 48 * 2 length (default: 0)
  #
+# @quote-generation-service: socket address for Quote Generation Service(QGS)

qapi/qom.json line 891

Quoted 10 lines
qapi/qom.json +3 -1
+#
  # Since: 8.2
  ##
  { 'struct': 'TdxGuestProperties',
    'data': { '*sept-ve-disable': 'bool',
              '*mrconfigid': 'str',
              '*mrowner': 'str',
-            '*mrownerconfig': 'str' } }
+            '*mrownerconfig': 'str',
+            '*quote-generation-service': 'str' } }

Why not type SocketAddress?

Quoted 5 lines
Quoted 3 lines

Yes, the code uses SocketAddress internally when it eventually calls qio_channel_socket_connect_async(), so we should directly use SocketAddress in the QAPI from the start.

Any benefit to directly use SocketAddress?

We don't want whatever code consumes the configuration to do a second level of parsing to convert the 'str' value into the 'SocketAddress' object it actually needs.

Quoted 5 lines

QEMU has a long history of having a second round of ad-hoc parsing of configuration and we've found it to be a serious maintenence burden. Thus we strive to have everything represented in QAPI using the desired final type, and avoid the second round of parsing.

Thanks for your explanation.

Quoted 9 lines
Quoted 6 lines

"quote-generation-service" here is optional, it seems not trivial to add and parse the SocketAddress type in QEMU command. After I change 'str' to 'SocketAddress' and specify the command like "-object tdx-guest,type=vsock,cid=2,port=1234...", it will report "invalid parameter cid".

The -object parameter supports JSON syntax for this reason

Quoted 7 lines
Quoted 4 lines
    -object '{"qom-type":"tdx-guest","quote-generation-service":{"type": "vsock", "cid":"2","port":"1234"}}'

libvirt will always use the JSON syntax for -object with a new enough QEMU.

The JSON syntax works for me. Then, we need to add some doc about using JSON syntax when quote-generation-service is required.

This limitation doesn't look reasonable to me.

@Daniel,

Is it acceptable by QEMU community?

Quoted 3 lines
Quoted 3 lines

With regards,
Daniel

#92 lore raw

Re: [PATCH v2 41/58] i386/tdx: handle TDG.VP.VMCALL<GetQuote>

Quoted 90 lines · Xiaoyao Li
Quoted 89 lines · Chenyi Qiang
Quoted 64 lines · Daniel P. Berrangé
Quoted 53 lines · Chenyi Qiang
Quoted 50 lines · Daniel P. Berrangé
Quoted 48 lines · Markus Armbruster
Quoted 34 lines · Xiaoyao Li

From: Isaku Yamahata <[email protected]>

For GetQuote, delegate a request to Quote Generation Service. Add property of address of quote generation server and On request, connect to the server, read request buffer from shared guest memory, send the request buffer to the server and store the response into shared guest memory and notify TD guest by interrupt.

"quote-generation-service" is a property to specify Quote Generation Service(QGS) in qemu socket address format. The examples of the supported format are "vsock:2:1234", "unix:/run/qgs", "localhost:1234".

command line example:

   qemu-system-x86_64 \
     -object 'tdx-guest,id=tdx0,quote-generation-service=localhost:1234' \
     -machine confidential-guest-support=tdx0
---
  qapi/qom.json         |   5 +-
  target/i386/kvm/tdx.c | 380 ++++++++++++++++++++++++++++++++++++++++++
  target/i386/kvm/tdx.h |   7 +
  3 files changed, 391 insertions(+), 1 deletion(-)
qapi/qom.json +1 -0
index 87c1d440f331..37139949d761 100644
@@ -879,13 +879,16 @@
  #
  # @mrownerconfig: MROWNERCONFIG SHA384 hex string of 48 * 2 length (default: 0)
  #
+# @quote-generation-service: socket address for Quote Generation Service(QGS)

qapi/qom.json line 891

Quoted 10 lines
qapi/qom.json +3 -1
+#
  # Since: 8.2
  ##
  { 'struct': 'TdxGuestProperties',
    'data': { '*sept-ve-disable': 'bool',
              '*mrconfigid': 'str',
              '*mrowner': 'str',
-            '*mrownerconfig': 'str' } }
+            '*mrownerconfig': 'str',
+            '*quote-generation-service': 'str' } }

Why not type SocketAddress?

Quoted 5 lines
Quoted 3 lines

Yes, the code uses SocketAddress internally when it eventually calls qio_channel_socket_connect_async(), so we should directly use SocketAddress in the QAPI from the start.

Any benefit to directly use SocketAddress?

We don't want whatever code consumes the configuration to do a second level of parsing to convert the 'str' value into the 'SocketAddress' object it actually needs.

Quoted 5 lines

QEMU has a long history of having a second round of ad-hoc parsing of configuration and we've found it to be a serious maintenence burden. Thus we strive to have everything represented in QAPI using the desired final type, and avoid the second round of parsing.

Thanks for your explanation.

Quoted 13 lines
Quoted 6 lines

"quote-generation-service" here is optional, it seems not trivial to add and parse the SocketAddress type in QEMU command. After I change 'str' to 'SocketAddress' and specify the command like "-object tdx-guest,type=vsock,cid=2,port=1234...", it will report "invalid parameter cid".

The -object parameter supports JSON syntax for this reason

    -object '{"qom-type":"tdx-guest","quote-generation-service":{"type": "vsock", "cid":"2","port":"1234"}}'

libvirt will always use the JSON syntax for -object with a new enough QEMU.

Quoted 8 lines
Quoted 2 lines

The JSON syntax works for me. Then, we need to add some doc about using JSON syntax when quote-generation-service is required.

This limitation doesn't look reasonable to me.

@Daniel,

Is it acceptable by QEMU community?

This is the expected approach for object types which have non-scalar properties.

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
author
#93 lore raw

Re: [PATCH v2 41/58] i386/tdx: handle TDG.VP.VMCALL<GetQuote>

Quoted 94 lines · Daniel P. Berrangé
Quoted 93 lines · Xiaoyao Li
Quoted 91 lines · Chenyi Qiang
Quoted 64 lines · Daniel P. Berrangé
Quoted 53 lines · Chenyi Qiang
Quoted 50 lines · Daniel P. Berrangé
Quoted 48 lines · Markus Armbruster
Quoted 34 lines · Xiaoyao Li

From: Isaku Yamahata <[email protected]>

For GetQuote, delegate a request to Quote Generation Service. Add property of address of quote generation server and On request, connect to the server, read request buffer from shared guest memory, send the request buffer to the server and store the response into shared guest memory and notify TD guest by interrupt.

"quote-generation-service" is a property to specify Quote Generation Service(QGS) in qemu socket address format. The examples of the supported format are "vsock:2:1234", "unix:/run/qgs", "localhost:1234".

command line example:

    qemu-system-x86_64 \
      -object 'tdx-guest,id=tdx0,quote-generation-service=localhost:1234' \
      -machine confidential-guest-support=tdx0
---
   qapi/qom.json         |   5 +-
   target/i386/kvm/tdx.c | 380 ++++++++++++++++++++++++++++++++++++++++++
   target/i386/kvm/tdx.h |   7 +
   3 files changed, 391 insertions(+), 1 deletion(-)
qapi/qom.json +1 -0
index 87c1d440f331..37139949d761 100644
@@ -879,13 +879,16 @@
   #
   # @mrownerconfig: MROWNERCONFIG SHA384 hex string of 48 * 2 length (default: 0)
   #
+# @quote-generation-service: socket address for Quote Generation Service(QGS)

qapi/qom.json line 891

Quoted 10 lines
qapi/qom.json +3 -1
+#
   # Since: 8.2
   ##
   { 'struct': 'TdxGuestProperties',
     'data': { '*sept-ve-disable': 'bool',
               '*mrconfigid': 'str',
               '*mrowner': 'str',
-            '*mrownerconfig': 'str' } }
+            '*mrownerconfig': 'str',
+            '*quote-generation-service': 'str' } }

Why not type SocketAddress?

Quoted 5 lines
Quoted 3 lines

Yes, the code uses SocketAddress internally when it eventually calls qio_channel_socket_connect_async(), so we should directly use SocketAddress in the QAPI from the start.

Any benefit to directly use SocketAddress?

We don't want whatever code consumes the configuration to do a second level of parsing to convert the 'str' value into the 'SocketAddress' object it actually needs.

Quoted 5 lines

QEMU has a long history of having a second round of ad-hoc parsing of configuration and we've found it to be a serious maintenence burden. Thus we strive to have everything represented in QAPI using the desired final type, and avoid the second round of parsing.

Thanks for your explanation.

Quoted 7 lines
Quoted 6 lines

"quote-generation-service" here is optional, it seems not trivial to add and parse the SocketAddress type in QEMU command. After I change 'str' to 'SocketAddress' and specify the command like "-object tdx-guest,type=vsock,cid=2,port=1234...", it will report "invalid parameter cid".

Quoted 6 lines

The -object parameter supports JSON syntax for this reason

     -object '{"qom-type":"tdx-guest","quote-generation-service":{"type": "vsock", "cid":"2","port":"1234"}}'

libvirt will always use the JSON syntax for -object with a new enough QEMU.

The JSON syntax works for me. Then, we need to add some doc about using JSON syntax when quote-generation-service is required.

Quoted 8 lines
Quoted 5 lines

This limitation doesn't look reasonable to me.

@Daniel,

Is it acceptable by QEMU community?

This is the expected approach for object types which have non-scalar properties.

Learned it.

thanks!

Quoted 2 lines

With regards,
Daniel

#94 lore raw

Re: [PATCH v2 41/58] i386/tdx: handle TDG.VP.VMCALL<GetQuote>

In reply to Chenyi Qiang

I sent this reply to your question on the same day, but it got eaten by malfunctioning servers, and I noticed only now after another failure made me dig through my logs. Sorry for the inconvenience!

Quoted 51 lines · Chenyi Qiang
Quoted 50 lines · Daniel P. Berrangé
Quoted 48 lines · Markus Armbruster
Quoted 34 lines · Xiaoyao Li

From: Isaku Yamahata <[email protected]>

For GetQuote, delegate a request to Quote Generation Service. Add property of address of quote generation server and On request, connect to the server, read request buffer from shared guest memory, send the request buffer to the server and store the response into shared guest memory and notify TD guest by interrupt.

"quote-generation-service" is a property to specify Quote Generation Service(QGS) in qemu socket address format. The examples of the supported format are "vsock:2:1234", "unix:/run/qgs", "localhost:1234".

command line example:

  qemu-system-x86_64 \
    -object 'tdx-guest,id=tdx0,quote-generation-service=localhost:1234' \
    -machine confidential-guest-support=tdx0
---
 qapi/qom.json         |   5 +-
 target/i386/kvm/tdx.c | 380 ++++++++++++++++++++++++++++++++++++++++++
 target/i386/kvm/tdx.h |   7 +
 3 files changed, 391 insertions(+), 1 deletion(-)
qapi/qom.json +1 -0
index 87c1d440f331..37139949d761 100644
@@ -879,13 +879,16 @@
 #
 # @mrownerconfig: MROWNERCONFIG SHA384 hex string of 48 * 2 length (default: 0)
 #
+# @quote-generation-service: socket address for Quote Generation Service(QGS)

qapi/qom.json line 891

Quoted 10 lines
qapi/qom.json +3 -1
+#
 # Since: 8.2
 ##
 { 'struct': 'TdxGuestProperties',
   'data': { '*sept-ve-disable': 'bool',
             '*mrconfigid': 'str',
             '*mrowner': 'str',
-            '*mrownerconfig': 'str' } }
+            '*mrownerconfig': 'str',
+            '*quote-generation-service': 'str' } }

Why not type SocketAddress?

Quoted 5 lines
Quoted 3 lines

Yes, the code uses SocketAddress internally when it eventually calls qio_channel_socket_connect_async(), so we should directly use SocketAddress in the QAPI from the start.

Any benefit to directly use SocketAddress?

Design principle: use JSON to encode structured data as text in QAPI/QMP.

Do: "mumble": [1, 2, 3]

Don't: "mumble": "1,2,3"

Do: "server": { "type": "inet", "host": "localhost", "port": "12345" }

Don't: "server": "host=localhost,port=12345"

We violate the principle in a couple of places. Some are arguably mistakes, some are pragmatic exceptions.

The principle implies "the only parser QAPI needs is the JSON parser".

The other benefit is consistency with existing interfaces. They use SocketAddress (a few old ones use SocketAddressLegacy).

Quoted 5 lines

"quote-generation-service" here is optional, it seems not trivial to add and parse the SocketAddress type in QEMU command. After I change 'str' to 'SocketAddress' and specify the command like "-object tdx-guest,type=vsock,cid=2,port=1234...", it will report "invalid parameter cid".

Try "quote-generation-service.port=1234".

author
#95 lore raw

PATCH v2 42/58 i386/tdx: register the fd read callback with the main loop to read the quote data

In reply to Xiaoyao Li

From: Chenyi Qiang <[email protected]>

When TD guest invokes getquote tdvmcall, QEMU will register a async qio task with default context when the qio channel is connected. However, as there is a blocking action (recvmsg()) in qio_channel_read() and it will block main thread and make TD guest have no response until the server returns.

Set the io channel non-blocking and register the socket fd with the main loop. Move the read operation into the callback. When the fd is readable, inovke the callback to handle the quote data.

