Skip to contentSkip to conversation

Qemu SGX virtualization

PATCH In review Yang Zhong 5 participants 61 messages Jul 19, 2021 – Sep 16, 2021

Subsystems (2)
lore
Share & tools
lore

Yang Zhong

author
#0 lore raw

v4 cover Qemu SGX virtualization

Since Sean Christopherson has left Intel and i am responsible for Qemu SGX upstream work. His @intel.com address will be bouncing and his new email( [email protected]) is also in CC lists.

This series is Qemu SGX virtualization implementation rebased on latest Qemu release. The numa support for SGX will be sent in another patchset once this basic SGX patchset are merged.

You can find Qemu repo here:

https://github.com/intel/qemu-sgx.git upstream

If you want to try SGX, you can directly install the linux release(at least 5.13.0-rc1+) since kvm SGX has been merged into linux release.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git

To simplify, you'd better install linux on host and guest, which can support SGX on host and guest kernel. And to me, use below reference command to boot SGX guest:

    #qemu-system-x86_64 \
        ...... \
        -cpu host,+sgx-provisionkey \
        -object memory-backend-epc,id=mem1,size=64M,prealloc=on \
        -object memory-backend-epc,id=mem2,size=28M \
        -M sgx-epc.0.memdev=mem1,sgx-epc.1.memdev=mem2

Overview
========

Intel Software Guard eXtensions (SGX) is a set of instructions and mechanisms for memory accesses in order to provide security accesses for sensitive applications and data. SGX allows an application to use it's pariticular address space as an *enclave*, which is a protected area provides confidentiality and integrity even in the presence of privileged malware. Accesses to the enclave memory area from any software not resident in the enclave are prevented, including those from privileged software.

SGX virtaulization
==================

The KVM SGX creates one new misc device, sgx_vepc, and Qemu will open '/dev/sgx_vepc' device node to mmap() host EPC memory to guest. The Qemu also adds 'sgx-epc' device to expose EPC sections to guest through CPUID and ACPI table. The Qemu SGX also supports multiple virtual EPC sections to guest, we just put them together physically contiguous for the sake of simplicity. The kernel SGX NUMA has been merged into Linux tip tree, we will support this function in the next phase.

Although the current host SGX subsystem can not support SGX2 feature, the KVM/Qemu implementation still expose this feature to guest. Guest SGX2 support doesn't have interaction with host kernel SGX driver, the SGX guest can normally use those new instructions.

As for SGX virtualization detailed infomation, please reference docs/intel-sgx.txt docuement(patch 33).

Changelog:
=========

(Changelog here is for global changes, please see each patch's changelog for changes made to specific patch.)

v3-->v4:
- Rebased the sgx patches into latest Qemu release.
- Moved sgx compound property setter/getter from MachineState to X86MachineState(Paolo).
- Re-defined struct SgxEPC, removed 'id' property and added struct SgxEPCList for

     sgx-epc.0.{memdev}(Paolo).
   - Removed g_malloc0(), and changed the 'SGXEPCState *sgx_epc' to 'SGXEPCState sgx_epc'
     in struct PCMachineState(Paolo).
   - Changed the SGX compound property cmdline from sgx-epc.{memdev}.0 to
     sgx-epc.0.{memdev}(Paolo).

- Removed the signature from the 'git format-patch' command(Jarkko).

v2-->v3:
- Rebased the sgx patches into latest Qemu release.
- Implemented the compound property for SGX, ref patch5, the command from '-sgx-epc'

     to '-M'(Paolo).

- Moved the sgx common code from sgx-epc.c to sgx.c. The sgx-epc.c is only responsible for virtual epc device.
- Removed the previous patch13(linux-headers: Add placeholder for KVM_CAP_SGX_ATTRIBUTE) because [email protected] updated Linux headers to 5.13-rc4 with commit 278f064e452.
- Updated the patch1 because ram_flags were changed by David Hildenbra.
- Added one patch24, which avoid reset operation caused by bios reset.
- Added one patch25, which make prealloc property consistent with Qemu cmdline during VM reset.

v1-->v2:
- Rebased the sgx patches into latest Qemu release.
- Unified the "share" and "protected" arguments with ram_flags in the memory_region_init_ram_from_fd()(Paolo).
- Added the new MemoryBackendEpcProperties and related documents(Eric Blake).
- Changed the KVM_CAP_SGX_ATTRIBUTE from 195 to 196(Kai).
- Changed the version and some grammar issues(Eric Blake).

Sean Christopherson (21):
memory: Add RAM_PROTECTED flag to skip IOMMU mappings
hostmem: Add hostmem-epc as a backend for SGX EPC
i386: Add 'sgx-epc' device to expose EPC sections to guest
vl: Add sgx compound properties to expose SGX EPC sections to guest
i386: Add primary SGX CPUID and MSR defines
i386: Add SGX CPUID leaf FEAT_SGX_12_0_EAX
i386: Add SGX CPUID leaf FEAT_SGX_12_0_EBX
i386: Add SGX CPUID leaf FEAT_SGX_12_1_EAX
i386: Add get/set/migrate support for SGX_LEPUBKEYHASH MSRs
i386: Add feature control MSR dependency when SGX is enabled
i386: Update SGX CPUID info according to hardware/KVM/user input
i386: kvm: Add support for exposing PROVISIONKEY to guest
i386: Propagate SGX CPUID sub-leafs to KVM
Adjust min CPUID level to 0x12 when SGX is enabled hw/i386/fw_cfg: Set SGX bits in feature control fw_cfg accordingly hw/i386/pc: Account for SGX EPC sections when calculating device
memory
i386/pc: Add e820 entry for SGX EPC section(s)
i386: acpi: Add SGX EPC entry to ACPI tables
q35: Add support for SGX EPC
i440fx: Add support for SGX EPC
doc: Add the SGX doc

Yang Zhong (12):
qom: Add memory-backend-epc ObjectOptions support
hostmem-epc: Add the reset interface for EPC backend reset
sgx-epc: Add the reset interface for sgx-epc virt device
sgx-epc: Avoid bios reset during sgx epc initialization
hostmem-epc: Make prealloc consistent with qemu cmdline during reset
qmp: Add query-sgx command
hmp: Add 'info sgx' command
i386: Add sgx_get_info() interface
bitops: Support 32 and 64 bit mask macro
qmp: Add the qmp_query_sgx_capabilities()
Kconfig: Add CONFIG_SGX support
sgx-epc: Add the fill_device_info() callback support

 backends/hostmem-epc.c                   | 118 ++++++++++
 backends/meson.build                     |   1 +
 configs/devices/i386-softmmu/default.mak |   1 +
 docs/intel-sgx.txt                       | 167 ++++++++++++++
 hmp-commands-info.hx                     |  15 ++
 hw/i386/Kconfig                          |   5 +
 hw/i386/acpi-build.c                     |  22 ++
 hw/i386/fw_cfg.c                         |  10 +-
 hw/i386/meson.build                      |   2 +
 hw/i386/pc.c                             |  15 +-
 hw/i386/pc_piix.c                        |   4 +
 hw/i386/pc_q35.c                         |   3 +
 hw/i386/sgx-epc.c                        | 265 +++++++++++++++++++++++
 hw/i386/sgx-stub.c                       |  13 ++
 hw/i386/sgx.c                            | 170 +++++++++++++++
 hw/i386/x86.c                            |  29 +++
 hw/vfio/common.c                         |   1 +
 include/exec/memory.h                    |  15 +-
 include/hw/i386/pc.h                     |  10 +
 include/hw/i386/sgx-epc.h                |  68 ++++++
 include/hw/i386/x86.h                    |   1 +
 include/monitor/hmp.h                    |   1 +
 include/qemu/bitops.h                    |   7 +
 monitor/hmp-cmds.c                       |  32 +++
 monitor/qmp-cmds.c                       |  19 ++
 qapi/machine.json                        |  52 ++++-
 qapi/misc.json                           |  61 ++++++
 qapi/qom.json                            |  19 ++
 qemu-options.hx                          |  10 +-
 softmmu/memory.c                         |   5 +
 softmmu/physmem.c                        |   3 +-
 stubs/meson.build                        |   1 +
 stubs/sgx-stub.c                         |  12 +
 target/i386/cpu.c                        | 168 +++++++++++++-
 target/i386/cpu.h                        |  16 ++
 target/i386/kvm/kvm.c                    |  75 +++++++
 target/i386/kvm/kvm_i386.h               |   2 +
 target/i386/machine.c                    |  20 ++
 tests/qtest/qmp-cmd-test.c               |   2 +
 39 files changed, 1430 insertions(+), 10 deletions(-)
 create mode 100644 backends/hostmem-epc.c
 create mode 100644 docs/intel-sgx.txt
 create mode 100644 hw/i386/sgx-epc.c
 create mode 100644 hw/i386/sgx-stub.c
 create mode 100644 hw/i386/sgx.c
 create mode 100644 include/hw/i386/sgx-epc.h
 create mode 100644 stubs/sgx-stub.c
author
#1 lore raw

PATCH v4 1/33 memory: Add RAM_PROTECTED flag to skip IOMMU mappings

From: Sean Christopherson <[email protected]>

Add a new RAMBlock flag to denote "protected" memory, i.e. memory that looks and acts like RAM but is inaccessible via normal mechanisms, including DMA. Use the flag to skip protected memory regions when mapping RAM for DMA in VFIO.

v1-->v2:
- Unified the "share" and "protected" arguments with ram_flags in the memory_region_init_ram_from_fd()(Paolo).

---
 hw/vfio/common.c      |  1 +
 include/exec/memory.h | 15 ++++++++++++++-
 softmmu/memory.c      |  5 +++++
 softmmu/physmem.c     |  3 ++-
 4 files changed, 22 insertions(+), 2 deletions(-)
hw/vfio/common.c +1 -0
index 8728d4d5c2..1289cfa8be 100644
@@ -562,6 +562,7 @@ static bool vfio_listener_skipped_section(MemoryRegionSection *section)
 {
     return (!memory_region_is_ram(section->mr) &&
             !memory_region_is_iommu(section->mr)) ||
+           memory_region_is_protected(section->mr) ||
            /*
             * Sizing an enabled 64-bit BAR can cause spurious mappings to
             * addresses in the upper part of the 64-bit address space.  These
include/exec/memory.h +14 -1
index c3d417d317..9446874d21 100644
@@ -190,6 +190,9 @@ typedef struct IOMMUTLBEvent {
  */
 #define RAM_NORESERVE (1 << 7)
 
+/* RAM that isn't accessible through normal means. */
+#define RAM_PROTECTED (1 << 8)
+
 static inline void iommu_notifier_init(IOMMUNotifier *n, IOMMUNotify fn,
                                        IOMMUNotifierFlag flags,
                                        hwaddr start, hwaddr end,
@@ -1267,7 +1270,7 @@ void memory_region_init_ram_from_file(MemoryRegion *mr,
  * @name: the name of the region.
  * @size: size of the region.
  * @ram_flags: RamBlock flags. Supported flags: RAM_SHARED, RAM_PMEM,
- *             RAM_NORESERVE.
+ *             RAM_NORESERVE, RAM_PROTECTED.
  * @fd: the fd to mmap.
  * @offset: offset within the file referenced by fd
  * @errp: pointer to Error*, to store an error if it happens.
@@ -1568,6 +1571,16 @@ static inline bool memory_region_is_romd(MemoryRegion *mr)
     return mr->rom_device && mr->romd_mode;
 }
 
+/**
+ * memory_region_is_protected: check whether a memory region is protected
+ *
+ * Returns %true if a memory region is protected RAM and cannot be accessed
+ * via standard mechanisms, e.g. DMA.
+ *
+ * @mr: the memory region being queried
+ */
+bool memory_region_is_protected(MemoryRegion *mr);
+
 /**
  * memory_region_get_iommu: check whether a memory region is an iommu
  *
softmmu/memory.c +5 -0
index bfedaf9c4d..54cd0e9824 100644
@@ -1811,6 +1811,11 @@ bool memory_region_is_ram_device(MemoryRegion *mr)
     return mr->ram_device;
 }
 
+bool memory_region_is_protected(MemoryRegion *mr)
+{
+    return mr->ram && (mr->ram_block->flags & RAM_PROTECTED);
+}
+
 uint8_t memory_region_get_dirty_log_mask(MemoryRegion *mr)
 {
     uint8_t mask = mr->dirty_log_mask;
softmmu/physmem.c +2 -1
index 3c1912a1a0..f3b646658f 100644
@@ -2052,7 +2052,8 @@ RAMBlock *qemu_ram_alloc_from_fd(ram_addr_t size, MemoryRegion *mr,
     int64_t file_size, file_align;
 
     /* Just support these ram flags by now. */
-    assert((ram_flags & ~(RAM_SHARED | RAM_PMEM | RAM_NORESERVE)) == 0);
+    assert((ram_flags & ~(RAM_SHARED | RAM_PMEM | RAM_NORESERVE |
+                          RAM_PROTECTED)) == 0);
 
     if (xen_enabled()) {
         error_setg(errp, "-mem-path not supported with Xen");
author
#2 lore raw

PATCH v4 2/33 hostmem: Add hostmem-epc as a backend for SGX EPC

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

EPC (Enclave Page Cahe) is a specialized type of memory used by Intel SGX (Software Guard Extensions). The SDM desribes EPC as:

The Enclave Page Cache (EPC) is the secure storage used to store enclave pages when they are a part of an executing enclave. For an EPC page, hardware performs additional access control checks to restrict access to the page. After the current page access checks and translations are performed, the hardware checks that the EPC page is accessible to the program currently executing. Generally an EPC page is only accessed by the owner of the executing enclave or an instruction which is setting up an EPC page.