---
 target/i386/kvm/tdx.c | 147 +++++++++++++++++++++++++++---------------
 1 file changed, 96 insertions(+), 51 deletions(-)
target/i386/kvm/tdx.c +96 -51
index 73d6cd88af9e..3cb2163a0335 100644
@@ -997,6 +997,8 @@ struct tdx_get_quote_task {
     uint32_t apic_id;
     hwaddr gpa;
     uint64_t buf_len;
+    char *out_data;
+    uint64_t out_len;
     struct tdx_get_quote_header hdr;
     int event_notify_interrupt;
     QIOChannelSocket *ioc;
@@ -1082,78 +1084,53 @@ static void tdx_td_notify(struct tdx_get_quote_task *t)
     }
 }
 
-/*
- * TODO: If QGS doesn't reply for long time, make it an error and interrupt
- * guest.
- */
-static void tdx_handle_get_quote_connected(QIOTask *task, gpointer opaque)
+static void tdx_get_quote_read(void *opaque)
 {
     struct tdx_get_quote_task *t = opaque;
+    ssize_t size = 0;
     Error *err = NULL;
-    char *in_data = NULL;
-    char *out_data = NULL;
-    size_t out_len;
-    ssize_t size;
     MachineState *ms;
     TdxGuest *tdx;
 
-    t->hdr.error_code = cpu_to_le64(TDX_VP_GET_QUOTE_ERROR);
-    if (qio_task_propagate_error(task, NULL)) {
-        t->hdr.error_code = cpu_to_le64(TDX_VP_GET_QUOTE_QGS_UNAVAILABLE);
-        goto error;
-    }
-
-    in_data = g_malloc(le32_to_cpu(t->hdr.in_len));
-    if (address_space_read(&address_space_memory, t->gpa + sizeof(t->hdr),
-                           MEMTXATTRS_UNSPECIFIED, in_data,
-                           le32_to_cpu(t->hdr.in_len)) != MEMTX_OK) {
-        goto error;
-    }
-
-    if (qio_channel_write_all(QIO_CHANNEL(t->ioc), in_data,
-                              le32_to_cpu(t->hdr.in_len), &err) ||
-        err) {
-        t->hdr.error_code = cpu_to_le64(TDX_VP_GET_QUOTE_QGS_UNAVAILABLE);
-        goto error;
-    }
-
-    out_data = g_malloc(t->buf_len);
-    out_len = 0;
-    size = 0;
     while (true) {
         char *buf;
         size_t buf_size;
 
-        if (out_len < t->buf_len) {
-            buf = out_data + out_len;
-            buf_size = t->buf_len - out_len;
+        if (t->out_len < t->buf_len) {
+            buf = t->out_data + t->out_len;
+            buf_size = t->buf_len - t->out_len;
         } else {
             /*
              * The received data is too large to fit in the shared GPA.
              * Discard the received data and try to know the data size.
              */
-            buf = out_data;
+            buf = t->out_data;
             buf_size = t->buf_len;
         }
 
         size = qio_channel_read(QIO_CHANNEL(t->ioc), buf, buf_size, &err);
-        if (err) {
+        if (!size) {
             break;
         }
-        if (size <= 0) {
-            break;
+
+        if (size < 0) {
+            if (size == QIO_CHANNEL_ERR_BLOCK) {
+                return;
+            } else {
+                break;
+            }
         }
-        out_len += size;
+        t->out_len += size;
     }
     /*
-     * Treat partial read as success and let the QGS client to handle it because
-     * the client knows better about the QGS.
+     * If partial read successfully but return error at last, also treat it
+     * as failure.
      */
-    if (out_len == 0 && (err || size < 0)) {
+    if (size < 0) {
         t->hdr.error_code = cpu_to_le64(TDX_VP_GET_QUOTE_QGS_UNAVAILABLE);
         goto error;
     }
-    if (out_len > 0 && out_len > t->buf_len) {
+    if (t->out_len > 0 && t->out_len > t->buf_len) {
         /*
          * There is no specific error code defined for this case(E2BIG) at the
          * moment.
@@ -1161,20 +1138,20 @@ static void tdx_handle_get_quote_connected(QIOTask *task, gpointer opaque)
          * update the error code.
          */
         t->hdr.error_code = cpu_to_le64(TDX_VP_GET_QUOTE_ERROR);
-        t->hdr.out_len = cpu_to_le32(out_len);
+        t->hdr.out_len = cpu_to_le32(t->out_len);
         goto error_hdr;
     }
 
     if (address_space_write(
             &address_space_memory, t->gpa + sizeof(t->hdr),
-            MEMTXATTRS_UNSPECIFIED, out_data, out_len) != MEMTX_OK) {
+            MEMTXATTRS_UNSPECIFIED, t->out_data, t->out_len) != MEMTX_OK) {
         goto error;
     }
     /*
      * Even if out_len == 0, it's a success.  It's up to the QGS-client contract
      * how to interpret the zero-sized message as return message.
      */
-    t->hdr.out_len = cpu_to_le32(out_len);
+    t->hdr.out_len = cpu_to_le32(t->out_len);
     t->hdr.error_code = cpu_to_le64(TDX_VP_GET_QUOTE_SUCCESS);
 
 error:
@@ -1185,14 +1162,15 @@ error_hdr:
     if (address_space_write(
             &address_space_memory, t->gpa,
             MEMTXATTRS_UNSPECIFIED, &t->hdr, sizeof(t->hdr)) != MEMTX_OK) {
-        error_report("TDX: failed to updsate GetQuote header.\n");
+        error_report("TDX: failed to update GetQuote header.");
     }
     tdx_td_notify(t);
 
+    qemu_set_fd_handler(t->ioc->fd, NULL, NULL, NULL);
     qio_channel_close(QIO_CHANNEL(t->ioc), &err);
     object_unref(OBJECT(t->ioc));
-    g_free(in_data);
-    g_free(out_data);
+    g_free(t->out_data);
+    g_free(t);
 
     /* Maintain the number of in-flight requests. */
     ms = MACHINE(qdev_get_machine());
@@ -1200,7 +1178,71 @@ error_hdr:
     qemu_mutex_lock(&tdx->lock);
     tdx->quote_generation_num--;
     qemu_mutex_unlock(&tdx->lock);
+}
+
+/*
+ * TODO: If QGS doesn't reply for long time, make it an error and interrupt
+ * guest.
+ */
+static void tdx_handle_get_quote_connected(QIOTask *task, gpointer opaque)
+{
+    struct tdx_get_quote_task *t = opaque;
+    Error *err = NULL;
+    char *in_data = NULL;
+    MachineState *ms;
+    TdxGuest *tdx;
+
+    t->hdr.error_code = cpu_to_le64(TDX_VP_GET_QUOTE_ERROR);
+    if (qio_task_propagate_error(task, NULL)) {
+        t->hdr.error_code = cpu_to_le64(TDX_VP_GET_QUOTE_QGS_UNAVAILABLE);
+        goto error;
+    }
+
+    in_data = g_malloc(le32_to_cpu(t->hdr.in_len));
+    if (!in_data) {
+        goto error;
+    }
+
+    if (address_space_read(&address_space_memory, t->gpa + sizeof(t->hdr),
+                           MEMTXATTRS_UNSPECIFIED, in_data,
+                           le32_to_cpu(t->hdr.in_len)) != MEMTX_OK) {
+        goto error;
+    }
+
+    qio_channel_set_blocking(QIO_CHANNEL(t->ioc), false, NULL);
+
+    if (qio_channel_write_all(QIO_CHANNEL(t->ioc), in_data,
+                              le32_to_cpu(t->hdr.in_len), &err) ||
+        err) {
+        t->hdr.error_code = cpu_to_le64(TDX_VP_GET_QUOTE_QGS_UNAVAILABLE);
+        goto error;
+    }
+
+    g_free(in_data);
+    qemu_set_fd_handler(t->ioc->fd, tdx_get_quote_read, NULL, t);
+
+    return;
+error:
+    t->hdr.out_len = cpu_to_le32(0);
 
+    if (address_space_write(
+            &address_space_memory, t->gpa,
+            MEMTXATTRS_UNSPECIFIED, &t->hdr, sizeof(t->hdr)) != MEMTX_OK) {
+        error_report("TDX: failed to update GetQuote header.\n");
+    }
+    tdx_td_notify(t);
+
+    qio_channel_close(QIO_CHANNEL(t->ioc), &err);
+    object_unref(OBJECT(t->ioc));
+    g_free(t);
+    g_free(in_data);
+
+    /* Maintain the number of in-flight requests. */
+    ms = MACHINE(qdev_get_machine());
+    tdx = TDX_GUEST(ms->cgs);
+    qemu_mutex_lock(&tdx->lock);
+    tdx->quote_generation_num--;
+    qemu_mutex_unlock(&tdx->lock);
     return;
 }
 
@@ -1269,6 +1311,8 @@ static void tdx_handle_get_quote(X86CPU *cpu, struct kvm_tdx_vmcall *vmcall)
     t->apic_id = tdx->event_notify_apic_id;
     t->gpa = gpa;
     t->buf_len = buf_len;
+    t->out_data = g_malloc(t->buf_len);
+    t->out_len = 0;
     t->hdr = hdr;
     t->ioc = ioc;
 
@@ -1279,13 +1323,14 @@ static void tdx_handle_get_quote(X86CPU *cpu, struct kvm_tdx_vmcall *vmcall)
         qemu_mutex_unlock(&tdx->lock);
         vmcall->status_code = TDG_VP_VMCALL_RETRY;
         object_unref(OBJECT(ioc));
+        g_free(t->out_data);
         g_free(t);
         return;
     }
     tdx->quote_generation_num++;
     t->event_notify_interrupt = tdx->event_notify_interrupt;
     qio_channel_socket_connect_async(
-        ioc, tdx->quote_generation, tdx_handle_get_quote_connected, t, g_free,
+        ioc, tdx->quote_generation, tdx_handle_get_quote_connected, t, NULL,
         NULL);
     qemu_mutex_unlock(&tdx->lock);
 
Signature · 1 line
2.34.1
#96 lore raw

Re: [PATCH v2 42/58] i386/tdx: register the fd read callback with the main loop to read the quote data

Quoted 12 lines · Xiaoyao Li

From: Chenyi Qiang <[email protected]>

When TD guest invokes getquote tdvmcall, QEMU will register a async qio task with default context when the qio channel is connected. However, as there is a blocking action (recvmsg()) in qio_channel_read() and it will block main thread and make TD guest have no response until the server returns.

Set the io channel non-blocking and register the socket fd with the main loop. Move the read operation into the callback. When the fd is readable, inovke the callback to handle the quote data.

Quoted 6 lines
---
 target/i386/kvm/tdx.c | 147 +++++++++++++++++++++++++++---------------
 1 file changed, 96 insertions(+), 51 deletions(-)

How about squashing this patch with the previous one? I think this patch is somewhat a bug fix for it to resolve the thread blocking issue.

author
#97 lore raw

PATCH v2 43/58 i386/tdx: setup a timer for the qio channel

In reply to Xiaoyao Li

From: Chenyi Qiang <[email protected]>

To avoid no response from QGS server, setup a timer for the transaction. If timeout, make it an error and interrupt guest. Define the threshold of time to 30s at present, maybe change to other value if not appropriate.

Extract the common cleanup code to make it more clear.

---
 target/i386/kvm/tdx.c | 151 ++++++++++++++++++++++++------------------
 1 file changed, 85 insertions(+), 66 deletions(-)
target/i386/kvm/tdx.c +85 -66
index 3cb2163a0335..fa658ce1f2e4 100644
@@ -1002,6 +1002,7 @@ struct tdx_get_quote_task {
     struct tdx_get_quote_header hdr;
     int event_notify_interrupt;
     QIOChannelSocket *ioc;
+    QEMUTimer timer;
 };
 
 struct x86_msi {
@@ -1084,13 +1085,48 @@ static void tdx_td_notify(struct tdx_get_quote_task *t)
     }
 }
 
+static void tdx_getquote_task_cleanup(struct tdx_get_quote_task *t, bool outlen_overflow)
+{
+    MachineState *ms;
+    TdxGuest *tdx;
+
+    if (t->hdr.error_code != cpu_to_le64(TDX_VP_GET_QUOTE_SUCCESS) && !outlen_overflow) {
+        t->hdr.out_len = cpu_to_le32(0);
+    }
+
+    /* Publish the response contents before marking this request completed. */
+    smp_wmb();
+    if (address_space_write(
+            &address_space_memory, t->gpa,
+            MEMTXATTRS_UNSPECIFIED, &t->hdr, sizeof(t->hdr)) != MEMTX_OK) {
+        error_report("TDX: failed to update GetQuote header.");
+    }
+    tdx_td_notify(t);
+
+    if (t->ioc->fd > 0) {
+        qemu_set_fd_handler(t->ioc->fd, NULL, NULL, NULL);
+    }
+    qio_channel_close(QIO_CHANNEL(t->ioc), NULL);
+    object_unref(OBJECT(t->ioc));
+    timer_del(&t->timer);
+    g_free(t->out_data);
+    g_free(t);
+
+    /* Maintain the number of in-flight requests. */
+    ms = MACHINE(qdev_get_machine());
+    tdx = TDX_GUEST(ms->cgs);
+    qemu_mutex_lock(&tdx->lock);
+    tdx->quote_generation_num--;
+    qemu_mutex_unlock(&tdx->lock);
+}
+
+
 static void tdx_get_quote_read(void *opaque)
 {
     struct tdx_get_quote_task *t = opaque;
     ssize_t size = 0;
     Error *err = NULL;
-    MachineState *ms;
-    TdxGuest *tdx;
+    bool outlen_overflow = false;
 
     while (true) {
         char *buf;
@@ -1135,11 +1171,12 @@ static void tdx_get_quote_read(void *opaque)
          * There is no specific error code defined for this case(E2BIG) at the
          * moment.
          * TODO: Once an error code for this case is defined in GHCI spec ,
-         * update the error code.
+         * update the error code and the tdx_getquote_task_cleanup() argument.
          */
         t->hdr.error_code = cpu_to_le64(TDX_VP_GET_QUOTE_ERROR);
         t->hdr.out_len = cpu_to_le32(t->out_len);
-        goto error_hdr;
+        outlen_overflow = true;
+        goto error;
     }
 
     if (address_space_write(
@@ -1155,94 +1192,76 @@ static void tdx_get_quote_read(void *opaque)
     t->hdr.error_code = cpu_to_le64(TDX_VP_GET_QUOTE_SUCCESS);
 
 error:
-    if (t->hdr.error_code != cpu_to_le64(TDX_VP_GET_QUOTE_SUCCESS)) {
-        t->hdr.out_len = cpu_to_le32(0);
-    }
-error_hdr:
-    if (address_space_write(
-            &address_space_memory, t->gpa,
-            MEMTXATTRS_UNSPECIFIED, &t->hdr, sizeof(t->hdr)) != MEMTX_OK) {
-        error_report("TDX: failed to update GetQuote header.");
-    }
-    tdx_td_notify(t);
+    tdx_getquote_task_cleanup(t, outlen_overflow);
+}
+
+#define TRANSACTION_TIMEOUT 30000
+
+static void getquote_timer_expired(void *opaque)
+{
+    struct tdx_get_quote_task *t = opaque;
+
+    tdx_getquote_task_cleanup(t, false);
+}
 
-    qemu_set_fd_handler(t->ioc->fd, NULL, NULL, NULL);
-    qio_channel_close(QIO_CHANNEL(t->ioc), &err);
-    object_unref(OBJECT(t->ioc));
-    g_free(t->out_data);
-    g_free(t);
+static void tdx_transaction_start(struct tdx_get_quote_task *t)
+{
+    int64_t time;
 
-    /* Maintain the number of in-flight requests. */
-    ms = MACHINE(qdev_get_machine());
-    tdx = TDX_GUEST(ms->cgs);
-    qemu_mutex_lock(&tdx->lock);
-    tdx->quote_generation_num--;
-    qemu_mutex_unlock(&tdx->lock);
+    time = qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL);
+    /*
+     * Timeout callback and fd callback both run in main loop thread,
+     * thus no need to worry about race condition.
+     */
+    qemu_set_fd_handler(t->ioc->fd, tdx_get_quote_read, NULL, t);
+    timer_init_ms(&t->timer, QEMU_CLOCK_VIRTUAL, getquote_timer_expired, t);
+    timer_mod(&t->timer, time + TRANSACTION_TIMEOUT);
 }
 
-/*
- * TODO: If QGS doesn't reply for long time, make it an error and interrupt
- * guest.
- */
 static void tdx_handle_get_quote_connected(QIOTask *task, gpointer opaque)
 {
     struct tdx_get_quote_task *t = opaque;
     Error *err = NULL;
     char *in_data = NULL;
-    MachineState *ms;
-    TdxGuest *tdx;
+    int ret = 0;
 
     t->hdr.error_code = cpu_to_le64(TDX_VP_GET_QUOTE_ERROR);
-    if (qio_task_propagate_error(task, NULL)) {
+    ret = qio_task_propagate_error(task, NULL);
+    if (ret) {
         t->hdr.error_code = cpu_to_le64(TDX_VP_GET_QUOTE_QGS_UNAVAILABLE);
-        goto error;
+        goto out;
     }
 
     in_data = g_malloc(le32_to_cpu(t->hdr.in_len));
     if (!in_data) {
-        goto error;
+        ret = -1;
+        goto out;
     }
 
-    if (address_space_read(&address_space_memory, t->gpa + sizeof(t->hdr),
-                           MEMTXATTRS_UNSPECIFIED, in_data,
-                           le32_to_cpu(t->hdr.in_len)) != MEMTX_OK) {
-        goto error;
+    ret = address_space_read(&address_space_memory, t->gpa + sizeof(t->hdr),
+                             MEMTXATTRS_UNSPECIFIED, in_data,
+                             le32_to_cpu(t->hdr.in_len));
+    if (ret) {
+        g_free(in_data);
+        goto out;
     }
 
     qio_channel_set_blocking(QIO_CHANNEL(t->ioc), false, NULL);
 
-    if (qio_channel_write_all(QIO_CHANNEL(t->ioc), in_data,
-                              le32_to_cpu(t->hdr.in_len), &err) ||
-        err) {
+    ret = qio_channel_write_all(QIO_CHANNEL(t->ioc), in_data,
+                              le32_to_cpu(t->hdr.in_len), &err);
+    if (ret) {
         t->hdr.error_code = cpu_to_le64(TDX_VP_GET_QUOTE_QGS_UNAVAILABLE);
-        goto error;
+        g_free(in_data);
+        goto out;
     }
 
-    g_free(in_data);
-    qemu_set_fd_handler(t->ioc->fd, tdx_get_quote_read, NULL, t);
-
-    return;
-error:
-    t->hdr.out_len = cpu_to_le32(0);
-
-    if (address_space_write(
-            &address_space_memory, t->gpa,
-            MEMTXATTRS_UNSPECIFIED, &t->hdr, sizeof(t->hdr)) != MEMTX_OK) {
-        error_report("TDX: failed to update GetQuote header.\n");
+out:
+    if (ret) {
+        tdx_getquote_task_cleanup(t, false);
+    } else {
+        tdx_transaction_start(t);
     }
-    tdx_td_notify(t);
-
-    qio_channel_close(QIO_CHANNEL(t->ioc), &err);
-    object_unref(OBJECT(t->ioc));
-    g_free(t);
-    g_free(in_data);
-
-    /* Maintain the number of in-flight requests. */
-    ms = MACHINE(qdev_get_machine());
-    tdx = TDX_GUEST(ms->cgs);
-    qemu_mutex_lock(&tdx->lock);
-    tdx->quote_generation_num--;
-    qemu_mutex_unlock(&tdx->lock);
     return;
 }
 
Signature · 1 line
2.34.1
#98 lore raw

Re: [PATCH v2 43/58] i386/tdx: setup a timer for the qio channel

Quoted 44 lines · Xiaoyao Li

From: Chenyi Qiang <[email protected]>

To avoid no response from QGS server, setup a timer for the transaction. If timeout, make it an error and interrupt guest. Define the threshold of time to 30s at present, maybe change to other value if not appropriate.

Extract the common cleanup code to make it more clear.

---
 target/i386/kvm/tdx.c | 151 ++++++++++++++++++++++++------------------
 1 file changed, 85 insertions(+), 66 deletions(-)
target/i386/kvm/tdx.c +15 -0
index 3cb2163a0335..fa658ce1f2e4 100644
@@ -1002,6 +1002,7 @@ struct tdx_get_quote_task {
     struct tdx_get_quote_header hdr;
     int event_notify_interrupt;
     QIOChannelSocket *ioc;
+    QEMUTimer timer;
 };
 
 struct x86_msi {
@@ -1084,13 +1085,48 @@ static void tdx_td_notify(struct tdx_get_quote_task *t)
     }
 }
 
+static void tdx_getquote_task_cleanup(struct tdx_get_quote_task *t, bool outlen_overflow)
+{
+    MachineState *ms;
+    TdxGuest *tdx;
+
+    if (t->hdr.error_code != cpu_to_le64(TDX_VP_GET_QUOTE_SUCCESS) && !outlen_overflow) {
+        t->hdr.out_len = cpu_to_le32(0);
+    }
+
+    /* Publish the response contents before marking this request completed. */
+    smp_wmb();
+    if (address_space_write(
+            &address_space_memory, t->gpa,
+            MEMTXATTRS_UNSPECIFIED, &t->hdr, sizeof(t->hdr)) != MEMTX_OK) {

target/i386/kvm/tdx.c static void tdx_td_notify(struct tdx_get_quote_task *t) · line 1111

Quoted 10 lines
target/i386/kvm/tdx.c +10 -0
+        error_report("TDX: failed to update GetQuote header.");
+    }
+    tdx_td_notify(t);
+
+    if (t->ioc->fd > 0) {
+        qemu_set_fd_handler(t->ioc->fd, NULL, NULL, NULL);
+    }
+    qio_channel_close(QIO_CHANNEL(t->ioc), NULL);
+    object_unref(OBJECT(t->ioc));
+    timer_del(&t->timer);

Xiaoyao, I guess you missed a bug fix patch here as t->timer could be uninitialized and then timer_del() will cause segv.

Quoted 11 lines
target/i386/kvm/tdx.c +11 -0
+    g_free(t->out_data);
+    g_free(t);
+
+    /* Maintain the number of in-flight requests. */
+    ms = MACHINE(qdev_get_machine());
+    tdx = TDX_GUEST(ms->cgs);
+    qemu_mutex_lock(&tdx->lock);
+    tdx->quote_generation_num--;
+    qemu_mutex_unlock(&tdx->lock);
+}
+
author
#99 lore raw

Re: [PATCH v2 43/58] i386/tdx: setup a timer for the qio channel

Quoted 52 lines · Chenyi Qiang
Quoted 49 lines · Xiaoyao Li

From: Chenyi Qiang <[email protected]>

To avoid no response from QGS server, setup a timer for the transaction. If timeout, make it an error and interrupt guest. Define the threshold of time to 30s at present, maybe change to other value if not appropriate.

Extract the common cleanup code to make it more clear.

---
  target/i386/kvm/tdx.c | 151 ++++++++++++++++++++++++------------------
  1 file changed, 85 insertions(+), 66 deletions(-)
target/i386/kvm/tdx.c +20 -0
index 3cb2163a0335..fa658ce1f2e4 100644
@@ -1002,6 +1002,7 @@ struct tdx_get_quote_task {
      struct tdx_get_quote_header hdr;
      int event_notify_interrupt;
      QIOChannelSocket *ioc;
+    QEMUTimer timer;
  };
 
  struct x86_msi {
@@ -1084,13 +1085,48 @@ static void tdx_td_notify(struct tdx_get_quote_task *t)
      }
  }
 
+static void tdx_getquote_task_cleanup(struct tdx_get_quote_task *t, bool outlen_overflow)
+{
+    MachineState *ms;
+    TdxGuest *tdx;
+
+    if (t->hdr.error_code != cpu_to_le64(TDX_VP_GET_QUOTE_SUCCESS) && !outlen_overflow) {
+        t->hdr.out_len = cpu_to_le32(0);
+    }
+
+    /* Publish the response contents before marking this request completed. */
+    smp_wmb();
+    if (address_space_write(
+            &address_space_memory, t->gpa,
+            MEMTXATTRS_UNSPECIFIED, &t->hdr, sizeof(t->hdr)) != MEMTX_OK) {
+        error_report("TDX: failed to update GetQuote header.");
+    }
+    tdx_td_notify(t);
+
+    if (t->ioc->fd > 0) {
Quoted 8 lines

target/i386/kvm/tdx.c static void tdx_td_notify(struct tdx_get_quote_task *t) · line 1111

Quoted 5 lines
target/i386/kvm/tdx.c +5 -0
+        qemu_set_fd_handler(t->ioc->fd, NULL, NULL, NULL);
+    }
+    qio_channel_close(QIO_CHANNEL(t->ioc), NULL);
+    object_unref(OBJECT(t->ioc));
+    timer_del(&t->timer);

Xiaoyao, I guess you missed a bug fix patch here as t->timer could be uninitialized and then timer_del() will cause segv.

Thanks for the reminding.
I'll update this patch to include the fix.

Thanks,
-Xiaoyao

Quoted 12 lines
Quoted 11 lines
target/i386/kvm/tdx.c +11 -0
+    g_free(t->out_data);
+    g_free(t);
+
+    /* Maintain the number of in-flight requests. */
+    ms = MACHINE(qdev_get_machine());
+    tdx = TDX_GUEST(ms->cgs);
+    qemu_mutex_lock(&tdx->lock);
+    tdx->quote_generation_num--;
+    qemu_mutex_unlock(&tdx->lock);
+}
+
author
#100 lore raw

PATCH v2 44/58 i386/tdx: handle TDG.VP.VMCALL<MapGPA> hypercall

In reply to Xiaoyao Li

From: Isaku Yamahata <[email protected]>

MapGPA is a hypercall to convert GPA from/to private GPA to/from shared GPA. As the conversion function is already implemented as kvm_convert_memory, wire it to TDX hypercall exit.

---
 accel/kvm/kvm-all.c   |  2 +-
 include/sysemu/kvm.h  |  2 ++
 target/i386/kvm/tdx.c | 37 +++++++++++++++++++++++++++++++++++++
 3 files changed, 40 insertions(+), 1 deletion(-)
accel/kvm/kvm-all.c +1 -1
index 8d53c89e9dbf..b0a18800fd63 100644
@@ -3059,7 +3059,7 @@ static void kvm_eat_signals(CPUState *cpu)
     } while (sigismember(&chkset, SIG_IPI));
 }
 
-static int kvm_convert_memory(hwaddr start, hwaddr size, bool to_private)
+int kvm_convert_memory(hwaddr start, hwaddr size, bool to_private)
 {
     MemoryRegionSection section;
     void *addr;
include/sysemu/kvm.h +2 -0
index d89ec87072d7..7e0f56d1c5c5 100644
@@ -584,4 +584,6 @@ uint32_t kvm_dirty_ring_size(void);
 
 int kvm_set_memory_attributes_private(hwaddr start, hwaddr size);
 int kvm_set_memory_attributes_shared(hwaddr start, hwaddr size);
+
+int kvm_convert_memory(hwaddr start, hwaddr size, bool to_private);
 #endif
target/i386/kvm/tdx.c +37 -0
index fa658ce1f2e4..0c43c1f7759f 100644
@@ -934,6 +934,7 @@ static void tdx_guest_class_init(ObjectClass *oc, void *data)
 {
 }
 
+#define TDG_VP_VMCALL_MAP_GPA                           0x10001ULL
 #define TDG_VP_VMCALL_GET_QUOTE                         0x10002ULL
 #define TDG_VP_VMCALL_SETUP_EVENT_NOTIFY_INTERRUPT      0x10004ULL
 
@@ -993,6 +994,39 @@ static hwaddr tdx_shared_bit(X86CPU *cpu)
     return (cpu->phys_bits > 48) ? BIT_ULL(51) : BIT_ULL(47);
 }
 
+static void tdx_handle_map_gpa(X86CPU *cpu, struct kvm_tdx_vmcall *vmcall)
+{
+    hwaddr shared_bit = tdx_shared_bit(cpu);
+    hwaddr gpa = vmcall->in_r12 & ~shared_bit;
+    bool private = !(vmcall->in_r12 & shared_bit);
+    hwaddr size = vmcall->in_r13;
+    int ret = 0;
+
+    vmcall->status_code = TDG_VP_VMCALL_INVALID_OPERAND;
+
+    if (!QEMU_IS_ALIGNED(gpa, 4096) || !QEMU_IS_ALIGNED(size, 4096)) {
+        vmcall->status_code = TDG_VP_VMCALL_ALIGN_ERROR;
+        return;
+    }
+
+    /* Overflow case. */
+    if (gpa + size < gpa) {
+        return;
+    }
+    if (gpa >= (1ULL << cpu->phys_bits) ||
+        gpa + size >= (1ULL << cpu->phys_bits)) {
+        return;
+    }
+
+    if (size > 0) {
+        ret = kvm_convert_memory(gpa, size, private);
+    }
+
+    if (!ret) {
+        vmcall->status_code = TDG_VP_VMCALL_SUCCESS;
+    }
+}
+
 struct tdx_get_quote_task {
     uint32_t apic_id;
     hwaddr gpa;
@@ -1384,6 +1418,9 @@ static void tdx_handle_vmcall(X86CPU *cpu, struct kvm_tdx_vmcall *vmcall)
     }
 
     switch (vmcall->subfunction) {
+    case TDG_VP_VMCALL_MAP_GPA:
+        tdx_handle_map_gpa(cpu, vmcall);
+        break;
     case TDG_VP_VMCALL_GET_QUOTE:
         tdx_handle_get_quote(cpu, vmcall);
         break;
Signature · 1 line
2.34.1
author
#101 lore raw

PATCH v2 45/58 i386/tdx: Limit the range size for MapGPA

In reply to Xiaoyao Li

From: Isaku Yamahata <[email protected]>

If the range for TDG.VP.VMCALL<MapGPA> is too large, process the limited size and return retry error. It's bad for VMM to take too long time, e.g. second order, with blocking vcpu execution. It results in too many missing timer interrupts.