Because of its unique requirements, Linux manages EPC separately from normal memory. Similar to memfd, the device /dev/sgx_vepc can be opened to obtain a file descriptor which can in turn be used to mmap() EPC memory.

v1-->v2:
- Unified the "share" and "protected" arguments with ram_flags in the memory_region_init_ram_from_fd()(Paolo).

---
 backends/hostmem-epc.c | 92 ++++++++++++++++++++++++++++++++++++++++++
 backends/meson.build   |  1 +
 2 files changed, 93 insertions(+)
 create mode 100644 backends/hostmem-epc.c
backends/hostmem-epc.c new file +92 -0
new file mode 100644
index 0000000000..b512a68cb0
@@ -0,0 +1,92 @@
+/*
+ * QEMU host SGX EPC memory backend
+ *
+ * Copyright (C) 2019 Intel Corporation
+ *
+ * Authors:
+ *   Sean Christopherson <[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 <sys/ioctl.h>
+
+#include "qemu/osdep.h"
+#include "qemu-common.h"
+#include "qom/object_interfaces.h"
+#include "qapi/error.h"
+#include "sysemu/hostmem.h"
+
+#define TYPE_MEMORY_BACKEND_EPC "memory-backend-epc"
+
+#define MEMORY_BACKEND_EPC(obj)                                        \
+    OBJECT_CHECK(HostMemoryBackendEpc, (obj), TYPE_MEMORY_BACKEND_EPC)
+
+typedef struct HostMemoryBackendEpc HostMemoryBackendEpc;
+
+struct HostMemoryBackendEpc {
+    HostMemoryBackend parent_obj;
+};
+
+static void
+sgx_epc_backend_memory_alloc(HostMemoryBackend *backend, Error **errp)
+{
+    uint32_t ram_flags;
+    char *name;
+    int fd;
+
+    if (!backend->size) {
+        error_setg(errp, "can't create backend with size 0");
+        return;
+    }
+
+    fd = qemu_open_old("/dev/sgx_vepc", O_RDWR);
+    if (fd < 0) {
+        error_setg_errno(errp, errno,
+                         "failed to open /dev/sgx_vepc to alloc SGX EPC");
+        return;
+    }
+
+    name = object_get_canonical_path(OBJECT(backend));
+    ram_flags = (backend->share ? RAM_SHARED : 0) | RAM_PROTECTED;
+    memory_region_init_ram_from_fd(&backend->mr, OBJECT(backend),
+                                   name, backend->size, ram_flags,
+                                   fd, 0, errp);
+    g_free(name);
+}
+
+static void sgx_epc_backend_instance_init(Object *obj)
+{
+    HostMemoryBackend *m = MEMORY_BACKEND(obj);
+
+    m->share = true;
+    m->merge = false;
+    m->dump = false;
+}
+
+static void sgx_epc_backend_class_init(ObjectClass *oc, void *data)
+{
+    HostMemoryBackendClass *bc = MEMORY_BACKEND_CLASS(oc);
+
+    bc->alloc = sgx_epc_backend_memory_alloc;
+}
+
+static const TypeInfo sgx_epc_backed_info = {
+    .name = TYPE_MEMORY_BACKEND_EPC,
+    .parent = TYPE_MEMORY_BACKEND,
+    .instance_init = sgx_epc_backend_instance_init,
+    .class_init = sgx_epc_backend_class_init,
+    .instance_size = sizeof(HostMemoryBackendEpc),
+};
+
+static void register_types(void)
+{
+    int fd = qemu_open_old("/dev/sgx_vepc", O_RDWR);
+    if (fd >= 0) {
+        close(fd);
+
+        type_register_static(&sgx_epc_backed_info);
+    }
+}
+
+type_init(register_types);
backends/meson.build +1 -0
index d4221831fc..46fd16b269 100644
@@ -16,5 +16,6 @@ softmmu_ss.add(when: ['CONFIG_VHOST_USER', 'CONFIG_VIRTIO'], if_true: files('vho
 softmmu_ss.add(when: 'CONFIG_VIRTIO_CRYPTO', if_true: files('cryptodev-vhost.c'))
 softmmu_ss.add(when: ['CONFIG_VIRTIO_CRYPTO', 'CONFIG_VHOST_CRYPTO'], if_true: files('cryptodev-vhost-user.c'))
 softmmu_ss.add(when: 'CONFIG_GIO', if_true: [files('dbus-vmstate.c'), gio])
+softmmu_ss.add(when: 'CONFIG_LINUX', if_true: files('hostmem-epc.c'))
 
 subdir('tpm')
author
#3 lore raw

PATCH v4 3/33 qom: Add memory-backend-epc ObjectOptions support

In reply to Yang Zhong

Add the new 'memory-backend-epc' user creatable QOM object in the ObjectOptions to support SGX since v6.1, or the sgx backend object cannot bootup.

v1-->v2:
- Added the new MemoryBackendEpcProperties and related documents, and updated the blurb(Eric Blake).

---
 qapi/qom.json | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
qapi/qom.json +19 -0
index 652be317b8..aba77aa97b 100644
@@ -642,6 +642,23 @@
             '*hugetlbsize': 'size',
             '*seal': 'bool' } }
 
+##
+# @MemoryBackendEpcProperties:
+#
+# Properties for memory-backend-epc objects.
+#
+# The @share boolean option is true by default with epc
+#
+# The @merge boolean option is false by default with epc
+#
+# The @dump boolean option is false by default with epc
+#
+# Since: 6.1
+##
+{ 'struct': 'MemoryBackendEpcProperties',
+  'base': 'MemoryBackendProperties',
+  'data': {} }
+
 ##
 # @PrManagerHelperProperties:
 #
@@ -792,6 +809,7 @@
     { 'name': 'memory-backend-memfd',
       'if': 'defined(CONFIG_LINUX)' },
     'memory-backend-ram',
+    'memory-backend-epc',
     'pef-guest',
     'pr-manager-helper',
     'qtest',
@@ -850,6 +868,7 @@
       'memory-backend-memfd':       { 'type': 'MemoryBackendMemfdProperties',
                                       'if': 'defined(CONFIG_LINUX)' },
       'memory-backend-ram':         'MemoryBackendProperties',
+      'memory-backend-epc':         'MemoryBackendEpcProperties',
       'pr-manager-helper':          'PrManagerHelperProperties',
       'qtest':                      'QtestProperties',
       'rng-builtin':                'RngProperties',
author
#4 lore raw

PATCH v4 4/33 i386: Add 'sgx-epc' device to expose EPC sections to guest

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

SGX EPC is enumerated through CPUID, i.e. EPC "devices" need to be realized prior to realizing the vCPUs themselves, which occurs long before generic devices are parsed and realized. Because of this, do not allow 'sgx-epc' devices to be instantiated after vCPUS have been created.

The 'sgx-epc' device is essentially a placholder at this time, it will be fully implemented in a future patch along with a dedicated command to create 'sgx-epc' devices.

---
 hw/i386/meson.build       |   1 +
 hw/i386/sgx-epc.c         | 161 ++++++++++++++++++++++++++++++++++++++
 include/hw/i386/sgx-epc.h |  44 +++++++++++
 3 files changed, 206 insertions(+)
 create mode 100644 hw/i386/sgx-epc.c
 create mode 100644 include/hw/i386/sgx-epc.h
hw/i386/meson.build +1 -0
index 80dad29f2b..27476b36bb 100644
@@ -5,6 +5,7 @@ i386_ss.add(files(
   'e820_memory_layout.c',
   'multiboot.c',
   'x86.c',
+  'sgx-epc.c',
 ))
 
 i386_ss.add(when: 'CONFIG_X86_IOMMU', if_true: files('x86-iommu.c'),
hw/i386/sgx-epc.c new file +161 -0

161 lines not on this page · raw message

include/hw/i386/sgx-epc.h new file +44 -0

44 lines not on this page · raw message

#5 lore raw

Re: [PATCH v4 04/33] i386: Add 'sgx-epc' device to expose EPC sections to guest

Quoted 47 lines · Yang Zhong

From: Sean Christopherson <[email protected]>

SGX EPC is enumerated through CPUID, i.e. EPC "devices" need to be realized prior to realizing the vCPUs themselves, which occurs long before generic devices are parsed and realized. Because of this, do not allow 'sgx-epc' devices to be instantiated after vCPUS have been created.

The 'sgx-epc' device is essentially a placholder at this time, it will be fully implemented in a future patch along with a dedicated command to create 'sgx-epc' devices.

---
 hw/i386/meson.build       |   1 +
 hw/i386/sgx-epc.c         | 161 ++++++++++++++++++++++++++++++++++++++
 include/hw/i386/sgx-epc.h |  44 +++++++++++
 3 files changed, 206 insertions(+)
 create mode 100644 hw/i386/sgx-epc.c
 create mode 100644 include/hw/i386/sgx-epc.h
hw/i386/meson.build +1 -0
index 80dad29f2b..27476b36bb 100644
@@ -5,6 +5,7 @@ i386_ss.add(files(
   'e820_memory_layout.c',
   'multiboot.c',
   'x86.c',
+  'sgx-epc.c',
 ))
 
 i386_ss.add(when: 'CONFIG_X86_IOMMU', if_true: files('x86-iommu.c'),
hw/i386/sgx-epc.c new file +7 -0
new file mode 100644
index 0000000000..aa487dea79
@@ -0,0 +1,161 @@
+/*
+ * SGX EPC device
+ *
+ * Copyright (C) 2019 Intel Corporation
+ *
+ * Authors:
+ *   Sean Christopherson <[email protected]>

hw/i386/sgx-epc.c line 17

Quoted 10 lines
hw/i386/sgx-epc.c new file +10 -0
+ *
+ * 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 "hw/i386/pc.h"
+#include "hw/i386/sgx-epc.h"
+#include "hw/mem/memory-device.h"
+#include "hw/qdev-properties.h"
+#include "monitor/qdev.h"

Is that include used?

hw/i386/sgx-epc.c line 20

Quoted 3 lines
hw/i386/sgx-epc.c new file +3 -0
+#include "qapi/error.h"
+#include "qapi/visitor.h"
+#include "qemu/config-file.h"

Ditto.

hw/i386/sgx-epc.c line 21

Quoted 1 line
hw/i386/sgx-epc.c new file +1 -0
+#include "qemu/error-report.h"

Ditto.

hw/i386/sgx-epc.c line 22

Quoted 1 line
hw/i386/sgx-epc.c new file +1 -0
+#include "qemu/option.h"

Ditto.

hw/i386/sgx-epc.c line 23

Quoted 1 line
hw/i386/sgx-epc.c new file +1 -0
+#include "qemu/units.h"

Ditto.

Quoted 2 lines
hw/i386/sgx-epc.c new file +2 -0
+#include "target/i386/cpu.h"
+#include "exec/address-spaces.h"
author
#6 lore raw

Re: [PATCH v4 04/33] i386: Add 'sgx-epc' device to expose EPC sections to guest

Quoted 71 lines · Philippe Mathieu-Daudé
Quoted 47 lines · Yang Zhong

From: Sean Christopherson <[email protected]>

SGX EPC is enumerated through CPUID, i.e. EPC "devices" need to be realized prior to realizing the vCPUs themselves, which occurs long before generic devices are parsed and realized. Because of this, do not allow 'sgx-epc' devices to be instantiated after vCPUS have been created.

The 'sgx-epc' device is essentially a placholder at this time, it will be fully implemented in a future patch along with a dedicated command to create 'sgx-epc' devices.

---
 hw/i386/meson.build       |   1 +
 hw/i386/sgx-epc.c         | 161 ++++++++++++++++++++++++++++++++++++++
 include/hw/i386/sgx-epc.h |  44 +++++++++++
 3 files changed, 206 insertions(+)
 create mode 100644 hw/i386/sgx-epc.c
 create mode 100644 include/hw/i386/sgx-epc.h
hw/i386/meson.build +1 -0
index 80dad29f2b..27476b36bb 100644
@@ -5,6 +5,7 @@ i386_ss.add(files(
   'e820_memory_layout.c',
   'multiboot.c',
   'x86.c',
+  'sgx-epc.c',
 ))
 
 i386_ss.add(when: 'CONFIG_X86_IOMMU', if_true: files('x86-iommu.c'),
hw/i386/sgx-epc.c new file +7 -0
new file mode 100644
index 0000000000..aa487dea79
@@ -0,0 +1,161 @@
+/*
+ * SGX EPC device
+ *
+ * Copyright (C) 2019 Intel Corporation
+ *
+ * Authors:
+ *   Sean Christopherson <[email protected]>

hw/i386/sgx-epc.c line 17

Quoted 10 lines
hw/i386/sgx-epc.c new file +10 -0
+ *
+ * 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 "hw/i386/pc.h"
+#include "hw/i386/sgx-epc.h"
+#include "hw/mem/memory-device.h"
+#include "hw/qdev-properties.h"
+#include "monitor/qdev.h"

Is that include used?

hw/i386/sgx-epc.c line 20

Quoted 3 lines
hw/i386/sgx-epc.c new file +3 -0
+#include "qapi/error.h"
+#include "qapi/visitor.h"
+#include "qemu/config-file.h"

Ditto.

hw/i386/sgx-epc.c line 21

Quoted 1 line
hw/i386/sgx-epc.c new file +1 -0
+#include "qemu/error-report.h"

Ditto.