---
 target/i386/kvm/tdx.c | 19 ++++++++++++++++++-
 1 file changed, 18 insertions(+), 1 deletion(-)
target/i386/kvm/tdx.c +18 -1
index 0c43c1f7759f..ced55be506d1 100644
@@ -994,12 +994,16 @@ static hwaddr tdx_shared_bit(X86CPU *cpu)
     return (cpu->phys_bits > 48) ? BIT_ULL(51) : BIT_ULL(47);
 }
 
+/* 64MB at most in one call. What value is appropriate? */
+#define TDX_MAP_GPA_MAX_LEN     (64 * 1024 * 1024)
+
 static void tdx_handle_map_gpa(X86CPU *cpu, struct kvm_tdx_vmcall *vmcall)
 {
     hwaddr shared_bit = tdx_shared_bit(cpu);
     hwaddr gpa = vmcall->in_r12 & ~shared_bit;
     bool private = !(vmcall->in_r12 & shared_bit);
     hwaddr size = vmcall->in_r13;
+    bool retry = false;
     int ret = 0;
 
     vmcall->status_code = TDG_VP_VMCALL_INVALID_OPERAND;
@@ -1018,12 +1022,25 @@ static void tdx_handle_map_gpa(X86CPU *cpu, struct kvm_tdx_vmcall *vmcall)
         return;
     }
 
+    if (size > TDX_MAP_GPA_MAX_LEN) {
+        retry = true;
+        size = TDX_MAP_GPA_MAX_LEN;
+    }
+
     if (size > 0) {
         ret = kvm_convert_memory(gpa, size, private);
     }
 
     if (!ret) {
-        vmcall->status_code = TDG_VP_VMCALL_SUCCESS;
+        if (retry) {
+            vmcall->status_code = TDG_VP_VMCALL_RETRY;
+            vmcall->out_r11 = gpa + size;
+            if (!private) {
+                vmcall->out_r11 |= shared_bit;
+            }
+        } else {
+            vmcall->status_code = TDG_VP_VMCALL_SUCCESS;
+        }
     }
 }
 
Signature · 1 line
2.34.1
#102 lore raw

Re: [PATCH v2 45/58] i386/tdx: Limit the range size for MapGPA

Quoted 6 lines · Xiaoyao Li

From: Isaku Yamahata <[email protected]>

If the range for TDG.VP.VMCALL<MapGPA> is too large, process the limited size and return retry error. It's bad for VMM to take too long time, e.g. second order, with blocking vcpu execution. It results in too many missing timer interrupts.

This patch requires the guest side patch. [1]
Unless with large guest memory, it's unlikely to hit the limit with KVM/qemu, though.

[1] https://lore.kernel.org/all/[email protected]/

Quoted 59 lines
---
 target/i386/kvm/tdx.c | 19 ++++++++++++++++++-
 1 file changed, 18 insertions(+), 1 deletion(-)
target/i386/kvm/tdx.c +18 -1
index 0c43c1f7759f..ced55be506d1 100644
@@ -994,12 +994,16 @@ static hwaddr tdx_shared_bit(X86CPU *cpu)
     return (cpu->phys_bits > 48) ? BIT_ULL(51) : BIT_ULL(47);
 }
 
+/* 64MB at most in one call. What value is appropriate? */
+#define TDX_MAP_GPA_MAX_LEN     (64 * 1024 * 1024)
+
 static void tdx_handle_map_gpa(X86CPU *cpu, struct kvm_tdx_vmcall *vmcall)
 {
     hwaddr shared_bit = tdx_shared_bit(cpu);
     hwaddr gpa = vmcall->in_r12 & ~shared_bit;
     bool private = !(vmcall->in_r12 & shared_bit);
     hwaddr size = vmcall->in_r13;
+    bool retry = false;
     int ret = 0;
 
     vmcall->status_code = TDG_VP_VMCALL_INVALID_OPERAND;
@@ -1018,12 +1022,25 @@ static void tdx_handle_map_gpa(X86CPU *cpu, struct kvm_tdx_vmcall *vmcall)
         return;
     }
 
+    if (size > TDX_MAP_GPA_MAX_LEN) {
+        retry = true;
+        size = TDX_MAP_GPA_MAX_LEN;
+    }
+
     if (size > 0) {
         ret = kvm_convert_memory(gpa, size, private);
     }
 
     if (!ret) {
-        vmcall->status_code = TDG_VP_VMCALL_SUCCESS;
+        if (retry) {
+            vmcall->status_code = TDG_VP_VMCALL_RETRY;
+            vmcall->out_r11 = gpa + size;
+            if (!private) {
+                vmcall->out_r11 |= shared_bit;
+            }
+        } else {
+            vmcall->status_code = TDG_VP_VMCALL_SUCCESS;
+        }
     }
 }
 
Signature · 1 line
2.34.1
Signature · 1 line
Isaku Yamahata <[email protected]>
author
#103 lore raw

PATCH v2 46/58 i386/tdx: Handle TDG.VP.VMCALL<REPORT_FATAL_ERROR>

In reply to Xiaoyao Li
---
 target/i386/kvm/tdx.c | 40 ++++++++++++++++++++++++++++++++++++++++
 1 file changed, 40 insertions(+)
target/i386/kvm/tdx.c +40 -0
index ced55be506d1..f111b46dac92 100644
@@ -936,6 +936,7 @@ static void tdx_guest_class_init(ObjectClass *oc, void *data)
 
 #define TDG_VP_VMCALL_MAP_GPA                           0x10001ULL
 #define TDG_VP_VMCALL_GET_QUOTE                         0x10002ULL
+#define TDG_VP_VMCALL_REPORT_FATAL_ERROR                0x10003ULL
 #define TDG_VP_VMCALL_SETUP_EVENT_NOTIFY_INTERRUPT      0x10004ULL
 
 #define TDG_VP_VMCALL_SUCCESS           0x0000000000000000ULL
@@ -1407,6 +1408,42 @@ static void tdx_handle_get_quote(X86CPU *cpu, struct kvm_tdx_vmcall *vmcall)
     vmcall->status_code = TDG_VP_VMCALL_SUCCESS;
 }
 
+static void tdx_handle_report_fatal_error(X86CPU *cpu,
+                                          struct kvm_tdx_vmcall *vmcall)
+{
+    uint64_t error_code = vmcall->in_r12;
+    char *message = NULL;
+
+    if (error_code & 0xffff) {
+        error_report("invalid error code of TDG.VP.VMCALL<REPORT_FATAL_ERROR>\n");
+        exit(1);
+    }
+
+    /* it has optional message */
+    if (vmcall->in_r14) {
+        uint64_t * tmp;
+
+#define GUEST_PANIC_INFO_TDX_MESSAGE_MAX        64
+        message = g_malloc0(GUEST_PANIC_INFO_TDX_MESSAGE_MAX + 1);
+
+        tmp = (uint64_t *)message;
+        /* The order is defined in TDX GHCI spec */
+        *(tmp++) = cpu_to_le64(vmcall->in_r14);
+        *(tmp++) = cpu_to_le64(vmcall->in_r15);
+        *(tmp++) = cpu_to_le64(vmcall->in_rbx);
+        *(tmp++) = cpu_to_le64(vmcall->in_rdi);
+        *(tmp++) = cpu_to_le64(vmcall->in_rsi);
+        *(tmp++) = cpu_to_le64(vmcall->in_r8);
+        *(tmp++) = cpu_to_le64(vmcall->in_r9);
+        *(tmp++) = cpu_to_le64(vmcall->in_rdx);
+        message[GUEST_PANIC_INFO_TDX_MESSAGE_MAX] = '\0';
+        assert((char *)tmp == message + GUEST_PANIC_INFO_TDX_MESSAGE_MAX);
+    }
+
+    error_report("TD guest reports fatal error. %s\n", message ? : "");
+    exit(1);
+}
+
 static void tdx_handle_setup_event_notify_interrupt(X86CPU *cpu,
                                                     struct kvm_tdx_vmcall *vmcall)
 {
@@ -1441,6 +1478,9 @@ static void tdx_handle_vmcall(X86CPU *cpu, struct kvm_tdx_vmcall *vmcall)
     case TDG_VP_VMCALL_GET_QUOTE:
         tdx_handle_get_quote(cpu, vmcall);
         break;
+    case TDG_VP_VMCALL_REPORT_FATAL_ERROR:
+        tdx_handle_report_fatal_error(cpu, vmcall);
+        break;
     case TDG_VP_VMCALL_SETUP_EVENT_NOTIFY_INTERRUPT:
         tdx_handle_setup_event_notify_interrupt(cpu, vmcall);
         break;
Signature · 1 line
2.34.1
author
#104 lore raw

PATCH v2 47/58 i386/tdx: Wire REPORT_FATAL_ERROR with GuestPanic facility

In reply to Xiaoyao Li

Originated-from: Isaku Yamahata <[email protected]>

---
 qapi/run-state.json   | 17 +++++++++++++--
 softmmu/runstate.c    | 49 +++++++++++++++++++++++++++++++++++++++++++
 target/i386/kvm/tdx.c | 24 ++++++++++++++++++++-
 3 files changed, 87 insertions(+), 3 deletions(-)
qapi/run-state.json +15 -2
index f216ba54ec4c..506bbe31541f 100644
@@ -499,7 +499,7 @@
 # Since: 2.9
 ##
 { 'enum': 'GuestPanicInformationType',
-  'data': [ 'hyper-v', 's390' ] }
+  'data': [ 'hyper-v', 's390', 'tdx' ] }
 
 ##
 # @GuestPanicInformation:
@@ -514,7 +514,8 @@
  'base': {'type': 'GuestPanicInformationType'},
  'discriminator': 'type',
  'data': {'hyper-v': 'GuestPanicInformationHyperV',
-          's390': 'GuestPanicInformationS390'}}
+          's390': 'GuestPanicInformationS390',
+          'tdx' : 'GuestPanicInformationTdx'}}
 
 ##
 # @GuestPanicInformationHyperV:
@@ -577,6 +578,18 @@
           'psw-addr': 'uint64',
           'reason': 'S390CrashReason'}}
 
+##
+# @GuestPanicInformationTdx:
+#
+# TDX GHCI TDG.VP.VMCALL<ReportFatalError> specific guest panic information
+#
+# Since: 8.2
+##
+{'struct': 'GuestPanicInformationTdx',
+ 'data': {'error-code': 'uint64',
+          'gpa': 'uint64',
+          'message': 'str'}}
+
 ##
 # @MEMORY_FAILURE:
 #
softmmu/runstate.c +49 -0

56 lines not on this page · raw message

target/i386/kvm/tdx.c +23 -1
index f111b46dac92..7efaa13f59e2 100644
@@ -18,6 +18,7 @@
 #include "qom/object_interfaces.h"
 #include "standard-headers/asm-x86/kvm_para.h"
 #include "sysemu/kvm.h"
+#include "sysemu/runstate.h"
 #include "sysemu/sysemu.h"
 #include "exec/address-spaces.h"
 #include "exec/ramblock.h"
@@ -1408,11 +1409,26 @@ static void tdx_handle_get_quote(X86CPU *cpu, struct kvm_tdx_vmcall *vmcall)
     vmcall->status_code = TDG_VP_VMCALL_SUCCESS;
 }
 
+static void tdx_panicked_on_fatal_error(X86CPU *cpu, uint64_t error_code,
+                                        uint64_t gpa, char *message)
+{
+    GuestPanicInformation *panic_info;
+
+    panic_info = g_new0(GuestPanicInformation, 1);
+    panic_info->type = GUEST_PANIC_INFORMATION_TYPE_TDX;
+    panic_info->u.tdx.error_code = error_code;
+    panic_info->u.tdx.gpa = gpa;
+    panic_info->u.tdx.message = (char *)message;
+
+    qemu_system_guest_panicked(panic_info);
+}
+
 static void tdx_handle_report_fatal_error(X86CPU *cpu,
                                           struct kvm_tdx_vmcall *vmcall)
 {
     uint64_t error_code = vmcall->in_r12;
     char *message = NULL;
+    uint64_t gpa = -1ull;
 
     if (error_code & 0xffff) {
         error_report("invalid error code of TDG.VP.VMCALL<REPORT_FATAL_ERROR>\n");
@@ -1441,7 +1457,13 @@ static void tdx_handle_report_fatal_error(X86CPU *cpu,
     }
 
     error_report("TD guest reports fatal error. %s\n", message ? : "");
-    exit(1);
+
+#define TDX_REPORT_FATAL_ERROR_GPA_VALID    BIT_ULL(63)
+    if (error_code & TDX_REPORT_FATAL_ERROR_GPA_VALID) {
+	gpa = vmcall->in_r13;
+    }
+
+    tdx_panicked_on_fatal_error(cpu, error_code, gpa, message);
 }
 
 static void tdx_handle_setup_event_notify_interrupt(X86CPU *cpu,
Signature · 1 line
2.34.1
#105 lore raw

Re: [PATCH v2 47/58] i386/tdx: Wire REPORT_FATAL_ERROR with GuestPanic facility

Quoted 13 lines · Xiaoyao Li

Originated-from: Isaku Yamahata <[email protected]>

---
 qapi/run-state.json   | 17 +++++++++++++--
 softmmu/runstate.c    | 49 +++++++++++++++++++++++++++++++++++++++++++
 target/i386/kvm/tdx.c | 24 ++++++++++++++++++++-
 3 files changed, 87 insertions(+), 3 deletions(-)
qapi/run-state.json +0 -0
index f216ba54ec4c..506bbe31541f 100644

qapi/run-state.json line 502

Quoted 5 lines
qapi/run-state.json +1 -1
@@ -499,7 +499,7 @@
 # Since: 2.9
 ##
 { 'enum': 'GuestPanicInformationType',
-  'data': [ 'hyper-v', 's390' ] }
+  'data': [ 'hyper-v', 's390', 'tdx' ] }

Missing documentation for the 'tdx' value

Quoted 14 lines
qapi/run-state.json +2 -1
 ##
 # @GuestPanicInformation:
@@ -514,7 +514,8 @@
  'base': {'type': 'GuestPanicInformationType'},
  'discriminator': 'type',
  'data': {'hyper-v': 'GuestPanicInformationHyperV',
-          's390': 'GuestPanicInformationS390'}}
+          's390': 'GuestPanicInformationS390',
+          'tdx' : 'GuestPanicInformationTdx'}}
 
 ##
 # @GuestPanicInformationHyperV:

qapi/run-state.json line 584

Quoted 7 lines
qapi/run-state.json +4 -0
@@ -577,6 +578,18 @@
           'psw-addr': 'uint64',
           'reason': 'S390CrashReason'}}
 
+##
+# @GuestPanicInformationTdx:
+#
+# TDX GHCI TDG.VP.VMCALL<ReportFatalError> specific guest panic information

Not documented any of the struct members. Especially please include the warning that 'message' comes from the guest and so must not be trusted, not assumed to be well formed.

Quoted 46 lines
qapi/run-state.json +8 -0
+#
+# Since: 8.2
+##
+{'struct': 'GuestPanicInformationTdx',
+ 'data': {'error-code': 'uint64',
+          'gpa': 'uint64',
+          'message': 'str'}}
+
 ##
 # @MEMORY_FAILURE:
 #
softmmu/runstate.c +27 -0
index f3bd86281813..cab11484ed7e 100644
@@ -518,7 +518,56 @@ void qemu_system_guest_panicked(GuestPanicInformation *info)
                           S390CrashReason_str(info->u.s390.reason),
                           info->u.s390.psw_mask,
                           info->u.s390.psw_addr);
+        } else if (info->type == GUEST_PANIC_INFORMATION_TYPE_TDX) {
+            char *buf = NULL;
+            bool printable = false;
+
+            /*
+             * Although message is defined as a json string, we shouldn't
+             * unconditionally treat it as is because the guest generated it and
+             * it's not necessarily trustable.
+             */
+            if (info->u.tdx.message) {
+                /* The caller guarantees the NUL-terminated string. */
+                int len = strlen(info->u.tdx.message);
+                int i;
+
+                printable = len > 0;
+                for (i = 0; i < len; i++) {
+                    if (!(0x20 <= info->u.tdx.message[i] &&
+                          info->u.tdx.message[i] <= 0x7e)) {
+                        printable = false;
+                        break;
+                    }
+                }
+
+                /* 3 = length of "%02x " */
+                buf = g_malloc(len * 3);
+                for (i = 0; i < len; i++) {
+                    if (info->u.tdx.message[i] == '\0') {

softmmu/runstate.c void qemu_system_guest_panicked(GuestPanicInformation *info) · line 557

Quoted 10 lines
softmmu/runstate.c +10 -0
+                        break;
+                    } else {
+                        sprintf(buf + 3 * i, "%02x ", info->u.tdx.message[i]);
+                    }
+                }
+                if (i > 0)
+                    /* replace the last ' '(space) to NUL */
+                    buf[i * 3 - 1] = '\0';
+                else
+                    buf[0] = '\0';

You're building this escaped buffer but...

softmmu/runstate.c void qemu_system_guest_panicked(GuestPanicInformation *info) · line 565

Quoted 8 lines
softmmu/runstate.c +8 -0
+            }
+
+            qemu_log_mask(LOG_GUEST_ERROR,
+                          //" TDX report fatal error:\"%s\" %s",
+                          " TDX report fatal error:\"%s\""
+                          "error: 0x%016" PRIx64 " gpa page: 0x%016" PRIx64 "\n",
+                          printable ? info->u.tdx.message : "",
+                          //buf ? buf : "",

...then not actually using it

Either delete the 'buf' code, or use it.

Quoted 48 lines
softmmu/runstate.c +4 -0
+                          info->u.tdx.error_code,
+                          info->u.tdx.gpa);
+            g_free(buf);
         }
+
         qapi_free_GuestPanicInformation(info);
     }
 }
target/i386/kvm/tdx.c +16 -0
index f111b46dac92..7efaa13f59e2 100644
@@ -18,6 +18,7 @@
 #include "qom/object_interfaces.h"
 #include "standard-headers/asm-x86/kvm_para.h"
 #include "sysemu/kvm.h"
+#include "sysemu/runstate.h"
 #include "sysemu/sysemu.h"
 #include "exec/address-spaces.h"
 #include "exec/ramblock.h"
@@ -1408,11 +1409,26 @@ static void tdx_handle_get_quote(X86CPU *cpu, struct kvm_tdx_vmcall *vmcall)
     vmcall->status_code = TDG_VP_VMCALL_SUCCESS;
 }
 
+static void tdx_panicked_on_fatal_error(X86CPU *cpu, uint64_t error_code,
+                                        uint64_t gpa, char *message)
+{
+    GuestPanicInformation *panic_info;
+
+    panic_info = g_new0(GuestPanicInformation, 1);
+    panic_info->type = GUEST_PANIC_INFORMATION_TYPE_TDX;
+    panic_info->u.tdx.error_code = error_code;
+    panic_info->u.tdx.gpa = gpa;
+    panic_info->u.tdx.message = (char *)message;
+
+    qemu_system_guest_panicked(panic_info);
+}
+
 static void tdx_handle_report_fatal_error(X86CPU *cpu,
                                           struct kvm_tdx_vmcall *vmcall)
 {
     uint64_t error_code = vmcall->in_r12;
     char *message = NULL;
+    uint64_t gpa = -1ull;
 
     if (error_code & 0xffff) {
         error_report("invalid error code of TDG.VP.VMCALL<REPORT_FATAL_ERROR>\n");

target/i386/kvm/tdx.c static void tdx_handle_report_fatal_error(X86CPU *cpu, · line 1459

Quoted 3 lines
target/i386/kvm/tdx.c +0 -0
@@ -1441,7 +1457,13 @@ static void tdx_handle_report_fatal_error(X86CPU *cpu,
     }
 
     error_report("TD guest reports fatal error. %s\n", message ? : "");

In tdx_panicked_on_fatal_error you're avoiding printing 'message' if it contains non-printable characters, but here you're printing it regardless.

Do we still need this error_report call at all ?

target/i386/kvm/tdx.c static void tdx_handle_report_fatal_error(X86CPU *cpu, · line 1463

Quoted 5 lines
target/i386/kvm/tdx.c +4 -1
-    exit(1);
+
+#define TDX_REPORT_FATAL_ERROR_GPA_VALID    BIT_ULL(63)
+    if (error_code & TDX_REPORT_FATAL_ERROR_GPA_VALID) {
+	gpa = vmcall->in_r13;

Bad indent

Quoted 9 lines
target/i386/kvm/tdx.c +3 -0
+    }
+
+    tdx_panicked_on_fatal_error(cpu, error_code, gpa, message);
 }
 
 static void tdx_handle_setup_event_notify_interrupt(X86CPU *cpu,
Signature · 1 line
2.34.1

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
author
#106 lore raw

Re: [PATCH v2 47/58] i386/tdx: Wire REPORT_FATAL_ERROR with GuestPanic facility

Quoted 39 lines · Daniel P. Berrangé
Quoted 13 lines · Xiaoyao Li

Originated-from: Isaku Yamahata <[email protected]>

---
  qapi/run-state.json   | 17 +++++++++++++--
  softmmu/runstate.c    | 49 +++++++++++++++++++++++++++++++++++++++++++
  target/i386/kvm/tdx.c | 24 ++++++++++++++++++++-
  3 files changed, 87 insertions(+), 3 deletions(-)
qapi/run-state.json +0 -0
index f216ba54ec4c..506bbe31541f 100644

qapi/run-state.json line 502

Quoted 5 lines
qapi/run-state.json +1 -1
@@ -499,7 +499,7 @@
  # Since: 2.9
  ##
  { 'enum': 'GuestPanicInformationType',
-  'data': [ 'hyper-v', 's390' ] }
+  'data': [ 'hyper-v', 's390', 'tdx' ] }

Missing documentation for the 'tdx' value

Quoted 17 lines
qapi/run-state.json +2 -1
  ##
  # @GuestPanicInformation:
@@ -514,7 +514,8 @@
   'base': {'type': 'GuestPanicInformationType'},
   'discriminator': 'type',
   'data': {'hyper-v': 'GuestPanicInformationHyperV',
-          's390': 'GuestPanicInformationS390'}}
+          's390': 'GuestPanicInformationS390',
+          'tdx' : 'GuestPanicInformationTdx'}}
 
  ##
  # @GuestPanicInformationHyperV:
@@ -577,6 +578,18 @@
            'psw-addr': 'uint64',
            'reason': 'S390CrashReason'}}
 
Quoted 8 lines

qapi/run-state.json line 584

Quoted 4 lines
qapi/run-state.json +4 -0
+##
+# @GuestPanicInformationTdx:
+#
+# TDX GHCI TDG.VP.VMCALL<ReportFatalError> specific guest panic information

Not documented any of the struct members. Especially please include the warning that 'message' comes from the guest and so must not be trusted, not assumed to be well formed.

Will do it in next version.

thanks!

Quoted 68 lines
Quoted 46 lines
qapi/run-state.json +8 -0
+#
+# Since: 8.2
+##
+{'struct': 'GuestPanicInformationTdx',
+ 'data': {'error-code': 'uint64',
+          'gpa': 'uint64',
+          'message': 'str'}}
+
  ##
  # @MEMORY_FAILURE:
  #
softmmu/runstate.c +27 -0
index f3bd86281813..cab11484ed7e 100644
@@ -518,7 +518,56 @@ void qemu_system_guest_panicked(GuestPanicInformation *info)
                            S390CrashReason_str(info->u.s390.reason),
                            info->u.s390.psw_mask,
                            info->u.s390.psw_addr);
+        } else if (info->type == GUEST_PANIC_INFORMATION_TYPE_TDX) {
+            char *buf = NULL;
+            bool printable = false;
+
+            /*
+             * Although message is defined as a json string, we shouldn't
+             * unconditionally treat it as is because the guest generated it and
+             * it's not necessarily trustable.
+             */
+            if (info->u.tdx.message) {
+                /* The caller guarantees the NUL-terminated string. */
+                int len = strlen(info->u.tdx.message);
+                int i;
+
+                printable = len > 0;
+                for (i = 0; i < len; i++) {
+                    if (!(0x20 <= info->u.tdx.message[i] &&
+                          info->u.tdx.message[i] <= 0x7e)) {
+                        printable = false;
+                        break;
+                    }
+                }
+
+                /* 3 = length of "%02x " */
+                buf = g_malloc(len * 3);
+                for (i = 0; i < len; i++) {
+                    if (info->u.tdx.message[i] == '\0') {

softmmu/runstate.c void qemu_system_guest_panicked(GuestPanicInformation *info) · line 557

Quoted 10 lines
softmmu/runstate.c +10 -0
+                        break;
+                    } else {
+                        sprintf(buf + 3 * i, "%02x ", info->u.tdx.message[i]);
+                    }
+                }
+                if (i > 0)
+                    /* replace the last ' '(space) to NUL */
+                    buf[i * 3 - 1] = '\0';
+                else
+                    buf[0] = '\0';

You're building this escaped buffer but...

softmmu/runstate.c void qemu_system_guest_panicked(GuestPanicInformation *info) · line 565

Quoted 8 lines
softmmu/runstate.c +8 -0
+            }
+
+            qemu_log_mask(LOG_GUEST_ERROR,
+                          //" TDX report fatal error:\"%s\" %s",
+                          " TDX report fatal error:\"%s\""
+                          "error: 0x%016" PRIx64 " gpa page: 0x%016" PRIx64 "\n",
+                          printable ? info->u.tdx.message : "",
+                          //buf ? buf : "",
Quoted 3 lines

...then not actually using it

Either delete the 'buf' code, or use it.

Sorry for posting some internal testing version.
Does below look good to you?