Quoted 8 lines

hw/i386/sgx-epc.c line 22

Quoted 1 line
hw/i386/sgx-epc.c new file +1 -0
+#include "qemu/option.h"

Ditto.

hw/i386/sgx-epc.c line 23

Quoted 1 line
hw/i386/sgx-epc.c new file +1 -0
+#include "qemu/units.h"

Ditto.

Thanks Philippe, those header files have been removed in the Paolo's gitlab(sgx branch), thanks!

Yang

Quoted 2 lines
Quoted 2 lines
hw/i386/sgx-epc.c new file +2 -0
+#include "target/i386/cpu.h"
+#include "exec/address-spaces.h"
author
#7 lore raw

PATCH v4 5/33 vl: Add sgx compound properties to expose SGX EPC sections to guest

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

Because SGX EPC is enumerated through CPUID, EPC "devices" need to be realized prior to realizing the vCPUs themselves, i.e. long before generic devices are parsed and realized. From a virtualization perspective, the CPUID aspect also means that EPC sections cannot be hotplugged without paravirtualizing the guest kernel (hardware does not support hotplugging as EPC sections must be locked down during pre-boot to provide EPC's security properties).

So even though EPC sections could be realized through the generic -devices command, they need to be created much earlier for them to actually be usable by the guest. Place all EPC sections in a contiguous block, somewhat arbitrarily starting after RAM above 4g. Ensuring EPC is in a contiguous region simplifies calculations, e.g. device memory base, PCI hole, etc..., allows dynamic calculation of the total EPC size, e.g. exposing EPC to guests does not require -maxmem, and last but not least allows all of EPC to be enumerated in a single ACPI entry, which is expected by some kernels, e.g. Windows 7 and 8.

The new compound properties command for sgx like below:
......
-object memory-backend-epc,id=mem1,size=28M,prealloc=on \
-object memory-backend-epc,id=mem2,size=10M \
-M sgx-epc.0.memdev=mem1,sgx-epc.1.memdev=mem2

v3-->v4:
- Moved sgx compound property setter/getter from MachineState to X86MachineState(Paolo).
- Re-defined struct SgxEPC, removed 'id' property and added struct SgxEPCList for sgx-epc.0.{memdev}(Paolo).
- Removed g_malloc0(), and changed the 'SGXEPCState *sgx_epc' to 'SGXEPCState sgx_epc' in struct PCMachineState(Paolo).
- Changed the SGX compound property cmdline from sgx-epc.{memdev}.0

    to sgx-epc.0.{memdev}(Paolo).

v2-->v3:
- Removed the QemuOptsList for sgx-epc virtual device and used the '-M' to replace '-sgx-epc' with compound properties(Paolo).

---
 hw/i386/meson.build       |  1 +
 hw/i386/sgx-epc.c         | 20 ++++++++----
 hw/i386/sgx.c             | 67 +++++++++++++++++++++++++++++++++++++++
 hw/i386/x86.c             | 29 +++++++++++++++++
 include/hw/i386/pc.h      |  6 ++++
 include/hw/i386/sgx-epc.h | 14 ++++++++
 include/hw/i386/x86.h     |  1 +
 qapi/machine.json         | 26 +++++++++++++++
 qemu-options.hx           | 10 ++++--
 9 files changed, 166 insertions(+), 8 deletions(-)
 create mode 100644 hw/i386/sgx.c
hw/i386/meson.build +1 -0
index 27476b36bb..fefce9e4ba 100644
@@ -6,6 +6,7 @@ i386_ss.add(files(
   'multiboot.c',
   'x86.c',
   'sgx-epc.c',
+  'sgx.c'
 ))
 
 i386_ss.add(when: 'CONFIG_X86_IOMMU', if_true: files('x86-iommu.c'),
hw/i386/sgx-epc.c +14 -6
index aa487dea79..924dea22f0 100644
@@ -14,13 +14,8 @@
 #include "hw/i386/sgx-epc.h"
 #include "hw/mem/memory-device.h"
 #include "hw/qdev-properties.h"
-#include "monitor/qdev.h"
 #include "qapi/error.h"
 #include "qapi/visitor.h"
-#include "qemu/config-file.h"
-#include "qemu/error-report.h"
-#include "qemu/option.h"
-#include "qemu/units.h"
 #include "target/i386/cpu.h"
 #include "exec/address-spaces.h"
 
@@ -56,6 +51,8 @@ static void sgx_epc_realize(DeviceState *dev, Error **errp)
 {
     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
     X86MachineState *x86ms = X86_MACHINE(pcms);
+    MemoryDeviceState *md = MEMORY_DEVICE(dev);
+    SGXEPCState *sgx_epc = &pcms->sgx_epc;
     SGXEPCDevice *epc = SGX_EPC(dev);
     const char *path;
 
@@ -74,7 +71,18 @@ static void sgx_epc_realize(DeviceState *dev, Error **errp)
         return;
     }
 
-    error_setg(errp, "'" TYPE_SGX_EPC "' not supported");
+    epc->addr = sgx_epc->base + sgx_epc->size;
+
+    memory_region_add_subregion(&sgx_epc->mr, epc->addr - sgx_epc->base,
+                                host_memory_backend_get_memory(epc->hostmem));
+
+    host_memory_backend_set_mapped(epc->hostmem, true);
+
+    sgx_epc->sections = g_renew(SGXEPCDevice *, sgx_epc->sections,
+                                sgx_epc->nr_sections + 1);
+    sgx_epc->sections[sgx_epc->nr_sections++] = epc;
+
+    sgx_epc->size += memory_device_get_region_size(md, errp);
 }
 
 static void sgx_epc_unrealize(DeviceState *dev)
hw/i386/sgx.c new file +67 -0

67 lines not on this page · raw message

hw/i386/x86.c +29 -0
index 00448ed55a..41ef9a84a9 100644
@@ -30,6 +30,8 @@
 #include "qapi/error.h"
 #include "qapi/qmp/qerror.h"
 #include "qapi/qapi-visit-common.h"
+#include "qapi/clone-visitor.h"
+#include "qapi/qapi-visit-machine.h"
 #include "qapi/visitor.h"
 #include "sysemu/qtest.h"
 #include "sysemu/whpx.h"
@@ -1263,6 +1265,27 @@ static void x86_machine_set_bus_lock_ratelimit(Object *obj, Visitor *v,
     visit_type_uint64(v, name, &x86ms->bus_lock_ratelimit, errp);
 }
 
+static void machine_get_sgx_epc(Object *obj, Visitor *v, const char *name,
+                                void *opaque, Error **errp)
+{
+    X86MachineState *x86ms = X86_MACHINE(obj);
+    SgxEPCList *list = x86ms->sgx_epc_list;
+
+    visit_type_SgxEPCList(v, name, &list, errp);
+}
+
+static void machine_set_sgx_epc(Object *obj, Visitor *v, const char *name,
+                                void *opaque, Error **errp)
+{
+    X86MachineState *x86ms = X86_MACHINE(obj);
+    SgxEPCList *list;
+
+    list = x86ms->sgx_epc_list;
+    visit_type_SgxEPCList(v, name, &x86ms->sgx_epc_list, errp);
+
+    qapi_free_SgxEPCList(list);
+}
+
 static void x86_machine_initfn(Object *obj)
 {
     X86MachineState *x86ms = X86_MACHINE(obj);
@@ -1322,6 +1345,12 @@ static void x86_machine_class_init(ObjectClass *oc, void *data)
                                 x86_machine_set_bus_lock_ratelimit, NULL, NULL);
     object_class_property_set_description(oc, X86_MACHINE_BUS_LOCK_RATELIMIT,
             "Set the ratelimit for the bus locks acquired in VMs");
+
+    object_class_property_add(oc, "sgx-epc", "SgxEPC",
+        machine_get_sgx_epc, machine_set_sgx_epc,
+        NULL, NULL);
+    object_class_property_set_description(oc, "sgx-epc",
+        "SGX EPC device");
 }
 
 static const TypeInfo x86_machine_info = {
include/hw/i386/pc.h +6 -0
index 88dffe7517..2fd6e6193b 100644
@@ -12,6 +12,7 @@
 #include "hw/acpi/acpi_dev_interface.h"
 #include "hw/hotplug.h"
 #include "qom/object.h"
+#include "hw/i386/sgx-epc.h"
 
 #define HPET_INTCAP "hpet-intcap"
 
@@ -53,6 +54,8 @@ typedef struct PCMachineState {
 
     /* ACPI Memory hotplug IO base address */
     hwaddr memhp_io_base;
+
+    SGXEPCState sgx_epc;
 } PCMachineState;
 
 #define PC_MACHINE_ACPI_DEVICE_PROP "acpi-device"
@@ -199,6 +202,9 @@ void pc_madt_cpu_entry(AcpiDeviceIf *adev, int uid,
 extern GlobalProperty pc_compat_6_0[];
 extern const size_t pc_compat_6_0_len;
 
+/* sgx-epc.c */
+void pc_machine_init_sgx_epc(PCMachineState *pcms);
+
 extern GlobalProperty pc_compat_5_2[];
 extern const size_t pc_compat_5_2_len;
 
include/hw/i386/sgx-epc.h +14 -0
index 5fd9ae2d0c..2b2490892b 100644
@@ -41,4 +41,18 @@ typedef struct SGXEPCDevice {
     HostMemoryBackend *hostmem;
 } SGXEPCDevice;
 
+/*
+ * @base: address in guest physical address space where EPC regions start
+ * @mr: address space container for memory devices
+ */
+typedef struct SGXEPCState {
+    uint64_t base;
+    uint64_t size;
+
+    MemoryRegion mr;
+
+    struct SGXEPCDevice **sections;
+    int nr_sections;
+} SGXEPCState;
+
 #endif
include/hw/i386/x86.h +1 -0
index 6e9244a82c..23267a3674 100644
@@ -62,6 +62,7 @@ struct X86MachineState {
     unsigned pci_irq_mask;
     unsigned apic_id_limit;
     uint16_t boot_cpus;
+    SgxEPCList *sgx_epc_list;
 
     OnOffAuto smm;
     OnOffAuto acpi;
qapi/machine.json +26 -0
index c3210ee1fb..3bbe5b590e 100644
@@ -1194,6 +1194,32 @@
           }
 }
 
+##
+# @SgxEPC:
+#
+# Sgx EPC cmdline information
+#
+# @memdev: memory backend linked with device
+#
+# Since: 6.1
+##
+{ 'struct': 'SgxEPC',
+  'data': { 'memdev': 'str' }
+}
+
+##
+# @SgxEPCProperties:
+#
+# Properties for SgxEPC objects.
+#
+# @sgx-epc: sgx epc section properties.
+#
+# Since: 6.1
+##
+{ 'struct': 'SgxEPCProperties',
+  'data': { 'sgx-epc': ['SgxEPC'] }
+}
+
 ##
 # @MemoryDeviceInfo:
 #
qemu-options.hx +8 -2
index 0c9ddc0274..dd05426f58 100644
@@ -122,8 +122,14 @@ SRST
         -m 512M
 ERST
 
-HXCOMM Deprecated by -machine
-DEF("M", HAS_ARG, QEMU_OPTION_M, "", QEMU_ARCH_ALL)
+DEF("M", HAS_ARG, QEMU_OPTION_M,
+    "                sgx-epc.0.memdev=memid\n",
+    QEMU_ARCH_ALL)
+
+SRST
+``sgx-epc.0.memdev=@var{memid}``
+    Define an SGX EPC section.
+ERST
 
 DEF("cpu", HAS_ARG, QEMU_OPTION_cpu,
     "-cpu cpu        select CPU ('-cpu help' for list)\n", QEMU_ARCH_ALL)
author
#8 lore raw

PATCH v4 6/33 i386: Add primary SGX CPUID and MSR defines

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

Add CPUID defines for SGX and SGX Launch Control (LC), as well as defines for their associated FEATURE_CONTROL MSR bits. Define the Launch Enclave Public Key Hash MSRs (LE Hash MSRs), which exist when SGX LC is present (in CPUID), and are writable when SGX LC is enabled (in FEATURE_CONTROL).