@@ -518,7 +518,56 @@ void qemu_system_guest_panicked(GuestPanicInformation *info)

                           S390CrashReason_str(info->u.s390.reason),
                           info->u.s390.psw_mask,
                           info->u.s390.psw_addr);
+        } else if (info->type == GUEST_PANIC_INFORMATION_TYPE_TDX) {
+            bool printable = false;
+            char *buf = NULL;
+            int len = 0, i;
+
+            /*
+             * Although message is defined as a json string, we shouldn't
+             * unconditionally treat it as is because the guest generated it and
+             * it's not necessarily trustable.
+             */

+ if (info->u.tdx.message) {
+ /* The caller guarantees the NUL-terminated string. */
+ len = strlen(info->u.tdx.message);
+
+ printable = len > 0;
+ for (i = 0; i < len; i++) {
+ if (!(0x20 <= info->u.tdx.message[i] &&
+ info->u.tdx.message[i] <= 0x7e)) {
+ printable = false;
+ break;
+ }
+ }
+ }
+
+ if (!printable && len) {

+                /* 3 = length of "%02x " */
+                buf = g_malloc(len * 3);
+                for (i = 0; i < len; i++) {

+ if (info->u.tdx.message[i] == '\0') {
+ break;
+ } else {
+ sprintf(buf + 3 * i, "%02x ", info->u.tdx.message[i]);
+ }
+ }
+ if (i > 0)
+ /* replace the last ' '(space) to NUL */
+ buf[i * 3 - 1] = '\0';
+ else
+ buf[0] = '\0';
+ }
+
+ qemu_log_mask(LOG_GUEST_ERROR,

+                          " TDX guest reports fatal error:\"%s\""
+                          " error code: 0x%016" PRIx64 " gpa page: 0x%016" PRIx64 "\n",
+                          printable ? info->u.tdx.message : buf,
+                          info->u.tdx.error_code,
+                          info->u.tdx.gpa);
+            g_free(buf);
         }
Quoted 50 lines
Quoted 50 lines
softmmu/runstate.c +4 -0
+                          info->u.tdx.error_code,
+                          info->u.tdx.gpa);
+            g_free(buf);
          }
+
          qapi_free_GuestPanicInformation(info);
      }
  }
target/i386/kvm/tdx.c +16 -0
index f111b46dac92..7efaa13f59e2 100644
@@ -18,6 +18,7 @@
  #include "qom/object_interfaces.h"
  #include "standard-headers/asm-x86/kvm_para.h"
  #include "sysemu/kvm.h"
+#include "sysemu/runstate.h"
  #include "sysemu/sysemu.h"
  #include "exec/address-spaces.h"
  #include "exec/ramblock.h"
@@ -1408,11 +1409,26 @@ static void tdx_handle_get_quote(X86CPU *cpu, struct kvm_tdx_vmcall *vmcall)
      vmcall->status_code = TDG_VP_VMCALL_SUCCESS;
  }
 
+static void tdx_panicked_on_fatal_error(X86CPU *cpu, uint64_t error_code,
+                                        uint64_t gpa, char *message)
+{
+    GuestPanicInformation *panic_info;
+
+    panic_info = g_new0(GuestPanicInformation, 1);
+    panic_info->type = GUEST_PANIC_INFORMATION_TYPE_TDX;
+    panic_info->u.tdx.error_code = error_code;
+    panic_info->u.tdx.gpa = gpa;
+    panic_info->u.tdx.message = (char *)message;
+
+    qemu_system_guest_panicked(panic_info);
+}
+
  static void tdx_handle_report_fatal_error(X86CPU *cpu,
                                            struct kvm_tdx_vmcall *vmcall)
  {
      uint64_t error_code = vmcall->in_r12;
      char *message = NULL;
+    uint64_t gpa = -1ull;
 
      if (error_code & 0xffff) {
          error_report("invalid error code of TDG.VP.VMCALL<REPORT_FATAL_ERROR>\n");
@@ -1441,7 +1457,13 @@ static void tdx_handle_report_fatal_error(X86CPU *cpu,
      }
 
Quoted 4 lines

target/i386/kvm/tdx.c static void tdx_handle_report_fatal_error(X86CPU *cpu, · line 1459

Quoted 1 line
target/i386/kvm/tdx.c +0 -0
      error_report("TD guest reports fatal error. %s\n", message ? : "");

In tdx_panicked_on_fatal_error you're avoiding printing 'message' if it contains non-printable characters, but here you're printing it regardless.

I guess you meant qemu_system_guest_panicked().

Quoted 1 line

Do we still need this error_report call at all ?

yes. It can and should be dropped before I sent to maillist. I keep it internally for testing purpose because qemu_log_mask() doesn't get printed by default.

Quoted 7 lines

target/i386/kvm/tdx.c static void tdx_handle_report_fatal_error(X86CPU *cpu, · line 1463

Quoted 5 lines
target/i386/kvm/tdx.c +4 -1
-    exit(1);
+
+#define TDX_REPORT_FATAL_ERROR_GPA_VALID    BIT_ULL(63)
+    if (error_code & TDX_REPORT_FATAL_ERROR_GPA_VALID) {
+	gpa = vmcall->in_r13;

Bad indent

Fixed.

Quoted 12 lines
Quoted 9 lines
target/i386/kvm/tdx.c +3 -0
+    }
+
+    tdx_panicked_on_fatal_error(cpu, error_code, gpa, message);
  }
 
  static void tdx_handle_setup_event_notify_interrupt(X86CPU *cpu,
Signature · 1 line
2.34.1

With regards,
Daniel

#107 lore raw

Re: [PATCH v2 47/58] i386/tdx: Wire REPORT_FATAL_ERROR with GuestPanic facility

Quoted 149 lines · Xiaoyao Li
Quoted 89 lines · Daniel P. Berrangé
Quoted 66 lines · Xiaoyao Li

Originated-from: Isaku Yamahata <[email protected]>

---
  qapi/run-state.json   | 17 +++++++++++++--
  softmmu/runstate.c    | 49 +++++++++++++++++++++++++++++++++++++++++++
  target/i386/kvm/tdx.c | 24 ++++++++++++++++++++-
  3 files changed, 87 insertions(+), 3 deletions(-)
qapi/run-state.json +9 -1
index f216ba54ec4c..506bbe31541f 100644
@@ -499,7 +499,7 @@
  # Since: 2.9
  ##
  { 'enum': 'GuestPanicInformationType',
-  'data': [ 'hyper-v', 's390' ] }
+  'data': [ 'hyper-v', 's390', 'tdx' ] }
 
 
+#
+# Since: 8.2
+##
+{'struct': 'GuestPanicInformationTdx',
+ 'data': {'error-code': 'uint64',
+          'gpa': 'uint64',
+          'message': 'str'}}
+
  ##
  # @MEMORY_FAILURE:
  #
softmmu/runstate.c +27 -0
index f3bd86281813..cab11484ed7e 100644
@@ -518,7 +518,56 @@ void qemu_system_guest_panicked(GuestPanicInformation *info)
                            S390CrashReason_str(info->u.s390.reason),
                            info->u.s390.psw_mask,
                            info->u.s390.psw_addr);
+        } else if (info->type == GUEST_PANIC_INFORMATION_TYPE_TDX) {
+            char *buf = NULL;
+            bool printable = false;
+
+            /*
+             * Although message is defined as a json string, we shouldn't
+             * unconditionally treat it as is because the guest generated it and
+             * it's not necessarily trustable.
+             */
+            if (info->u.tdx.message) {
+                /* The caller guarantees the NUL-terminated string. */
+                int len = strlen(info->u.tdx.message);
+                int i;
+
+                printable = len > 0;
+                for (i = 0; i < len; i++) {
+                    if (!(0x20 <= info->u.tdx.message[i] &&
+                          info->u.tdx.message[i] <= 0x7e)) {
+                        printable = false;
+                        break;
+                    }
+                }
+
+                /* 3 = length of "%02x " */
+                buf = g_malloc(len * 3);
+                for (i = 0; i < len; i++) {
+                    if (info->u.tdx.message[i] == '\0') {

softmmu/runstate.c void qemu_system_guest_panicked(GuestPanicInformation *info) · line 557

Quoted 10 lines
softmmu/runstate.c +10 -0
+                        break;
+                    } else {
+                        sprintf(buf + 3 * i, "%02x ", info->u.tdx.message[i]);
+                    }
+                }
+                if (i > 0)
+                    /* replace the last ' '(space) to NUL */
+                    buf[i * 3 - 1] = '\0';
+                else
+                    buf[0] = '\0';

You're building this escaped buffer but...

softmmu/runstate.c void qemu_system_guest_panicked(GuestPanicInformation *info) · line 565

Quoted 8 lines
softmmu/runstate.c +8 -0
+            }
+
+            qemu_log_mask(LOG_GUEST_ERROR,
+                          //" TDX report fatal error:\"%s\" %s",
+                          " TDX report fatal error:\"%s\""
+                          "error: 0x%016" PRIx64 " gpa page: 0x%016" PRIx64 "\n",
+                          printable ? info->u.tdx.message : "",
+                          //buf ? buf : "",
Quoted 3 lines

...then not actually using it

Either delete the 'buf' code, or use it.

Sorry for posting some internal testing version.
Does below look good to you?

softmmu/runstate.c +8 -0
@@ -518,7 +518,56 @@ void qemu_system_guest_panicked(GuestPanicInformation
 *info)
                           S390CrashReason_str(info->u.s390.reason),
                           info->u.s390.psw_mask,
                           info->u.s390.psw_addr);
+        } else if (info->type == GUEST_PANIC_INFORMATION_TYPE_TDX) {
+            bool printable = false;
+            char *buf = NULL;
+            int len = 0, i;
+
+            /*
+             * Although message is defined as a json string, we shouldn't
+             * unconditionally treat it as is because the guest generated

it and

softmmu/runstate.c +24 -0
+             * it's not necessarily trustable.
+             */
+            if (info->u.tdx.message) {
+                /* The caller guarantees the NUL-terminated string. */
+                len = strlen(info->u.tdx.message);
+
+                printable = len > 0;
+                for (i = 0; i < len; i++) {
+                    if (!(0x20 <= info->u.tdx.message[i] &&
+                          info->u.tdx.message[i] <= 0x7e)) {
+                        printable = false;
+                        break;
+                    }
+                }
+            }
+
+            if (!printable && len) {
+                /* 3 = length of "%02x " */
+                buf = g_malloc(len * 3);
+                for (i = 0; i < len; i++) {
+                    if (info->u.tdx.message[i] == '\0') {
+                        break;
+                    } else {
+                        sprintf(buf + 3 * i, "%02x ",

info->u.tdx.message[i]);

softmmu/runstate.c +12 -0
+                    }
+                }
+                if (i > 0)
+                    /* replace the last ' '(space) to NUL */
+                    buf[i * 3 - 1] = '\0';
+                else
+                    buf[0] = '\0';
+            }
+
+            qemu_log_mask(LOG_GUEST_ERROR,
+                          " TDX guest reports fatal error:\"%s\""
+                          " error code: 0x%016" PRIx64 " gpa page: 0x%016"

PRIx64 "\n",

softmmu/runstate.c void qemu_system_guest_panicked(GuestPanicInformation *info) · line 569

Quoted 5 lines
softmmu/runstate.c +4 -0
+                          printable ? info->u.tdx.message : buf,
+                          info->u.tdx.error_code,
+                          info->u.tdx.gpa);
+            g_free(buf);
         }

Ok that makes more sense now. BTW, probably a nice idea to create a separate helper method that santizes the guest provided JSON into the safe 'buf' string.

With regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
author
#108 lore raw

Re: [PATCH v2 47/58] i386/tdx: Wire REPORT_FATAL_ERROR with GuestPanic facility

Quoted 152 lines · Daniel P. Berrangé
Quoted 151 lines · Xiaoyao Li
Quoted 89 lines · Daniel P. Berrangé
Quoted 66 lines · Xiaoyao Li

Originated-from: Isaku Yamahata <[email protected]>

---
   qapi/run-state.json   | 17 +++++++++++++--
   softmmu/runstate.c    | 49 +++++++++++++++++++++++++++++++++++++++++++
   target/i386/kvm/tdx.c | 24 ++++++++++++++++++++-
   3 files changed, 87 insertions(+), 3 deletions(-)
qapi/run-state.json +9 -1
index f216ba54ec4c..506bbe31541f 100644
@@ -499,7 +499,7 @@
   # Since: 2.9
   ##
   { 'enum': 'GuestPanicInformationType',
-  'data': [ 'hyper-v', 's390' ] }
+  'data': [ 'hyper-v', 's390', 'tdx' ] }
 
 
+#
+# Since: 8.2
+##
+{'struct': 'GuestPanicInformationTdx',
+ 'data': {'error-code': 'uint64',
+          'gpa': 'uint64',
+          'message': 'str'}}
+
   ##
   # @MEMORY_FAILURE:
   #
softmmu/runstate.c +27 -0
index f3bd86281813..cab11484ed7e 100644
@@ -518,7 +518,56 @@ void qemu_system_guest_panicked(GuestPanicInformation *info)
                             S390CrashReason_str(info->u.s390.reason),
                             info->u.s390.psw_mask,
                             info->u.s390.psw_addr);
+        } else if (info->type == GUEST_PANIC_INFORMATION_TYPE_TDX) {
+            char *buf = NULL;
+            bool printable = false;
+
+            /*
+             * Although message is defined as a json string, we shouldn't
+             * unconditionally treat it as is because the guest generated it and
+             * it's not necessarily trustable.
+             */
+            if (info->u.tdx.message) {
+                /* The caller guarantees the NUL-terminated string. */
+                int len = strlen(info->u.tdx.message);
+                int i;
+
+                printable = len > 0;
+                for (i = 0; i < len; i++) {
+                    if (!(0x20 <= info->u.tdx.message[i] &&
+                          info->u.tdx.message[i] <= 0x7e)) {
+                        printable = false;
+                        break;
+                    }
+                }
+
+                /* 3 = length of "%02x " */
+                buf = g_malloc(len * 3);
+                for (i = 0; i < len; i++) {
+                    if (info->u.tdx.message[i] == '\0') {

softmmu/runstate.c void qemu_system_guest_panicked(GuestPanicInformation *info) · line 557

Quoted 10 lines
softmmu/runstate.c +10 -0
+                        break;
+                    } else {
+                        sprintf(buf + 3 * i, "%02x ", info->u.tdx.message[i]);
+                    }
+                }
+                if (i > 0)
+                    /* replace the last ' '(space) to NUL */
+                    buf[i * 3 - 1] = '\0';
+                else
+                    buf[0] = '\0';

You're building this escaped buffer but...

softmmu/runstate.c void qemu_system_guest_panicked(GuestPanicInformation *info) · line 565

Quoted 8 lines
softmmu/runstate.c +8 -0
+            }
+
+            qemu_log_mask(LOG_GUEST_ERROR,
+                          //" TDX report fatal error:\"%s\" %s",
+                          " TDX report fatal error:\"%s\""
+                          "error: 0x%016" PRIx64 " gpa page: 0x%016" PRIx64 "\n",
+                          printable ? info->u.tdx.message : "",
+                          //buf ? buf : "",
Quoted 3 lines

...then not actually using it

Either delete the 'buf' code, or use it.