---
 target/i386/cpu.c |  4 ++--
 target/i386/cpu.h | 12 ++++++++++++
 2 files changed, 14 insertions(+), 2 deletions(-)
target/i386/cpu.c +2 -2
index 48b55ebd0a..7796d1caa6 100644
@@ -794,7 +794,7 @@ FeatureWordInfo feature_word_info[FEATURE_WORDS] = {
     [FEAT_7_0_EBX] = {
         .type = CPUID_FEATURE_WORD,
         .feat_names = {
-            "fsgsbase", "tsc-adjust", NULL, "bmi1",
+            "fsgsbase", "tsc-adjust", "sgx", "bmi1",
             "hle", "avx2", NULL, "smep",
             "bmi2", "erms", "invpcid", "rtm",
             NULL, NULL, "mpx", NULL,
@@ -820,7 +820,7 @@ FeatureWordInfo feature_word_info[FEATURE_WORDS] = {
             "la57", NULL, NULL, NULL,
             NULL, NULL, "rdpid", NULL,
             "bus-lock-detect", "cldemote", NULL, "movdiri",
-            "movdir64b", NULL, NULL, "pks",
+            "movdir64b", NULL, "sgxlc", "pks",
         },
         .cpuid = {
             .eax = 7,
target/i386/cpu.h +12 -0
index 5d98a4e7c0..af33f956fe 100644
@@ -377,9 +377,17 @@ typedef enum X86Seg {
 #define MSR_IA32_PKRS                   0x6e1
 
 #define FEATURE_CONTROL_LOCKED                    (1<<0)
+#define FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX  (1ULL << 1)
 #define FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX (1<<2)
+#define FEATURE_CONTROL_SGX_LC                    (1ULL << 17)
+#define FEATURE_CONTROL_SGX                       (1ULL << 18)
 #define FEATURE_CONTROL_LMCE                      (1<<20)
 
+#define MSR_IA32_SGXLEPUBKEYHASH0       0x8c
+#define MSR_IA32_SGXLEPUBKEYHASH1       0x8d
+#define MSR_IA32_SGXLEPUBKEYHASH2       0x8e
+#define MSR_IA32_SGXLEPUBKEYHASH3       0x8f
+
 #define MSR_P6_PERFCTR0                 0xc1
 
 #define MSR_IA32_SMBASE                 0x9e
@@ -701,6 +709,8 @@ typedef uint64_t FeatureWordArray[FEATURE_WORDS];
 
 /* Support RDFSBASE/RDGSBASE/WRFSBASE/WRGSBASE */
 #define CPUID_7_0_EBX_FSGSBASE          (1U << 0)
+/* Support SGX */
+#define CPUID_7_0_EBX_SGX               (1U << 2)
 /* 1st Group of Advanced Bit Manipulation Extensions */
 #define CPUID_7_0_EBX_BMI1              (1U << 3)
 /* Hardware Lock Elision */
@@ -788,6 +798,8 @@ typedef uint64_t FeatureWordArray[FEATURE_WORDS];
 #define CPUID_7_0_ECX_MOVDIRI           (1U << 27)
 /* Move 64 Bytes as Direct Store Instruction */
 #define CPUID_7_0_ECX_MOVDIR64B         (1U << 28)
+/* Support SGX Launch Control */
+#define CPUID_7_0_ECX_SGX_LC            (1U << 30)
 /* Protection Keys for Supervisor-mode Pages */
 #define CPUID_7_0_ECX_PKS               (1U << 31)
 
author
#9 lore raw

PATCH v4 7/33 i386: Add SGX CPUID leaf FEAT_SGX_12_0_EAX

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

CPUID leaf 12_0_EAX is an Intel-defined feature bits leaf enumerating the CPU's SGX capabilities, e.g. supported SGX instruction sets. Currently there are four enumerated capabilities:

- SGX1 instruction set, i.e. "base" SGX
- SGX2 instruction set for dynamic EPC management
- ENCLV instruction set for VMM oversubscription of EPC
- ENCLS-C instruction set for thread safe variants of ENCLS

---
 target/i386/cpu.c | 20 ++++++++++++++++++++
 target/i386/cpu.h |  1 +
 2 files changed, 21 insertions(+)
target/i386/cpu.c +20 -0
index 7796d1caa6..13289e0547 100644
@@ -653,6 +653,7 @@ void x86_cpu_vendor_words2str(char *dst, uint32_t vendor1,
           /* missing:
           CPUID_XSAVE_XSAVEC, CPUID_XSAVE_XSAVES */
 #define TCG_14_0_ECX_FEATURES 0
+#define TCG_SGX_12_0_EAX_FEATURES 0
 
 FeatureWordInfo feature_word_info[FEATURE_WORDS] = {
     [FEAT_1_EDX] = {
@@ -1181,6 +1182,25 @@ FeatureWordInfo feature_word_info[FEATURE_WORDS] = {
         .tcg_features = TCG_14_0_ECX_FEATURES,
      },
 
+    [FEAT_SGX_12_0_EAX] = {
+        .type = CPUID_FEATURE_WORD,
+        .feat_names = {
+            "sgx1", "sgx2", NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+        },
+        .cpuid = {
+            .eax = 0x12,
+            .needs_ecx = true, .ecx = 0,
+            .reg = R_EAX,
+        },
+        .tcg_features = TCG_SGX_12_0_EAX_FEATURES,
+    },
 };
 
 typedef struct FeatureMask {
target/i386/cpu.h +1 -0
index af33f956fe..21830b2d79 100644
@@ -561,6 +561,7 @@ typedef enum FeatureWord {
     FEAT_VMX_BASIC,
     FEAT_VMX_VMFUNC,
     FEAT_14_0_ECX,
+    FEAT_SGX_12_0_EAX,  /* CPUID[EAX=0x12,ECX=0].EAX (SGX) */
     FEATURE_WORDS,
 } FeatureWord;
 
author
#10 lore raw

PATCH v4 8/33 i386: Add SGX CPUID leaf FEAT_SGX_12_0_EBX

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

CPUID leaf 12_0_EBX is an Intel-defined feature bits leaf enumerating the platform's SGX extended capabilities. Currently there is a single capabilitiy:

- EXINFO: record information about #PFs and #GPs in the enclave's SSA

---
 target/i386/cpu.c | 21 +++++++++++++++++++++
 target/i386/cpu.h |  1 +
 2 files changed, 22 insertions(+)
target/i386/cpu.c +21 -0
index 13289e0547..0525a4cf7f 100644
@@ -654,6 +654,7 @@ void x86_cpu_vendor_words2str(char *dst, uint32_t vendor1,
           CPUID_XSAVE_XSAVEC, CPUID_XSAVE_XSAVES */
 #define TCG_14_0_ECX_FEATURES 0
 #define TCG_SGX_12_0_EAX_FEATURES 0
+#define TCG_SGX_12_0_EBX_FEATURES 0
 
 FeatureWordInfo feature_word_info[FEATURE_WORDS] = {
     [FEAT_1_EDX] = {
@@ -1201,6 +1202,26 @@ FeatureWordInfo feature_word_info[FEATURE_WORDS] = {
         },
         .tcg_features = TCG_SGX_12_0_EAX_FEATURES,
     },
+
+    [FEAT_SGX_12_0_EBX] = {
+        .type = CPUID_FEATURE_WORD,
+        .feat_names = {
+            "sgx-exinfo" , NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+        },
+        .cpuid = {
+            .eax = 0x12,
+            .needs_ecx = true, .ecx = 0,
+            .reg = R_EBX,
+        },
+        .tcg_features = TCG_SGX_12_0_EBX_FEATURES,
+    },
 };
 
 typedef struct FeatureMask {
target/i386/cpu.h +1 -0
index 21830b2d79..a71d1c8fd0 100644
@@ -562,6 +562,7 @@ typedef enum FeatureWord {
     FEAT_VMX_VMFUNC,
     FEAT_14_0_ECX,
     FEAT_SGX_12_0_EAX,  /* CPUID[EAX=0x12,ECX=0].EAX (SGX) */
+    FEAT_SGX_12_0_EBX,  /* CPUID[EAX=0x12,ECX=0].EBX (SGX MISCSELECT[31:0]) */
     FEATURE_WORDS,
 } FeatureWord;
 
author
#11 lore raw

PATCH v4 9/33 i386: Add SGX CPUID leaf FEAT_SGX_12_1_EAX

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

CPUID leaf 12_1_EAX is an Intel-defined feature bits leaf enumerating the platform's SGX capabilities that may be utilized by an enclave, e.g. whether or not an enclave can gain access to the provision key. Currently there are six capabilities:

- INIT: set when the enclave has has been initialized by EINIT. Cannot be set by software, i.e. forced to zero in CPUID.
- DEBUG: permits a debugger to read/write into the enclave.
- MODE64BIT: the enclave runs in 64-bit mode
- PROVISIONKEY: grants has access to the provision key
- EINITTOKENKEY: grants access to the EINIT token key, i.e. the enclave can generate EINIT tokens
- KSS: Key Separation and Sharing enabled for the enclave.

Note that the entirety of CPUID.0x12.0x1, i.e. all registers, enumerates the allowed ATTRIBUTES (128 bits), but only bits 31:0 are directly exposed to the user (via FEAT_12_1_EAX). Bits 63:32 are currently all reserved and bits 127:64 correspond to the allowed XSAVE Feature Request Mask, which is calculated based on other CPU features, e.g. XSAVE, MPX, AVX, etc... and is not exposed to the user.

---
 target/i386/cpu.c | 21 +++++++++++++++++++++
 target/i386/cpu.h |  1 +
 2 files changed, 22 insertions(+)
target/i386/cpu.c +21 -0
index 0525a4cf7f..840f825431 100644
@@ -655,6 +655,7 @@ void x86_cpu_vendor_words2str(char *dst, uint32_t vendor1,
 #define TCG_14_0_ECX_FEATURES 0
 #define TCG_SGX_12_0_EAX_FEATURES 0
 #define TCG_SGX_12_0_EBX_FEATURES 0
+#define TCG_SGX_12_1_EAX_FEATURES 0
 
 FeatureWordInfo feature_word_info[FEATURE_WORDS] = {
     [FEAT_1_EDX] = {
@@ -1222,6 +1223,26 @@ FeatureWordInfo feature_word_info[FEATURE_WORDS] = {
         },
         .tcg_features = TCG_SGX_12_0_EBX_FEATURES,
     },
+
+    [FEAT_SGX_12_1_EAX] = {
+        .type = CPUID_FEATURE_WORD,
+        .feat_names = {
+            NULL, "sgx-debug", "sgx-mode64", NULL,
+            "sgx-provisionkey", "sgx-tokenkey", NULL, "sgx-kss",
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+            NULL, NULL, NULL, NULL,
+        },
+        .cpuid = {
+            .eax = 0x12,
+            .needs_ecx = true, .ecx = 1,
+            .reg = R_EAX,
+        },
+        .tcg_features = TCG_SGX_12_1_EAX_FEATURES,
+    },
 };
 
 typedef struct FeatureMask {
target/i386/cpu.h +1 -0
index a71d1c8fd0..e4d46cca80 100644
@@ -563,6 +563,7 @@ typedef enum FeatureWord {
     FEAT_14_0_ECX,
     FEAT_SGX_12_0_EAX,  /* CPUID[EAX=0x12,ECX=0].EAX (SGX) */
     FEAT_SGX_12_0_EBX,  /* CPUID[EAX=0x12,ECX=0].EBX (SGX MISCSELECT[31:0]) */
+    FEAT_SGX_12_1_EAX,  /* CPUID[EAX=0x12,ECX=1].EAX (SGX ATTRIBUTES[31:0]) */
     FEATURE_WORDS,
 } FeatureWord;
 
author
#12 lore raw

PATCH v4 10/33 i386: Add get/set/migrate support for SGX_LEPUBKEYHASH MSRs

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

On real hardware, on systems that supports SGX Launch Control, those MSRs are initialized to digest of Intel's signing key; on systems that don't support SGX Launch Control, those MSRs are not available but hardware always uses digest of Intel's signing key in EINIT.

KVM advertises SGX LC via CPUID if and only if the MSRs are writable. Unconditionally initialize those MSRs to digest of Intel's signing key when CPU is realized and reset to reflect the fact. This avoids potential bug in case kvm_arch_put_registers() is called before kvm_arch_get_registers() is called, in which case guest's virtual SGX_LEPUBKEYHASH MSRs will be set to 0, although KVM initializes those to digest of Intel's signing key by default, since KVM allows those MSRs to be updated by Qemu to support live migration.

Save/restore the SGX Launch Enclave Public Key Hash MSRs if SGX Launch Control (LC) is exposed to the guest. Likewise, migrate the MSRs if they are writable by the guest.