Sorry for posting some internal testing version.
Does below look good to you?

softmmu/runstate.c +8 -0
@@ -518,7 +518,56 @@ void qemu_system_guest_panicked(GuestPanicInformation
 *info)
                            S390CrashReason_str(info->u.s390.reason),
                            info->u.s390.psw_mask,
                            info->u.s390.psw_addr);
+        } else if (info->type == GUEST_PANIC_INFORMATION_TYPE_TDX) {
+            bool printable = false;
+            char *buf = NULL;
+            int len = 0, i;
+
+            /*
+             * Although message is defined as a json string, we shouldn't
+             * unconditionally treat it as is because the guest generated

it and

softmmu/runstate.c +24 -0
+             * it's not necessarily trustable.
+             */
+            if (info->u.tdx.message) {
+                /* The caller guarantees the NUL-terminated string. */
+                len = strlen(info->u.tdx.message);
+
+                printable = len > 0;
+                for (i = 0; i < len; i++) {
+                    if (!(0x20 <= info->u.tdx.message[i] &&
+                          info->u.tdx.message[i] <= 0x7e)) {
+                        printable = false;
+                        break;
+                    }
+                }
+            }
+
+            if (!printable && len) {
+                /* 3 = length of "%02x " */
+                buf = g_malloc(len * 3);
+                for (i = 0; i < len; i++) {
+                    if (info->u.tdx.message[i] == '\0') {
+                        break;
+                    } else {
+                        sprintf(buf + 3 * i, "%02x ",

info->u.tdx.message[i]);

softmmu/runstate.c +12 -0
+                    }
+                }
+                if (i > 0)
+                    /* replace the last ' '(space) to NUL */
+                    buf[i * 3 - 1] = '\0';
+                else
+                    buf[0] = '\0';
+            }
+
+            qemu_log_mask(LOG_GUEST_ERROR,
+                          " TDX guest reports fatal error:\"%s\""
+                          " error code: 0x%016" PRIx64 " gpa page: 0x%016"

PRIx64 "\n",

(quoted diff) +2 -0
+                          printable ? info->u.tdx.message : buf,
+                          info->u.tdx.error_code,
Quoted 9 lines

softmmu/runstate.c void qemu_system_guest_panicked(GuestPanicInformation *info) · line 569

Quoted 3 lines
softmmu/runstate.c +2 -0
+                          info->u.tdx.gpa);
+            g_free(buf);
          }

Ok that makes more sense now. BTW, probably a nice idea to create a separate helper method that santizes the guest provided JSON into the safe 'buf' string.

OK. Thanks for the suggestion.

Will do it in next version.

Quoted 2 lines

With regards,
Daniel

author
#109 lore raw

PATCH v2 48/58 i386/tdx: Disable SMM for TDX VMs

In reply to Xiaoyao Li

TDX doesn't support SMM and VMM cannot emulate SMM for TDX VMs because VMM cannot manipulate TDX VM's memory.

Disable SMM for TDX VMs and error out if user requests to enable SMM.

---
 target/i386/kvm/tdx.c | 8 ++++++++
 1 file changed, 8 insertions(+)
target/i386/kvm/tdx.c +8 -0
index 7efaa13f59e2..f9d03ab0f461 100644
@@ -678,9 +678,17 @@ static Notifier tdx_machine_done_notify = {
 
 int tdx_kvm_init(MachineState *ms, Error **errp)
 {
+    X86MachineState *x86ms = X86_MACHINE(ms);
     TdxGuest *tdx = (TdxGuest *)object_dynamic_cast(OBJECT(ms->cgs),
                                                     TYPE_TDX_GUEST);
 
+    if (x86ms->smm == ON_OFF_AUTO_AUTO) {
+        x86ms->smm = ON_OFF_AUTO_OFF;
+    } else if (x86ms->smm == ON_OFF_AUTO_ON) {
+        error_setg(errp, "TDX VM doesn't support SMM");
+        return -EINVAL;
+    }
+
     if (!tdx_caps) {
         get_tdx_capabilities();
     }
Signature · 1 line
2.34.1
author
#110 lore raw

PATCH v2 49/58 i386/tdx: Disable PIC for TDX VMs

In reply to Xiaoyao Li

Legacy PIC (8259) cannot be supported for TDX VMs since TDX module doesn't allow directly interrupt injection. Using posted interrupts for the PIC is not a viable option as the guest BIOS/kernel will not do EOI for PIC IRQs, i.e. will leave the vIRR bit set.

Hence disable PIC for TDX VMs and error out if user wants PIC.

---
 target/i386/kvm/tdx.c | 7 +++++++
 1 file changed, 7 insertions(+)
target/i386/kvm/tdx.c +7 -0
index f9d03ab0f461..23ecd84a9e21 100644
@@ -689,6 +689,13 @@ int tdx_kvm_init(MachineState *ms, Error **errp)
         return -EINVAL;
     }
 
+    if (x86ms->pic == ON_OFF_AUTO_AUTO) {
+        x86ms->pic = ON_OFF_AUTO_OFF;
+    } else if (x86ms->pic == ON_OFF_AUTO_ON) {
+        error_setg(errp, "TDX VM doesn't support PIC");
+        return -EINVAL;
+    }
+
     if (!tdx_caps) {
         get_tdx_capabilities();
     }
Signature · 1 line
2.34.1
author
#111 lore raw

PATCH v2 50/58 i386/tdx: Don't allow system reset for TDX VMs

In reply to Xiaoyao Li

TDX CPU state is protected and thus vcpu state cann't be reset by VMM.

---
 target/i386/kvm/kvm.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
target/i386/kvm/kvm.c +1 -1
index 601683d836c8..50b0218a8044 100644
@@ -5918,7 +5918,7 @@ bool kvm_has_waitpkg(void)
 
 bool kvm_arch_cpu_check_are_resettable(void)
 {
-    return !sev_es_enabled();
+    return !sev_es_enabled() && !is_tdx_vm();
 }
 
 #define ARCH_REQ_XCOMP_GUEST_PERM       0x1025
Signature · 1 line
2.34.1
author
#112 lore raw

PATCH v2 51/58 i386/tdx: LMCE is not supported for TDX

In reply to Xiaoyao Li

LMCE is not supported TDX since KVM doesn't provide emulation for MSR_IA32_FEAT_CTL.

---
 target/i386/kvm/kvm-cpu.c | 5 +++++
 1 file changed, 5 insertions(+)
target/i386/kvm/kvm-cpu.c +5 -0
index 7237378a7d4e..bec8b5f918e7 100644
@@ -15,6 +15,7 @@
 #include "sysemu/sysemu.h"
 #include "hw/boards.h"
 
+#include "tdx.h"
 #include "kvm_i386.h"
 #include "hw/core/accel-cpu.h"
 
@@ -59,6 +60,10 @@ static bool lmce_supported(void)
     if (kvm_ioctl(kvm_state, KVM_X86_GET_MCE_CAP_SUPPORTED, &mce_cap) < 0) {
         return false;
     }
+
+    if (is_tdx_vm())
+        return false;
+
     return !!(mce_cap & MCG_LMCE_P);
 }
 
Signature · 1 line
2.34.1
author
#113 lore raw

PATCH v2 52/58 hw/i386: add eoi_intercept_unsupported member to X86MachineState

In reply to Xiaoyao Li

Add a new bool member, eoi_intercept_unsupported, to X86MachineState with default value false. Set true for TDX VM.

Inability to intercept eoi causes impossibility to emulate level triggered interrupt to be re-injected when level is still kept active. which affects interrupt controller emulation.

---
 hw/i386/x86.c         | 1 +
 include/hw/i386/x86.h | 1 +
 target/i386/kvm/tdx.c | 2 ++
 3 files changed, 4 insertions(+)
hw/i386/x86.c +1 -0
index a0c9f4d646e2..567384484244 100644
@@ -1426,6 +1426,7 @@ static void x86_machine_initfn(Object *obj)
     x86ms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
     x86ms->bus_lock_ratelimit = 0;
     x86ms->above_4g_mem_start = 4 * GiB;
+    x86ms->eoi_intercept_unsupported = false;
 }
 
 static void x86_machine_class_init(ObjectClass *oc, void *data)
include/hw/i386/x86.h +1 -0
index a3d03f78cefe..c4bfb67b03c7 100644
@@ -61,6 +61,7 @@ struct X86MachineState {
 
     /* CPU and apic information: */
     bool apic_xrupt_override;
+    bool eoi_intercept_unsupported;
     unsigned pci_irq_mask;
     unsigned apic_id_limit;
     uint16_t boot_cpus;
target/i386/kvm/tdx.c +2 -0
index 23ecd84a9e21..9c017cf16d0d 100644
@@ -696,6 +696,8 @@ int tdx_kvm_init(MachineState *ms, Error **errp)
         return -EINVAL;
     }
 
+    x86ms->eoi_intercept_unsupported = true;
+
     if (!tdx_caps) {
         get_tdx_capabilities();
     }
Signature · 1 line
2.34.1
author
#114 lore raw

PATCH v2 53/58 hw/i386: add option to forcibly report edge trigger in acpi tables

In reply to Xiaoyao Li

From: Isaku Yamahata <[email protected]>

When level trigger isn't supported on x86 platform, forcibly report edge trigger in acpi tables.

---
 hw/i386/acpi-build.c  | 99 ++++++++++++++++++++++++++++---------------
 hw/i386/acpi-common.c | 50 ++++++++++++++++------
 2 files changed, 104 insertions(+), 45 deletions(-)
hw/i386/acpi-build.c +66 -33
index 9c74fa17ade0..0505514480bd 100644
@@ -976,7 +976,8 @@ static void build_dbg_aml(Aml *table)
     aml_append(table, scope);
 }
 
-static Aml *build_link_dev(const char *name, uint8_t uid, Aml *reg)
+static Aml *build_link_dev(const char *name, uint8_t uid, Aml *reg,
+                           bool level_trigger_unsupported)
 {
     Aml *dev;
     Aml *crs;
@@ -988,7 +989,10 @@ static Aml *build_link_dev(const char *name, uint8_t uid, Aml *reg)
     aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
 
     crs = aml_resource_template();
-    aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
+    aml_append(crs, aml_interrupt(AML_CONSUMER,
+                                  level_trigger_unsupported ?
+                                  AML_EDGE : AML_LEVEL,
+                                  AML_ACTIVE_HIGH,
                                   AML_SHARED, irqs, ARRAY_SIZE(irqs)));
     aml_append(dev, aml_name_decl("_PRS", crs));
 
@@ -1012,7 +1016,8 @@ static Aml *build_link_dev(const char *name, uint8_t uid, Aml *reg)
     return dev;
  }
 
-static Aml *build_gsi_link_dev(const char *name, uint8_t uid, uint8_t gsi)
+static Aml *build_gsi_link_dev(const char *name, uint8_t uid,
+                               uint8_t gsi, bool level_trigger_unsupported)
 {
     Aml *dev;
     Aml *crs;
@@ -1025,7 +1030,10 @@ static Aml *build_gsi_link_dev(const char *name, uint8_t uid, uint8_t gsi)
 
     crs = aml_resource_template();
     irqs = gsi;
-    aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
+    aml_append(crs, aml_interrupt(AML_CONSUMER,
+                                  level_trigger_unsupported ?
+                                  AML_EDGE : AML_LEVEL,
+                                  AML_ACTIVE_HIGH,
                                   AML_SHARED, &irqs, 1));
     aml_append(dev, aml_name_decl("_PRS", crs));
 
@@ -1044,7 +1052,7 @@ static Aml *build_gsi_link_dev(const char *name, uint8_t uid, uint8_t gsi)
 }
 
 /* _CRS method - get current settings */
-static Aml *build_iqcr_method(bool is_piix4)
+static Aml *build_iqcr_method(bool is_piix4, bool level_trigger_unsupported)
 {
     Aml *if_ctx;
     uint32_t irqs;
@@ -1052,7 +1060,9 @@ static Aml *build_iqcr_method(bool is_piix4)
     Aml *crs = aml_resource_template();
 
     irqs = 0;
-    aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
+    aml_append(crs, aml_interrupt(AML_CONSUMER,
+                                  level_trigger_unsupported ?
+                                  AML_EDGE : AML_LEVEL,
                                   AML_ACTIVE_HIGH, AML_SHARED, &irqs, 1));
     aml_append(method, aml_name_decl("PRR0", crs));
 
@@ -1086,7 +1096,7 @@ static Aml *build_irq_status_method(void)
     return method;
 }
 
-static void build_piix4_pci0_int(Aml *table)
+static void build_piix4_pci0_int(Aml *table, bool level_trigger_unsupported)
 {
     Aml *dev;
     Aml *crs;
@@ -1099,12 +1109,16 @@ static void build_piix4_pci0_int(Aml *table)
     aml_append(sb_scope, pci0_scope);
 
     aml_append(sb_scope, build_irq_status_method());
-    aml_append(sb_scope, build_iqcr_method(true));
+    aml_append(sb_scope, build_iqcr_method(true, level_trigger_unsupported));
 