---
 target/i386/cpu.c     | 17 ++++++++++++++++-
 target/i386/cpu.h     |  1 +
 target/i386/kvm/kvm.c | 22 ++++++++++++++++++++++
 target/i386/machine.c | 20 ++++++++++++++++++++
 4 files changed, 59 insertions(+), 1 deletion(-)
target/i386/cpu.c +16 -1
index 840f825431..cea4307930 100644
@@ -5673,6 +5673,16 @@ void cpu_x86_cpuid(CPUX86State *env, uint32_t index, uint32_t count,
     }
 }
 
+#ifndef CONFIG_USER_ONLY
+static void x86_cpu_set_sgxlepubkeyhash(CPUX86State *env)
+{
+    env->msr_ia32_sgxlepubkeyhash[0] = 0xa6053e051270b7acULL;
+    env->msr_ia32_sgxlepubkeyhash[1] = 0x6cfbe8ba8b3b413dULL;
+    env->msr_ia32_sgxlepubkeyhash[2] = 0xc4916d99f2b3735dULL;
+    env->msr_ia32_sgxlepubkeyhash[3] = 0xd4f8c05909f9bb3bULL;
+}
+#endif
+
 static void x86_cpu_reset(DeviceState *dev)
 {
     CPUState *s = CPU(dev);
@@ -5804,6 +5814,8 @@ static void x86_cpu_reset(DeviceState *dev)
     if (kvm_enabled()) {
         kvm_arch_reset_vcpu(cpu);
     }
+
+    x86_cpu_set_sgxlepubkeyhash(env);
 #endif
 }
 
@@ -6186,6 +6198,10 @@ static void x86_cpu_realizefn(DeviceState *dev, Error **errp)
            & CPUID_EXT2_AMD_ALIASES);
     }
 
+#ifndef CONFIG_USER_ONLY
+    x86_cpu_set_sgxlepubkeyhash(env);
+#endif
+
     /*
      * note: the call to the framework needs to happen after feature expansion,
      * but before the checks/modifications to ucode_rev, mwait, phys_bits.
@@ -6871,7 +6887,6 @@ static const TypeInfo x86_cpu_type_info = {
     .class_init = x86_cpu_common_class_init,
 };
 
-
 /* "base" CPU model, used by query-cpu-model-expansion */
 static void x86_cpu_base_class_init(ObjectClass *oc, void *data)
 {
target/i386/cpu.h +1 -0
index e4d46cca80..892c0dfab4 100644
@@ -1499,6 +1499,7 @@ typedef struct CPUX86State {
     uint64_t mcg_status;
     uint64_t msr_ia32_misc_enable;
     uint64_t msr_ia32_feature_control;
+    uint64_t msr_ia32_sgxlepubkeyhash[4];
 
     uint64_t msr_fixed_ctr_ctrl;
     uint64_t msr_global_ctrl;
target/i386/kvm/kvm.c +22 -0
index 59ed8327ac..d4bf054ebe 100644
@@ -3096,6 +3096,17 @@ static int kvm_put_msrs(X86CPU *cpu, int level)
             }
         }
 
+        if (env->features[FEAT_7_0_ECX] & CPUID_7_0_ECX_SGX_LC) {
+            kvm_msr_entry_add(cpu, MSR_IA32_SGXLEPUBKEYHASH0,
+                              env->msr_ia32_sgxlepubkeyhash[0]);
+            kvm_msr_entry_add(cpu, MSR_IA32_SGXLEPUBKEYHASH1,
+                              env->msr_ia32_sgxlepubkeyhash[1]);
+            kvm_msr_entry_add(cpu, MSR_IA32_SGXLEPUBKEYHASH2,
+                              env->msr_ia32_sgxlepubkeyhash[2]);
+            kvm_msr_entry_add(cpu, MSR_IA32_SGXLEPUBKEYHASH3,
+                              env->msr_ia32_sgxlepubkeyhash[3]);
+        }
+
         /* Note: MSR_IA32_FEATURE_CONTROL is written separately, see
          *       kvm_put_msr_feature_control. */
     }
@@ -3435,6 +3446,13 @@ static int kvm_get_msrs(X86CPU *cpu)
         }
     }
 
+    if (env->features[FEAT_7_0_ECX] & CPUID_7_0_ECX_SGX_LC) {
+        kvm_msr_entry_add(cpu, MSR_IA32_SGXLEPUBKEYHASH0, 0);
+        kvm_msr_entry_add(cpu, MSR_IA32_SGXLEPUBKEYHASH1, 0);
+        kvm_msr_entry_add(cpu, MSR_IA32_SGXLEPUBKEYHASH2, 0);
+        kvm_msr_entry_add(cpu, MSR_IA32_SGXLEPUBKEYHASH3, 0);
+    }
+
     ret = kvm_vcpu_ioctl(CPU(cpu), KVM_GET_MSRS, cpu->kvm_msr_buf);
     if (ret < 0) {
         return ret;
@@ -3724,6 +3742,10 @@ static int kvm_get_msrs(X86CPU *cpu)
         case MSR_IA32_RTIT_ADDR0_A ... MSR_IA32_RTIT_ADDR3_B:
             env->msr_rtit_addrs[index - MSR_IA32_RTIT_ADDR0_A] = msrs[i].data;
             break;
+        case MSR_IA32_SGXLEPUBKEYHASH0 ... MSR_IA32_SGXLEPUBKEYHASH3:
+            env->msr_ia32_sgxlepubkeyhash[index - MSR_IA32_SGXLEPUBKEYHASH0] =
+                           msrs[i].data;
+            break;
         }
     }
 
target/i386/machine.c +20 -0
index f6f094f1c9..099a4c36f7 100644
@@ -1396,6 +1396,25 @@ static const VMStateDescription vmstate_msr_tsx_ctrl = {
     }
 };
 
+static bool intel_sgx_msrs_needed(void *opaque)
+{
+    X86CPU *cpu = opaque;
+    CPUX86State *env = &cpu->env;
+
+    return !!(env->features[FEAT_7_0_ECX] & CPUID_7_0_ECX_SGX_LC);
+}
+
+static const VMStateDescription vmstate_msr_intel_sgx = {
+    .name = "cpu/intel_sgx",
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .needed = intel_sgx_msrs_needed,
+    .fields = (VMStateField[]) {
+        VMSTATE_UINT64_ARRAY(env.msr_ia32_sgxlepubkeyhash, X86CPU, 4),
+        VMSTATE_END_OF_LIST()
+    }
+};
+
 const VMStateDescription vmstate_x86_cpu = {
     .name = "cpu",
     .version_id = 12,
@@ -1531,6 +1550,7 @@ const VMStateDescription vmstate_x86_cpu = {
         &vmstate_nested_state,
 #endif
         &vmstate_msr_tsx_ctrl,
+        &vmstate_msr_intel_sgx,
         NULL
     }
 };
#13 lore raw

Re: [PATCH v4 10/33] i386: Add get/set/migrate support for SGX_LEPUBKEYHASH MSRs

Quoted 37 lines · Yang Zhong

From: Sean Christopherson <[email protected]>

On real hardware, on systems that supports SGX Launch Control, those MSRs are initialized to digest of Intel's signing key; on systems that don't support SGX Launch Control, those MSRs are not available but hardware always uses digest of Intel's signing key in EINIT.

KVM advertises SGX LC via CPUID if and only if the MSRs are writable. Unconditionally initialize those MSRs to digest of Intel's signing key when CPU is realized and reset to reflect the fact. This avoids potential bug in case kvm_arch_put_registers() is called before kvm_arch_get_registers() is called, in which case guest's virtual SGX_LEPUBKEYHASH MSRs will be set to 0, although KVM initializes those to digest of Intel's signing key by default, since KVM allows those MSRs to be updated by Qemu to support live migration.

Save/restore the SGX Launch Enclave Public Key Hash MSRs if SGX Launch Control (LC) is exposed to the guest. Likewise, migrate the MSRs if they are writable by the guest.

---
 target/i386/cpu.c     | 17 ++++++++++++++++-
 target/i386/cpu.h     |  1 +
 target/i386/kvm/kvm.c | 22 ++++++++++++++++++++++
 target/i386/machine.c | 20 ++++++++++++++++++++
 4 files changed, 59 insertions(+), 1 deletion(-)
target/i386/cpu.c +0 -0
index 840f825431..cea4307930 100644
@@ -5673,6 +5673,16 @@ void cpu_x86_cpuid(CPUX86State *env, uint32_t index, uint32_t count,
     }
 }

target/i386/cpu.c void cpu_x86_cpuid(CPUX86State *env, uint32_t index, uint32_t count, · line 5684

Quoted 10 lines
target/i386/cpu.c +9 -0
 
+#ifndef CONFIG_USER_ONLY
+static void x86_cpu_set_sgxlepubkeyhash(CPUX86State *env)
+{
+    env->msr_ia32_sgxlepubkeyhash[0] = 0xa6053e051270b7acULL;
+    env->msr_ia32_sgxlepubkeyhash[1] = 0x6cfbe8ba8b3b413dULL;
+    env->msr_ia32_sgxlepubkeyhash[2] = 0xc4916d99f2b3735dULL;
+    env->msr_ia32_sgxlepubkeyhash[3] = 0xd4f8c05909f9bb3bULL;
+}
+#endif

Maybe easier to move the #ifdef'ry inside the function.

Where these values come from btw?

Quoted 8 lines
target/i386/cpu.c +4 -0
@@ -6186,6 +6198,10 @@ static void x86_cpu_realizefn(DeviceState *dev, Error **errp)
            & CPUID_EXT2_AMD_ALIASES);
     }
 
+#ifndef CONFIG_USER_ONLY
+    x86_cpu_set_sgxlepubkeyhash(env);
+#endif
+
author
#14 lore raw

Re: [PATCH v4 10/33] i386: Add get/set/migrate support for SGX_LEPUBKEYHASH MSRs

Quoted 39 lines · Philippe Mathieu-Daudé
Quoted 38 lines · Yang Zhong

From: Sean Christopherson <[email protected]>

On real hardware, on systems that supports SGX Launch Control, those MSRs are initialized to digest of Intel's signing key; on systems that don't support SGX Launch Control, those MSRs are not available but hardware always uses digest of Intel's signing key in EINIT.

KVM advertises SGX LC via CPUID if and only if the MSRs are writable. Unconditionally initialize those MSRs to digest of Intel's signing key when CPU is realized and reset to reflect the fact. This avoids potential bug in case kvm_arch_put_registers() is called before kvm_arch_get_registers() is called, in which case guest's virtual SGX_LEPUBKEYHASH MSRs will be set to 0, although KVM initializes those to digest of Intel's signing key by default, since KVM allows those MSRs to be updated by Qemu to support live migration.

Save/restore the SGX Launch Enclave Public Key Hash MSRs if SGX Launch Control (LC) is exposed to the guest. Likewise, migrate the MSRs if they are writable by the guest.

---
 target/i386/cpu.c     | 17 ++++++++++++++++-
 target/i386/cpu.h     |  1 +
 target/i386/kvm/kvm.c | 22 ++++++++++++++++++++++
 target/i386/machine.c | 20 ++++++++++++++++++++
 4 files changed, 59 insertions(+), 1 deletion(-)
target/i386/cpu.c +0 -0
index 840f825431..cea4307930 100644
@@ -5673,6 +5673,16 @@ void cpu_x86_cpuid(CPUX86State *env, uint32_t index, uint32_t count,
     }
 }
 
Quoted 12 lines

target/i386/cpu.c void cpu_x86_cpuid(CPUX86State *env, uint32_t index, uint32_t count, · line 5684

Quoted 9 lines
target/i386/cpu.c +9 -0
+#ifndef CONFIG_USER_ONLY
+static void x86_cpu_set_sgxlepubkeyhash(CPUX86State *env)
+{
+    env->msr_ia32_sgxlepubkeyhash[0] = 0xa6053e051270b7acULL;
+    env->msr_ia32_sgxlepubkeyhash[1] = 0x6cfbe8ba8b3b413dULL;
+    env->msr_ia32_sgxlepubkeyhash[2] = 0xc4916d99f2b3735dULL;
+    env->msr_ia32_sgxlepubkeyhash[3] = 0xd4f8c05909f9bb3bULL;
+}
+#endif

Maybe easier to move the #ifdef'ry inside the function.

Thanks for comments, since this function is pure void function, we can move this #ifdef into function.

Quoted 1 line

Where these values come from btw?

Those MSR values are intel default values, which were defined in Skylake platform.

Yang

Quoted 9 lines
Quoted 9 lines
target/i386/cpu.c +4 -0
@@ -6186,6 +6198,10 @@ static void x86_cpu_realizefn(DeviceState *dev, Error **errp)
            & CPUID_EXT2_AMD_ALIASES);
     }
 
+#ifndef CONFIG_USER_ONLY
+    x86_cpu_set_sgxlepubkeyhash(env);
+#endif
+
#15 lore raw

Re: [PATCH v4 10/33] i386: Add get/set/migrate support for SGX_LEPUBKEYHASH MSRs

Quoted 53 lines · Yang Zhong
Quoted 39 lines · Philippe Mathieu-Daudé
Quoted 38 lines · Yang Zhong

From: Sean Christopherson <[email protected]>

On real hardware, on systems that supports SGX Launch Control, those MSRs are initialized to digest of Intel's signing key; on systems that don't support SGX Launch Control, those MSRs are not available but hardware always uses digest of Intel's signing key in EINIT.

KVM advertises SGX LC via CPUID if and only if the MSRs are writable. Unconditionally initialize those MSRs to digest of Intel's signing key when CPU is realized and reset to reflect the fact. This avoids potential bug in case kvm_arch_put_registers() is called before kvm_arch_get_registers() is called, in which case guest's virtual SGX_LEPUBKEYHASH MSRs will be set to 0, although KVM initializes those to digest of Intel's signing key by default, since KVM allows those MSRs to be updated by Qemu to support live migration.

Save/restore the SGX Launch Enclave Public Key Hash MSRs if SGX Launch Control (LC) is exposed to the guest. Likewise, migrate the MSRs if they are writable by the guest.

---
 target/i386/cpu.c     | 17 ++++++++++++++++-
 target/i386/cpu.h     |  1 +
 target/i386/kvm/kvm.c | 22 ++++++++++++++++++++++
 target/i386/machine.c | 20 ++++++++++++++++++++
 4 files changed, 59 insertions(+), 1 deletion(-)
target/i386/cpu.c +0 -0
index 840f825431..cea4307930 100644
@@ -5673,6 +5673,16 @@ void cpu_x86_cpuid(CPUX86State *env, uint32_t index, uint32_t count,
     }
 }
 
Quoted 12 lines

target/i386/cpu.c void cpu_x86_cpuid(CPUX86State *env, uint32_t index, uint32_t count, · line 5684

Quoted 9 lines
target/i386/cpu.c +9 -0
+#ifndef CONFIG_USER_ONLY
+static void x86_cpu_set_sgxlepubkeyhash(CPUX86State *env)
+{
+    env->msr_ia32_sgxlepubkeyhash[0] = 0xa6053e051270b7acULL;
+    env->msr_ia32_sgxlepubkeyhash[1] = 0x6cfbe8ba8b3b413dULL;
+    env->msr_ia32_sgxlepubkeyhash[2] = 0xc4916d99f2b3735dULL;
+    env->msr_ia32_sgxlepubkeyhash[3] = 0xd4f8c05909f9bb3bULL;
+}
+#endif

Maybe easier to move the #ifdef'ry inside the function.

Quoted 6 lines

Thanks for comments, since this function is pure void function, we can move this #ifdef into function.

Quoted 1 line

Where these values come from btw?

Those MSR values are intel default values, which were defined in Skylake platform.

Could you add a comment (and reference if possible) about them please?

Thanks,

Phil.

author
#16 lore raw

PATCH v4 11/33 i386: Add feature control MSR dependency when SGX is enabled

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

SGX adds multiple flags to FEATURE_CONTROL to enable SGX and Flexible Launch Control.

---
 target/i386/kvm/kvm.c | 5 +++++
 1 file changed, 5 insertions(+)
target/i386/kvm/kvm.c +5 -0
index d4bf054ebe..8d017fb5a6 100644
@@ -1865,6 +1865,11 @@ int kvm_arch_init_vcpu(CPUState *cs)
                                   !!(c->ecx & CPUID_EXT_SMX);
     }
 
+    c = cpuid_find_entry(&cpuid_data.cpuid, 7, 0);
+    if (c && (c->ebx & CPUID_7_0_EBX_SGX)) {
+        has_msr_feature_control = true;
+    }
+
     if (env->mcg_cap & MCG_LMCE_P) {
         has_msr_mcg_ext_ctl = has_msr_feature_control = true;
     }
author
#17 lore raw

PATCH v4 12/33 i386: Update SGX CPUID info according to hardware/KVM/user input

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

Expose SGX to the guest if and only if KVM is enabled and supports virtualization of SGX. While the majority of ENCLS can be emulated to some degree, because SGX uses a hardware-based root of trust, the attestation aspects of SGX cannot be emulated in software, i.e. ultimately emulation will fail as software cannot generate a valid quote/report. The complexity of partially emulating SGX in Qemu far outweighs the value added, e.g. an SGX specific simulator for userspace applications can emulate SGX for development and testing purposes.

Note, access to the PROVISIONKEY is not yet advertised to the guest as KVM blocks access to the PROVISIONKEY by default and requires userspace to provide additional credentials (via ioctl()) to expose PROVISIONKEY.

v3-->v4:
- Replaced g_malloc0() with directly ....

---
 hw/i386/sgx.c             | 17 +++++++++
 include/hw/i386/sgx-epc.h |  2 +
 target/i386/cpu.c         | 77 +++++++++++++++++++++++++++++++++++++++
 3 files changed, 96 insertions(+)
hw/i386/sgx.c +17 -0
index e77deb0b00..5f988c6368 100644
@@ -18,6 +18,23 @@
 #include "qapi/error.h"
 #include "exec/address-spaces.h"
 
+int sgx_epc_get_section(int section_nr, uint64_t *addr, uint64_t *size)
+{
+    PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
+    SGXEPCDevice *epc;
+
+    if (pcms->sgx_epc.size == 0 || pcms->sgx_epc.nr_sections <= section_nr) {
+        return 1;
+    }
+
+    epc = pcms->sgx_epc.sections[section_nr];
+
+    *addr = epc->addr;
+    *size = memory_device_get_region_size(MEMORY_DEVICE(epc), &error_fatal);
+
+    return 0;
+}
+
 static int sgx_epc_set_property(void *opaque, const char *name,
                                 const char *value, Error **errp)
 {
include/hw/i386/sgx-epc.h +2 -0
index 2b2490892b..f85fd2a4ca 100644
@@ -55,4 +55,6 @@ typedef struct SGXEPCState {
     int nr_sections;
 } SGXEPCState;
 
+int sgx_epc_get_section(int section_nr, uint64_t *addr, uint64_t *size);
+
 #endif
target/i386/cpu.c +77 -0

95 lines not on this page · raw message

author
#18 lore raw

PATCH v4 13/33 i386: kvm: Add support for exposing PROVISIONKEY to guest

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

If the guest want to fully use SGX, the guest needs to be able to access provisioning key. Add a new KVM_CAP_SGX_ATTRIBUTE to KVM to support provisioning key to KVM guests.

---
 target/i386/cpu.c          |  5 ++++-
 target/i386/kvm/kvm.c      | 29 +++++++++++++++++++++++++++++
 target/i386/kvm/kvm_i386.h |  2 ++
 3 files changed, 35 insertions(+), 1 deletion(-)
target/i386/cpu.c +4 -1
index 8adfe80312..6ddeba7461 100644
@@ -5509,7 +5509,10 @@ void cpu_x86_cpuid(CPUX86State *env, uint32_t index, uint32_t count,
             *ecx |= XSTATE_FP_MASK | XSTATE_SSE_MASK;
 
             /* Access to PROVISIONKEY requires additional credentials. */
-            *eax &= ~(1U << 4);
+            if ((*eax & (1U << 4)) &&
+                !kvm_enable_sgx_provisioning(cs->kvm_state)) {
+                *eax &= ~(1U << 4);
+            }
         }
 #endif
         break;
target/i386/kvm/kvm.c +29 -0
index 8d017fb5a6..05a1c370c3 100644
@@ -4633,6 +4633,35 @@ void kvm_arch_update_guest_debug(CPUState *cpu, struct kvm_guest_debug *dbg)
     }
 }
 
+static bool has_sgx_provisioning;
+
+static bool __kvm_enable_sgx_provisioning(KVMState *s)
+{
+    int fd, ret;
+
+    if (!kvm_vm_check_extension(s, KVM_CAP_SGX_ATTRIBUTE)) {
+        return false;
+    }
+
+    fd = qemu_open_old("/dev/sgx_provision", O_RDONLY);
+    if (fd < 0) {
+        return false;
+    }
+
+    ret = kvm_vm_enable_cap(s, KVM_CAP_SGX_ATTRIBUTE, 0, fd);
+    if (ret) {
+        error_report("Could not enable SGX PROVISIONKEY: %s", strerror(-ret));
+        exit(1);
+    }
+    close(fd);
+    return true;
+}
+
+bool kvm_enable_sgx_provisioning(KVMState *s)
+{
+    return MEMORIZE(__kvm_enable_sgx_provisioning(s), has_sgx_provisioning);
+}
+
 static bool host_supports_vmx(void)
 {
     uint32_t ecx, unused;
target/i386/kvm/kvm_i386.h +2 -0
index 54667b35f0..a978509d50 100644
@@ -51,4 +51,6 @@ bool kvm_hyperv_expand_features(X86CPU *cpu, Error **errp);
 
 uint64_t kvm_swizzle_msi_ext_dest_id(uint64_t address);
 
+bool kvm_enable_sgx_provisioning(KVMState *s);
+
 #endif
author
#19 lore raw

PATCH v4 14/33 i386: Propagate SGX CPUID sub-leafs to KVM

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

The SGX sub-leafs are enumerated at CPUID 0x12. Indices 0 and 1 are always present when SGX is supported, and enumerate SGX features and capabilities. Indices >=2 are directly correlated with the platform's EPC sections. Because the number of EPC sections is dynamic and user defined, the number of SGX sub-leafs is "NULL" terminated.

---
 target/i386/kvm/kvm.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
target/i386/kvm/kvm.c +19 -0
index 05a1c370c3..c8179dcfb7 100644
@@ -1691,6 +1691,25 @@ int kvm_arch_init_vcpu(CPUState *cs)
             }
             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: {
             uint32_t times;
 
author
#20 lore raw

PATCH v4 15/33 Adjust min CPUID level to 0x12 when SGX is enabled

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

SGX capabilities are enumerated through CPUID_0x12.

---
 target/i386/cpu.c | 5 +++++
 1 file changed, 5 insertions(+)
target/i386/cpu.c +5 -0
index 6ddeba7461..1f2f6bdf2d 100644
@@ -6125,6 +6125,11 @@ void x86_cpu_expand_features(X86CPU *cpu, Error **errp)
         if (sev_enabled()) {
             x86_cpu_adjust_level(cpu, &env->cpuid_min_xlevel, 0x8000001F);
         }
+
+        /* SGX requires CPUID[0x12] for EPC enumeration */
+        if (env->features[FEAT_7_0_EBX] & CPUID_7_0_EBX_SGX) {
+            x86_cpu_adjust_level(cpu, &env->cpuid_min_level, 0x12);
+        }
     }
 
     /* Set cpuid_*level* based on cpuid_min_*level, if not explicitly set */
author
#21 lore raw

PATCH v4 16/33 hw/i386/fw_cfg: Set SGX bits in feature control fw_cfg accordingly

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

Request SGX an SGX Launch Control to be enabled in FEATURE_CONTROL when the features are exposed to the guest. Our design is the SGX Launch Control bit will be unconditionally set in FEATURE_CONTROL, which is unlike host bios.

---
 hw/i386/fw_cfg.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
hw/i386/fw_cfg.c +9 -1
index 4e68d5dea4..f6d036dfbe 100644
@@ -159,7 +159,7 @@ void fw_cfg_build_feature_control(MachineState *ms, FWCfgState *fw_cfg)
 {
     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
     CPUX86State *env = &cpu->env;
-    uint32_t unused, ecx, edx;
+    uint32_t unused, ebx, ecx, edx;
     uint64_t feature_control_bits = 0;
     uint64_t *val;
 
@@ -174,6 +174,14 @@ void fw_cfg_build_feature_control(MachineState *ms, FWCfgState *fw_cfg)
         feature_control_bits |= FEATURE_CONTROL_LMCE;
     }
 
+    cpu_x86_cpuid(env, 0x7, 0, &unused, &ebx, &ecx, &unused);
+    if (ebx & CPUID_7_0_EBX_SGX) {
+        feature_control_bits |= FEATURE_CONTROL_SGX;
+    }
+    if (ecx & CPUID_7_0_ECX_SGX_LC) {
+        feature_control_bits |= FEATURE_CONTROL_SGX_LC;
+    }
+
     if (!feature_control_bits) {
         return;
     }
author
#22 lore raw

PATCH v4 17/33 hw/i386/pc: Account for SGX EPC sections when calculating device memory

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

Add helpers to detect if SGX EPC exists above 4g, and if so, where SGX EPC above 4g ends. Use the helpers to adjust the device memory range if SGX EPC exists above 4g.

For multiple virtual EPC sections, we just put them together physically contiguous for the simplicity because we don't support EPC NUMA affinity now. Once the SGX EPC NUMA support in the kernel SGX driver, we will support this in the future.

Note that SGX EPC is currently hardcoded to reside above 4g.

---
 hw/i386/pc.c              | 11 ++++++++++-
 include/hw/i386/sgx-epc.h |  7 +++++++
 2 files changed, 17 insertions(+), 1 deletion(-)
hw/i386/pc.c +10 -1
index c2b9d62a35..109e835890 100644
@@ -920,8 +920,15 @@ void pc_memory_init(PCMachineState *pcms,
             exit(EXIT_FAILURE);
         }
 
+        if (pcms->sgx_epc.size != 0) {
+            machine->device_memory->base = sgx_epc_above_4g_end(&pcms->sgx_epc);
+        } else {
+            machine->device_memory->base =
+                0x100000000ULL + x86ms->above_4g_mem_size;
+        }
+
         machine->device_memory->base =
-            ROUND_UP(0x100000000ULL + x86ms->above_4g_mem_size, 1 * GiB);
+            ROUND_UP(machine->device_memory->base, 1 * GiB);
 
         if (pcmc->enforce_aligned_dimm) {
             /* size device region assuming 1G page max alignment per slot */
@@ -1006,6 +1013,8 @@ uint64_t pc_pci_hole64_start(void)
         if (!pcmc->broken_reserved_end) {
             hole64_start += memory_region_size(&ms->device_memory->mr);
         }
+    } else if (pcms->sgx_epc.size != 0) {
+            hole64_start = sgx_epc_above_4g_end(&pcms->sgx_epc);
     } else {
         hole64_start = 0x100000000ULL + x86ms->above_4g_mem_size;
     }
include/hw/i386/sgx-epc.h +7 -0
index f85fd2a4ca..155e8fad3e 100644
@@ -57,4 +57,11 @@ typedef struct SGXEPCState {
 
 int sgx_epc_get_section(int section_nr, uint64_t *addr, uint64_t *size);
 
+static inline uint64_t sgx_epc_above_4g_end(SGXEPCState *sgx_epc)
+{
+    assert(sgx_epc != NULL && sgx_epc->base >= 0x100000000ULL);
+
+    return sgx_epc->base + sgx_epc->size;
+}
+
 #endif
author
#23 lore raw

PATCH v4 18/33 i386/pc: Add e820 entry for SGX EPC section(s)

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

Note that SGX EPC is currently guaranteed to reside in a single contiguous chunk of memory regardless of the number of EPC sections.