-    aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQ0")));
-    aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQ1")));
-    aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQ2")));
-    aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQ3")));
+    aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQ0"),
+                                        level_trigger_unsupported));
+    aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQ1"),
+                                        level_trigger_unsupported));
+    aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQ2"),
+                                        level_trigger_unsupported));
+    aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQ3"),
+                                        level_trigger_unsupported));
 
     dev = aml_device("LNKS");
     {
@@ -1113,7 +1127,9 @@ static void build_piix4_pci0_int(Aml *table)
 
         crs = aml_resource_template();
         irqs = 9;
-        aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
+        aml_append(crs, aml_interrupt(AML_CONSUMER,
+                                      level_trigger_unsupported ?
+                                      AML_EDGE : AML_LEVEL,
                                       AML_ACTIVE_HIGH, AML_SHARED,
                                       &irqs, 1));
         aml_append(dev, aml_name_decl("_PRS", crs));
@@ -1199,7 +1215,7 @@ static Aml *build_q35_routing_table(const char *str)
     return pkg;
 }
 
-static void build_q35_pci0_int(Aml *table)
+static void build_q35_pci0_int(Aml *table, bool level_trigger_unsupported)
 {
     Aml *method;
     Aml *sb_scope = aml_scope("_SB");
@@ -1238,25 +1254,41 @@ static void build_q35_pci0_int(Aml *table)
     aml_append(sb_scope, pci0_scope);
 
     aml_append(sb_scope, build_irq_status_method());
-    aml_append(sb_scope, build_iqcr_method(false));
+    aml_append(sb_scope, build_iqcr_method(false, level_trigger_unsupported));
 
-    aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQA")));
-    aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQB")));
-    aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQC")));
-    aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQD")));
-    aml_append(sb_scope, build_link_dev("LNKE", 4, aml_name("PRQE")));
-    aml_append(sb_scope, build_link_dev("LNKF", 5, aml_name("PRQF")));
-    aml_append(sb_scope, build_link_dev("LNKG", 6, aml_name("PRQG")));
-    aml_append(sb_scope, build_link_dev("LNKH", 7, aml_name("PRQH")));
+    aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQA"),
+                                        level_trigger_unsupported));
+    aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQB"),
+                                        level_trigger_unsupported));
+    aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQC"),
+                                        level_trigger_unsupported));
+    aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQD"),
+                                        level_trigger_unsupported));
+    aml_append(sb_scope, build_link_dev("LNKE", 4, aml_name("PRQE"),
+                                        level_trigger_unsupported));
+    aml_append(sb_scope, build_link_dev("LNKF", 5, aml_name("PRQF"),
+                                        level_trigger_unsupported));
+    aml_append(sb_scope, build_link_dev("LNKG", 6, aml_name("PRQG"),
+                                        level_trigger_unsupported));
+    aml_append(sb_scope, build_link_dev("LNKH", 7, aml_name("PRQH"),
+                                        level_trigger_unsupported));
 
-    aml_append(sb_scope, build_gsi_link_dev("GSIA", 0x10, 0x10));
-    aml_append(sb_scope, build_gsi_link_dev("GSIB", 0x11, 0x11));
-    aml_append(sb_scope, build_gsi_link_dev("GSIC", 0x12, 0x12));
-    aml_append(sb_scope, build_gsi_link_dev("GSID", 0x13, 0x13));
-    aml_append(sb_scope, build_gsi_link_dev("GSIE", 0x14, 0x14));
-    aml_append(sb_scope, build_gsi_link_dev("GSIF", 0x15, 0x15));
-    aml_append(sb_scope, build_gsi_link_dev("GSIG", 0x16, 0x16));
-    aml_append(sb_scope, build_gsi_link_dev("GSIH", 0x17, 0x17));
+    aml_append(sb_scope, build_gsi_link_dev("GSIA", 0x10, 0x10,
+                                            level_trigger_unsupported));
+    aml_append(sb_scope, build_gsi_link_dev("GSIB", 0x11, 0x11,
+                                            level_trigger_unsupported));
+    aml_append(sb_scope, build_gsi_link_dev("GSIC", 0x12, 0x12,
+                                            level_trigger_unsupported));
+    aml_append(sb_scope, build_gsi_link_dev("GSID", 0x13, 0x13,
+                                            level_trigger_unsupported));
+    aml_append(sb_scope, build_gsi_link_dev("GSIE", 0x14, 0x14,
+                                            level_trigger_unsupported));
+    aml_append(sb_scope, build_gsi_link_dev("GSIF", 0x15, 0x15,
+                                            level_trigger_unsupported));
+    aml_append(sb_scope, build_gsi_link_dev("GSIG", 0x16, 0x16,
+                                            level_trigger_unsupported));
+    aml_append(sb_scope, build_gsi_link_dev("GSIH", 0x17, 0x17,
+                                            level_trigger_unsupported));
 
     aml_append(table, sb_scope);
 }
@@ -1436,6 +1468,7 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
     PCMachineState *pcms = PC_MACHINE(machine);
     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
     X86MachineState *x86ms = X86_MACHINE(machine);
+    bool level_trigger_unsupported = x86ms->eoi_intercept_unsupported;
     AcpiMcfgInfo mcfg;
     bool mcfg_valid = !!acpi_get_mcfg(&mcfg);
     uint32_t nr_mem = machine->ram_slots;
@@ -1469,7 +1502,7 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
         if (pm->pcihp_bridge_en || pm->pcihp_root_en) {
             build_x86_acpi_pci_hotplug(dsdt, pm->pcihp_io_base);
         }
-        build_piix4_pci0_int(dsdt);
+        build_piix4_pci0_int(dsdt, level_trigger_unsupported);
     } else if (q35) {
         sb_scope = aml_scope("_SB");
         dev = aml_device("PCI0");
@@ -1514,7 +1547,7 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
         if (pm->pcihp_bridge_en) {
             build_x86_acpi_pci_hotplug(dsdt, pm->pcihp_io_base);
         }
-        build_q35_pci0_int(dsdt);
+        build_q35_pci0_int(dsdt, level_trigger_unsupported);
     }
 
     if (misc->has_hpet) {
hw/i386/acpi-common.c +38 -12

62 lines not on this page · raw message

Signature · 1 line
2.34.1
author
#115 lore raw

PATCH v2 54/58 i386/tdx: Don't synchronize guest tsc for TDs

In reply to Xiaoyao Li

From: Isaku Yamahata <[email protected]>

TSC of TDs is not accessible and KVM doesn't allow access of MSR_IA32_TSC for TDs. To avoid the assert() in kvm_get_tsc, make kvm_synchronize_all_tsc() noop for TDs,

---
 target/i386/kvm/kvm.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
target/i386/kvm/kvm.c +1 -1
index 50b0218a8044..3f18a8d0f40b 100644
@@ -298,7 +298,7 @@ void kvm_synchronize_all_tsc(void)
 {
     CPUState *cpu;
 
-    if (kvm_enabled()) {
+    if (kvm_enabled() && !is_tdx_vm()) {
         CPU_FOREACH(cpu) {
             run_on_cpu(cpu, do_kvm_synchronize_tsc, RUN_ON_CPU_NULL);
         }
Signature · 1 line
2.34.1
author
#116 lore raw

PATCH v2 55/58 i386/tdx: Only configure MSR_IA32_UCODE_REV in kvm_init_msrs() for TDs

In reply to Xiaoyao Li

For TDs, only MSR_IA32_UCODE_REV in kvm_init_msrs() can be configured by VMM, while the features enumerated/controlled by other MSRs except MSR_IA32_UCODE_REV in kvm_init_msrs() are not under control of VMM.

Only configure MSR_IA32_UCODE_REV for TDs.

---
 target/i386/kvm/kvm.c | 44 ++++++++++++++++++++++---------------------
 1 file changed, 23 insertions(+), 21 deletions(-)
target/i386/kvm/kvm.c +23 -21
index 3f18a8d0f40b..53d8d65f6667 100644
@@ -3426,32 +3426,34 @@ static void kvm_init_msrs(X86CPU *cpu)
     CPUX86State *env = &cpu->env;
 
     kvm_msr_buf_reset(cpu);
-    if (has_msr_arch_capabs) {
-        kvm_msr_entry_add(cpu, MSR_IA32_ARCH_CAPABILITIES,
-                          env->features[FEAT_ARCH_CAPABILITIES]);
-    }
-
-    if (has_msr_core_capabs) {
-        kvm_msr_entry_add(cpu, MSR_IA32_CORE_CAPABILITY,
-                          env->features[FEAT_CORE_CAPABILITY]);
-    }
-
-    if (has_msr_perf_capabs && cpu->enable_pmu) {
-        kvm_msr_entry_add_perf(cpu, env->features);
+
+    if (!is_tdx_vm()) {
+        if (has_msr_arch_capabs) {
+            kvm_msr_entry_add(cpu, MSR_IA32_ARCH_CAPABILITIES,
+                                env->features[FEAT_ARCH_CAPABILITIES]);
+        }
+
+        if (has_msr_core_capabs) {
+            kvm_msr_entry_add(cpu, MSR_IA32_CORE_CAPABILITY,
+                                env->features[FEAT_CORE_CAPABILITY]);
+        }
+
+        if (has_msr_perf_capabs && cpu->enable_pmu) {
+            kvm_msr_entry_add_perf(cpu, env->features);
+        }
+
+        /*
+         * Older kernels do not include VMX MSRs in KVM_GET_MSR_INDEX_LIST, but
+         * all kernels with MSR features should have them.
+         */
+        if (kvm_feature_msrs && cpu_has_vmx(env)) {
+            kvm_msr_entry_add_vmx(cpu, env->features);
+        }
     }
 
     if (has_msr_ucode_rev) {
         kvm_msr_entry_add(cpu, MSR_IA32_UCODE_REV, cpu->ucode_rev);
     }
-
-    /*
-     * Older kernels do not include VMX MSRs in KVM_GET_MSR_INDEX_LIST, but
-     * all kernels with MSR features should have them.
-     */
-    if (kvm_feature_msrs && cpu_has_vmx(env)) {
-        kvm_msr_entry_add_vmx(cpu, env->features);
-    }
-
     assert(kvm_buf_set_msrs(cpu) == 0);
 }
 
Signature · 1 line
2.34.1
author
#117 lore raw

PATCH v2 56/58 i386/tdx: Skip kvm_put_apicbase() for TDs

In reply to Xiaoyao Li

KVM doesn't allow wirting to MSR_IA32_APICBASE for TDs.

---
 target/i386/kvm/kvm.c | 5 +++++
 1 file changed, 5 insertions(+)
target/i386/kvm/kvm.c +5 -0
index 53d8d65f6667..d542351983cd 100644
@@ -3208,6 +3208,11 @@ void kvm_put_apicbase(X86CPU *cpu, uint64_t value)
 {
     int ret;
 
+    /* TODO: Allow accessing guest state for debug TDs. */
+    if (is_tdx_vm()) {
+        return;
+    }
+
     ret = kvm_put_one_msr(cpu, MSR_IA32_APICBASE, value);
     assert(ret == 1);
 }
Signature · 1 line
2.34.1
author
#118 lore raw

PATCH v2 57/58 i386/tdx: Don't get/put guest state for TDX VMs

In reply to Xiaoyao Li

From: Sean Christopherson <[email protected]>

Don't get/put state of TDX VMs since accessing/mutating guest state of production TDs is not supported.

Note, it will be allowed for a debug TD. Corresponding support will be introduced when debug TD support is implemented in the future.

---
 target/i386/kvm/kvm.c | 11 +++++++++++
 1 file changed, 11 insertions(+)
target/i386/kvm/kvm.c +11 -0
index d542351983cd..1422c79aca40 100644
@@ -4852,6 +4852,11 @@ int kvm_arch_put_registers(CPUState *cpu, int level)
 
     assert(cpu_is_stopped(cpu) || qemu_cpu_is_self(cpu));
 
+    /* TODO: Allow accessing guest state for debug TDs. */
+    if (is_tdx_vm()) {
+        return 0;
+    }
+
     /*
      * Put MSR_IA32_FEATURE_CONTROL first, this ensures the VM gets out of VMX
      * root operation upon vCPU reset. kvm_put_msr_feature_control() should also
@@ -4962,6 +4967,12 @@ int kvm_arch_get_registers(CPUState *cs)
     if (ret < 0) {
         goto out;
     }
+
+    /* TODO: Allow accessing guest state for debug TDs. */
+    if (is_tdx_vm()) {
+        return 0;
+    }
+
     ret = kvm_getput_regs(cpu, 0);
     if (ret < 0) {
         goto out;
Signature · 1 line
2.34.1
author
#119 lore raw

PATCH v2 58/58 docs: Add TDX documentation

In reply to Xiaoyao Li

Add docs/system/i386/tdx.rst for TDX support, and add tdx in confidential-guest-support.rst

---
Changes since v1:
- Add prerequisite of private gmem;
- update example command to launch TD;

Changes since RFC v4:
- add the restriction that kernel-irqchip must be split

---
 docs/system/confidential-guest-support.rst |   1 +
 docs/system/i386/tdx.rst                   | 114 +++++++++++++++++++++
 docs/system/target-i386.rst                |   1 +
 3 files changed, 116 insertions(+)
 create mode 100644 docs/system/i386/tdx.rst
docs/system/confidential-guest-support.rst +1 -0
index 0c490dbda2b7..66129fbab64c 100644
@@ -38,6 +38,7 @@ Supported mechanisms
 Currently supported confidential guest mechanisms are:
 
 * AMD Secure Encrypted Virtualization (SEV) (see :doc:`i386/amd-memory-encryption`)
+* Intel Trust Domain Extension (TDX) (see :doc:`i386/tdx`)
 * POWER Protected Execution Facility (PEF) (see :ref:`power-papr-protected-execution-facility-pef`)
 * s390x Protected Virtualization (PV) (see :doc:`s390x/protvirt`)
 
docs/system/i386/tdx.rst new file +114 -0

114 lines not on this page · raw message

docs/system/target-i386.rst +1 -0
index 1b8a1f248abb..4d58cdbc4e06 100644
@@ -29,6 +29,7 @@ Architectural features
    i386/kvm-pv
    i386/sgx
    i386/amd-memory-encryption
+   i386/tdx
 
 OS requirements
 ~~~~~~~~~~~~~~~
Signature · 1 line
2.34.1