---
 hw/i386/pc.c | 4 ++++
 1 file changed, 4 insertions(+)
hw/i386/pc.c +4 -0
index 109e835890..80d3b5e62d 100644
@@ -890,6 +890,10 @@ void pc_memory_init(PCMachineState *pcms,
         e820_add_entry(0x100000000ULL, x86ms->above_4g_mem_size, E820_RAM);
     }
 
+    if (pcms->sgx_epc.size != 0) {
+        e820_add_entry(pcms->sgx_epc.base, pcms->sgx_epc.size, E820_RESERVED);
+    }
+
     if (!pcmc->has_reserved_memory &&
         (machine->ram_slots ||
          (machine->maxram_size > machine->ram_size))) {
author
#24 lore raw

PATCH v4 19/33 i386: acpi: Add SGX EPC entry to ACPI tables

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

The ACPI Device entry for SGX EPC is essentially a hack whose primary purpose is to provide software with a way to autoprobe SGX support, e.g. to allow software to implement SGX support as a driver. Details on the individual EPC sections are not enumerated through ACPI tables, i.e. software must enumerate the EPC sections via CPUID. Furthermore, software expects to see only a single EPC Device in the ACPI tables regardless of the number of EPC sections in the system.

However, several versions of Windows do rely on the ACPI tables to enumerate the address and size of the EPC. So, regardless of the number of EPC sections exposed to the guest, create exactly *one* EPC device with a _CRS entry that spans the entirety of all EPC sections (which are guaranteed to be contiguous in Qemu).

Note, NUMA support for EPC memory is intentionally not considered as enumerating EPC NUMA information is not yet defined for bare metal.

---
 hw/i386/acpi-build.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)
hw/i386/acpi-build.c +22 -0
index 17836149fe..efebe640ce 100644
@@ -1829,6 +1829,28 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
     }
 #endif
 
+    if (pcms->sgx_epc.size != 0) {
+        uint64_t epc_base = pcms->sgx_epc.base;
+        uint64_t epc_size = pcms->sgx_epc.size;
+
+        dev = aml_device("EPC");
+        aml_append(dev, aml_name_decl("_HID", aml_eisaid("INT0E0C")));
+        aml_append(dev, aml_name_decl("_STR",
+                                      aml_unicode("Enclave Page Cache 1.0")));
+        crs = aml_resource_template();
+        aml_append(crs,
+                   aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED,
+                                    AML_MAX_FIXED, AML_NON_CACHEABLE,
+                                    AML_READ_WRITE, 0, epc_base,
+                                    epc_base + epc_size - 1, 0, epc_size));
+        aml_append(dev, aml_name_decl("_CRS", crs));
+
+        method = aml_method("_STA", 0, AML_NOTSERIALIZED);
+        aml_append(method, aml_return(aml_int(0x0f)));
+        aml_append(dev, method);
+
+        aml_append(sb_scope, dev);
+    }
     aml_append(dsdt, sb_scope);
 
     /* copy AML table into ACPI tables blob and patch header there */
author
#25 lore raw

PATCH v4 20/33 q35: Add support for SGX EPC

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

Enable SGX EPC virtualization, which is currently only support by KVM.

---
 hw/i386/pc_q35.c | 3 +++
 1 file changed, 3 insertions(+)
hw/i386/pc_q35.c +3 -0
index 04b4a4788d..799c4fc61c 100644
@@ -178,6 +178,9 @@ static void pc_q35_init(MachineState *machine)
         x86ms->below_4g_mem_size = machine->ram_size;
     }
 
+    if (x86ms->sgx_epc_list) {
+        pc_machine_init_sgx_epc(pcms);
+    }
     x86_cpus_init(x86ms, pcmc->default_cpu_version);
 
     kvmclock_create(pcmc->kvmclock_create_always);
author
#26 lore raw

PATCH v4 21/33 i440fx: Add support for SGX EPC

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

Enable SGX EPC virtualization, which is currently only support by KVM.

---
 hw/i386/pc_piix.c | 4 ++++
 1 file changed, 4 insertions(+)
hw/i386/pc_piix.c +4 -0
index 30b8bd6ea9..c0a7552b88 100644
@@ -154,6 +154,10 @@ static void pc_init1(MachineState *machine,
         }
     }
 
+    if (x86ms->sgx_epc_list) {
+        pc_machine_init_sgx_epc(pcms);
+    }
+
     x86_cpus_init(x86ms, pcmc->default_cpu_version);
 
     if (pcmc->kvmclock_enabled) {
author
#27 lore raw

PATCH v4 22/33 hostmem-epc: Add the reset interface for EPC backend reset

In reply to Yang Zhong

Add the sgx_memory_backend_reset() interface to handle EPC backend reset when VM is reset. This reset function will destroy previous backend memory region and re-mmap the EPC section for guest.

---
 backends/hostmem-epc.c | 16 ++++++++++++++++
 include/hw/i386/pc.h   |  2 ++
 2 files changed, 18 insertions(+)
backends/hostmem-epc.c +16 -0
index b512a68cb0..3bd1535d82 100644
@@ -16,6 +16,7 @@
 #include "qom/object_interfaces.h"
 #include "qapi/error.h"
 #include "sysemu/hostmem.h"
+#include "hw/i386/pc.h"
 
 #define TYPE_MEMORY_BACKEND_EPC "memory-backend-epc"
 
@@ -55,6 +56,21 @@ sgx_epc_backend_memory_alloc(HostMemoryBackend *backend, Error **errp)
     g_free(name);
 }
 
+void sgx_memory_backend_reset(HostMemoryBackend *backend, int fd,
+                              Error **errp)
+{
+    MemoryRegion *mr = &backend->mr;
+
+    mr->enabled = false;
+
+    /* destroy the old memory region if it exist */
+    if (fd > 0 && mr->destructor) {
+        mr->destructor(mr);
+    }
+
+    sgx_epc_backend_memory_alloc(backend, errp);
+}
+
 static void sgx_epc_backend_instance_init(Object *obj)
 {
     HostMemoryBackend *m = MEMORY_BACKEND(obj);
include/hw/i386/pc.h +2 -0
index 2fd6e6193b..c725dfef5b 100644
@@ -204,6 +204,8 @@ extern const size_t pc_compat_6_0_len;
 
 /* sgx-epc.c */
 void pc_machine_init_sgx_epc(PCMachineState *pcms);
+void sgx_memory_backend_reset(HostMemoryBackend *backend, int fd,
+                              Error **errp);
 
 extern GlobalProperty pc_compat_5_2[];
 extern const size_t pc_compat_5_2_len;
#28 lore raw

Re: [PATCH v4 22/33] hostmem-epc: Add the reset interface for EPC backend reset

Quoted 5 lines · Yang Zhong
backends/hostmem-epc.c +5 -0
+void sgx_memory_backend_reset(HostMemoryBackend *backend, int fd,
+                              Error **errp)
+{
+    MemoryRegion *mr = &backend->mr;
+

backends/hostmem-epc.c sgx_epc_backend_memory_alloc(HostMemoryBackend *backend, Error **errp) · line 73

Quoted 10 lines
backends/hostmem-epc.c +10 -0
+    mr->enabled = false;
+
+    /* destroy the old memory region if it exist */
+    if (fd > 0 && mr->destructor) {
+        mr->destructor(mr);
+    }
+
+    sgx_epc_backend_memory_alloc(backend, errp);
+}
+

Jarkko, Sean, Kai,

this I think is problematic because it has a race window while /dev/sgx_vepc is closed and then reopened. First, the vEPC space could be exhausted by somebody doing another mmap in the meanwhile. Second, somebody might (for whatever reason) remove /dev/sgx_vepc while QEMU runs.

Yang explained to me (offlist) that this is needed because Windows fails to reboot without it. We would need a way to ask Linux to reinitialize the vEPC, that doesn't involve munmap/mmap; this could be for example fallocate(FALLOC_FL_ZERO_RANGE).

What do you all think?

Paolo

#29 lore raw

Re: [PATCH v4 22/33] hostmem-epc: Add the reset interface for EPC backend reset

Quoted 19 lines · Paolo Bonzini
Quoted 5 lines · Yang Zhong
backends/hostmem-epc.c +5 -0
+void sgx_memory_backend_reset(HostMemoryBackend *backend, int fd,
+                              Error **errp)
+{
+    MemoryRegion *mr = &backend->mr;
+

backends/hostmem-epc.c sgx_epc_backend_memory_alloc(HostMemoryBackend *backend, Error **errp) · line 73

Quoted 10 lines
backends/hostmem-epc.c +10 -0
+    mr->enabled = false;
+
+    /* destroy the old memory region if it exist */
+    if (fd > 0 && mr->destructor) {
+        mr->destructor(mr);
+    }
+
+    sgx_epc_backend_memory_alloc(backend, errp);
+}
+

Jarkko, Sean, Kai,

Quoted 4 lines

this I think is problematic because it has a race window while /dev/sgx_vepc is closed and then reopened. First, the vEPC space could be exhausted by somebody doing another mmap in the meanwhile. Second, somebody might (for whatever reason) remove /dev/sgx_vepc while QEMU runs.

Yep, both error cases are possible.

Quoted 6 lines

Yang explained to me (offlist) that this is needed because Windows fails to reboot without it. We would need a way to ask Linux to reinitialize the vEPC, that doesn't involve munmap/mmap; this could be for example fallocate(FALLOC_FL_ZERO_RANGE).

What do you all think?

Mechanically, FALLOC_FL_ZERO_RANGE could work, but it's definitely a stretch of the semantics as the underlying memory would not actually be zeroed.

The only other option that comes to mind is a dedicated ioctl().

#30 lore raw

Re: [PATCH v4 22/33] hostmem-epc: Add the reset interface for EPC backend reset

Quoted 9 lines · Sean Christopherson
Quoted 6 lines

Yang explained to me (offlist) that this is needed because Windows fails to reboot without it. We would need a way to ask Linux to reinitialize the vEPC, that doesn't involve munmap/mmap; this could be for example fallocate(FALLOC_FL_ZERO_RANGE).

What do you all think?

Mechanically, FALLOC_FL_ZERO_RANGE could work, but it's definitely a stretch of the semantics as the underlying memory would not actually be zeroed.

The contents are not visible to anyone, so they might well be zero (not entirely serious, but also not entirely unserious).

Quoted 1 line

The only other option that comes to mind is a dedicated ioctl().

If it is not too restrictive to do it for the whole mapping at once, that would be fine. If it makes sense to do it for a range, however, the effort of defining a ioctl() API is probably not worth it when fallocate() is available.

Either way, I suppose it would be just something like

	/* for fallocate; otherwise just use xa_for_each */
	size = min_t(unsigned long, size, -start);
	end = (start + size - 1) >> PAGE_SHIFT;
	start >>= PAGE_SHIFT;

	/*
	 * Removing in two passes lets us remove SECS pages as well,
	 * since they can only be EREMOVE'd after all their child pages.
	 * An EREMOVE failure on the second pass means that the SECS
	 * page still has children on another instance.  Since it is
	 * still in the xarray, bury our head in the sand and ignore
	 * it; sgx_vepc_release() will deal with it.
	 */
	LIST_HEAD(secs_pages);
        xa_for_each_range(&vepc->page_array, index, entry, start, end) {
                if (!sgx_vepc_free_page(entry))
                        list_add_tail(&epc_page->list, &secs_pages);
        }

        list_for_each_entry_safe(epc_page, tmp, &secs_pages, list) {
		list_del(&epc_page->list);
                sgx_vepc_free_page(epc_page);
        }

So another advantage of the ioctl would be e.g. being able to return the number of successfully EREMOVEd pages. But since QEMU would not do anything with the return value and _also_ bury its head in the sand, that would only be useful if you guys have other uses in mind.

Paolo

#31 lore raw

Re: [PATCH v4 22/33] hostmem-epc: Add the reset interface for EPC backend reset

Quoted 6 lines · Paolo Bonzini
Quoted 5 lines · Sean Christopherson
Quoted 5 lines

Yang explained to me (offlist) that this is needed because Windows fails to reboot without it. We would need a way to ask Linux to reinitialize the vEPC, that doesn't involve munmap/mmap; this could be for example fallocate(FALLOC_FL_ZERO_RANGE).

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

What do you all think?

Mechanically, FALLOC_FL_ZERO_RANGE could work, but it's definitely a stretch of the semantics as the underlying memory would not actually be zeroed.

The contents are not visible to anyone, so they might well be zero (not entirely serious, but also not entirely unserious).

Yeah, it wouldn't be a sticking point, just odd.

Quoted 4 lines
Quoted 1 line

The only other option that comes to mind is a dedicated ioctl().

If it is not too restrictive to do it for the whole mapping at once, that would be fine.

Oooh, rats. That reminds me of a complication. If QEMU creates multiple EPC sections, e.g. for a vNUMA setup, resetting each section individually will fail if the guest did an unclean RESET and a given enclaves has EPC pages from multiple sections. E.g. an SECS in vEPC[X] can have children in vEPC[0..N-1], and all those children need to be removed before the SECS can be removed. Yay SGX!

Quoted 29 lines

If it makes sense to do it for a range, however,
the effort of defining a ioctl() API is probably not worth it when fallocate() is available.

Either way, I suppose it would be just something like

	/* for fallocate; otherwise just use xa_for_each */
	size = min_t(unsigned long, size, -start);
	end = (start + size - 1) >> PAGE_SHIFT;
	start >>= PAGE_SHIFT;

	/*
	 * Removing in two passes lets us remove SECS pages as well,
	 * since they can only be EREMOVE'd after all their child pages.
	 * An EREMOVE failure on the second pass means that the SECS
	 * page still has children on another instance.  Since it is
	 * still in the xarray, bury our head in the sand and ignore
	 * it; sgx_vepc_release() will deal with it.
	 */
	LIST_HEAD(secs_pages);
        xa_for_each_range(&vepc->page_array, index, entry, start, end) {
                if (!sgx_vepc_free_page(entry))
                        list_add_tail(&epc_page->list, &secs_pages);
        }

        list_for_each_entry_safe(epc_page, tmp, &secs_pages, list) {
		list_del(&epc_page->list);
                sgx_vepc_free_page(epc_page);
        }
Quoted 4 lines

So another advantage of the ioctl would be e.g. being able to return the number of successfully EREMOVEd pages. But since QEMU would not do anything with the return value and _also_ bury its head in the sand, that would only be useful if you guys have other uses in mind.

There are two options: 1) QEMU has to handle "failure", or 2) the kernel provides an ioctl() that takes multiple vEPC fds and handles the SECS dependencies. #1 is probably the least awful option. For #2, in addition to the kernel having to deal with multiple fds, it would also need a second list_head object in each page so that it could track which pages failed to be removed. Using the existing list_head would work for now, but it won't work if/when an EPC cgroup is added.

Note, for #1, QEMU would have to potentially do three passes.

1. Remove child pages for a given vEPC.
2. Remove SECS for a given vEPC that were pinned by children in the same vEPC.
3. Remove SECS for all vEPC that were pinned by children in different vEPC.

#32 lore raw

Re: [PATCH v4 22/33] hostmem-epc: Add the reset interface for EPC backend reset

Quoted 20 lines · Sean Christopherson
Quoted 5 lines · Paolo Bonzini
Quoted 1 line · Sean Christopherson

The only other option that comes to mind is a dedicated ioctl().

If it is not too restrictive to do it for the whole mapping at once, that would be fine.

Oooh, rats. That reminds me of a complication. If QEMU creates multiple EPC sections, e.g. for a vNUMA setup, resetting each section individually will fail if the guest did an unclean RESET and a given enclaves has EPC pages from multiple sections. E.g. an SECS in vEPC[X] can have children in vEPC[0..N-1], and all those children need to be removed before the SECS can be removed. Yay SGX!

There are two options: 1) QEMU has to handle "failure", or 2) the kernel provides an ioctl() that takes multiple vEPC fds and handles the SECS dependencies. #1 is probably the least awful option. For #2, in addition to the kernel having to deal with multiple fds, it would also need a second list_head object in each page so that it could track which pages failed to be removed. Using the existing list_head would work for now, but it won't work if/when an EPC cgroup is added.

Quoted 5 lines

Note, for #1, QEMU would have to potentially do three passes.

1. Remove child pages for a given vEPC.
2. Remove SECS for a given vEPC that were pinned by children in the same vEPC.
3. Remove SECS for all vEPC that were pinned by children in different vEPC.

It's also possible that QEMU handles failure, but the kernel does two passes; then QEMU can just do two passes. The kernel will overall do four passes, but:

1) the second (SECS pinned by children in the same vEPC) would be cheaper than a full second pass

2) the fourth would actually do nothing, because there would be no pages failing the EREMOV'al.

A hypothetical other SGX client that only uses one vEPC will do the right thing with a single pass.

Paolo

#33 lore raw

Re: [PATCH v4 22/33] hostmem-epc: Add the reset interface for EPC backend reset

Quoted 27 lines · Paolo Bonzini
Quoted 20 lines · Sean Christopherson
Quoted 5 lines · Paolo Bonzini
Quoted 1 line · Sean Christopherson

The only other option that comes to mind is a dedicated ioctl().

If it is not too restrictive to do it for the whole mapping at once, that would be fine.

Oooh, rats. That reminds me of a complication. If QEMU creates multiple EPC sections, e.g. for a vNUMA setup, resetting each section individually will fail if the guest did an unclean RESET and a given enclaves has EPC pages from multiple sections. E.g. an SECS in vEPC[X] can have children in vEPC[0..N-1], and all those children need to be removed before the SECS can be removed. Yay SGX!

There are two options: 1) QEMU has to handle "failure", or 2) the kernel provides an ioctl() that takes multiple vEPC fds and handles the SECS dependencies. #1 is probably the least awful option. For #2, in addition to the kernel having to deal with multiple fds, it would also need a second list_head object in each page so that it could track which pages failed to be removed. Using the existing list_head would work for now, but it won't work if/when an EPC cgroup is added.

Quoted 5 lines

Note, for #1, QEMU would have to potentially do three passes.

1. Remove child pages for a given vEPC.
2. Remove SECS for a given vEPC that were pinned by children in the same vEPC.
3. Remove SECS for all vEPC that were pinned by children in different vEPC.

Quoted 6 lines

It's also possible that QEMU handles failure, but the kernel does two passes; then QEMU can just do two passes. The kernel will overall do four passes, but:

1) the second (SECS pinned by children in the same vEPC) would be cheaper than a full second pass

The problem is that this would require a list_head (or temp allocations) to track the SECS pages that failed the first time 'round. For vEPC destruction, the kernel can use sgx_epc_page.list because it can take the pages off the active/allocated list, but that's not an option in this case because the presumably-upcoming EPC cgroup needs to keep pages on the list to handle OOM.

The kernel's ioctl/syscall/whatever could return the number of pages that were not freed, or maybe just -EAGAIN, and userspace could use that to know it needs to do another reset to free everything.

My thought for QEMU was to do (bad pseudocode):

	/* Retry to EREMOVE pinned SECS pages if necessary. */
	ret = ioctl(SGX_VEPC_RESET, ...);
	if (ret)
		ret = ioctl(SGX_VEPC_RESET, ...);

	/*
	 * Tag the VM as needed yet another round of resets to ERMOVE SECS pages
	 * that were pinned across vEPC sections.
	 */
	vm->sgx_epc_final_reset_needed = !!ret;
Quoted 5 lines

2) the fourth would actually do nothing, because there would be no pages failing the EREMOV'al.

A hypothetical other SGX client that only uses one vEPC will do the right thing with a single pass.

#34 lore raw

Re: [PATCH v4 22/33] hostmem-epc: Add the reset interface for EPC backend reset

Il ven 10 set 2021, 22:21 Sean Christopherson <[email protected]> ha scritto:

Quoted 8 lines
Quoted 2 lines

It's also possible that QEMU handles failure, but the kernel does two passes; then QEMU can just do two passes. The kernel will overall do

four

Quoted 4 lines

passes, but:

1) the second (SECS pinned by children in the same vEPC) would be cheaper than a full second pass

Quoted 10 lines

The problem is that this would require a list_head (or temp allocations) to track
the SECS pages that failed the first time 'round. For vEPC destruction, the kernel
can use sgx_epc_page.list because it can take the pages off the active/allocated
list, but that's not an option in this case because the presumably-upcoming EPC
cgroup needs to keep pages on the list to handle OOM.

Good point, so yeah: let's go for a ioctl that does full removal, returning the number of failures. I will try and cobble up a patch unless Kai beats me to it.

Thanks for the quick discussion!

Paolo

Quoted 27 lines

The kernel's ioctl/syscall/whatever could return the number of pages that were
not freed, or maybe just -EAGAIN, and userspace could use that to know it needs
to do another reset to free everything.

My thought for QEMU was to do (bad pseudocode):

        /* Retry to EREMOVE pinned SECS pages if necessary. */
        ret = ioctl(SGX_VEPC_RESET, ...);
        if (ret)
                ret = ioctl(SGX_VEPC_RESET, ...);

        /*
         * Tag the VM as needed yet another round of resets to ERMOVE SECS
pages
         * that were pinned across vEPC sections.
         */
        vm->sgx_epc_final_reset_needed = !!ret;
Quoted 5 lines

2) the fourth would actually do nothing, because there would be no pages failing the EREMOV'al.

A hypothetical other SGX client that only uses one vEPC will do the right thing with a single pass.

#35 lore raw

Re: [PATCH v4 22/33] hostmem-epc: Add the reset interface for EPC backend reset

In reply to Paolo Bonzini
Quoted 19 lines · Paolo Bonzini
Quoted 5 lines · Yang Zhong
backends/hostmem-epc.c +5 -0
+void sgx_memory_backend_reset(HostMemoryBackend *backend, int fd,
+                              Error **errp)
+{
+    MemoryRegion *mr = &backend->mr;
+

backends/hostmem-epc.c sgx_epc_backend_memory_alloc(HostMemoryBackend *backend, Error **errp) · line 73

Quoted 10 lines
backends/hostmem-epc.c +10 -0
+    mr->enabled = false;
+
+    /* destroy the old memory region if it exist */
+    if (fd > 0 && mr->destructor) {
+        mr->destructor(mr);
+    }
+
+    sgx_epc_backend_memory_alloc(backend, errp);
+}
+

Jarkko, Sean, Kai,

Quoted 4 lines

this I think is problematic because it has a race window while /dev/sgx_vepc is closed and then reopened. First, the vEPC space could be exhausted by somebody doing another mmap in the meanwhile. Second, somebody might (for whatever reason) remove /dev/sgx_vepc while QEMU runs.

1: Why is it a problem that mmap() could fail?

2: Are you speaking about removing device node? If you have succesfully
mapped /dev/sgx_vepc, that should not have much effect (file refcount).

/Jarkko

#36 lore raw

Re: [PATCH v4 22/33] hostmem-epc: Add the reset interface for EPC backend reset

Quoted 20 lines · Jarkko Sakkinen
Quoted 19 lines · Paolo Bonzini
Quoted 5 lines · Yang Zhong
backends/hostmem-epc.c +5 -0
+void sgx_memory_backend_reset(HostMemoryBackend *backend, int fd,
+                              Error **errp)
+{
+    MemoryRegion *mr = &backend->mr;
+

backends/hostmem-epc.c sgx_epc_backend_memory_alloc(HostMemoryBackend *backend, Error **errp) · line 73

Quoted 10 lines
backends/hostmem-epc.c +10 -0
+    mr->enabled = false;
+
+    /* destroy the old memory region if it exist */
+    if (fd > 0 && mr->destructor) {
+        mr->destructor(mr);
+    }
+
+    sgx_epc_backend_memory_alloc(backend, errp);
+}
+

Jarkko, Sean, Kai,

Quoted 6 lines
Quoted 4 lines

this I think is problematic because it has a race window while /dev/sgx_vepc is closed and then reopened. First, the vEPC space could be exhausted by somebody doing another mmap in the meanwhile. Second, somebody might (for whatever reason) remove /dev/sgx_vepc while QEMU runs.

1: Why is it a problem that mmap() could fail?

The flow in question is QEMU's emulation of a guest RESET. If mmap() fails, QEMU either has to kill the VM or disable SGX. In either case, it's fatal to a running workload/VM.

Quoted 2 lines

2: Are you speaking about removing device node? If you have succesfully
mapped /dev/sgx_vepc, that should not have much effect (file refcount).

Paolo was calling out that doing munmap() before mmap() would allow /dev/sgx_vepc to be removed because QEMU would no longer hold a reference to /dev/sgx_vepc. That would again be fatal to the VM as QEMU would not be able to re-mmap() guest EPC.

#37 lore raw

Re: [PATCH v4 22/33] hostmem-epc: Add the reset interface for EPC backend reset

Quoted 26 lines · Sean Christopherson
Quoted 25 lines · Jarkko Sakkinen
Quoted 23 lines · Paolo Bonzini
Quoted 5 lines · Yang Zhong
backends/hostmem-epc.c +5 -0
+void sgx_memory_backend_reset(HostMemoryBackend *backend, int fd,
+                              Error **errp)
+{
+    MemoryRegion *mr = &backend->mr;
+

backends/hostmem-epc.c sgx_epc_backend_memory_alloc(HostMemoryBackend *backend, Error **errp) · line 73

Quoted 10 lines
backends/hostmem-epc.c +10 -0
+    mr->enabled = false;
+
+    /* destroy the old memory region if it exist */
+    if (fd > 0 && mr->destructor) {
+        mr->destructor(mr);
+    }
+
+    sgx_epc_backend_memory_alloc(backend, errp);
+}
+

Jarkko, Sean, Kai,

this I think is problematic because it has a race window while /dev/sgx_vepc is closed and then reopened. First, the vEPC space could be exhausted by somebody doing another mmap in the meanwhile. Second, somebody might (for whatever reason) remove /dev/sgx_vepc while QEMU runs.

Quoted 5 lines
Quoted 1 line

1: Why is it a problem that mmap() could fail?

The flow in question is QEMU's emulation of a guest RESET. If mmap() fails, QEMU either has to kill the VM or disable SGX. In either case, it's fatal to a running workload/VM.

Thanks for the explanations.

Isn't this more about badly configured system/workloads? That's at least for me the existential question.

I'm interested of legit workloads where this behaviour could still cause any issues.

I'd guess than in e.g. data center environment, you'd have firly strict orchestration for this type of resource so that you know that workloads have an appropriate bandwidth.

/Jarkko

author
#38 lore raw

PATCH v4 23/33 sgx-epc: Add the reset interface for sgx-epc virt device

In reply to Yang Zhong

If the VM is reset, we need make sure sgx virt epc in clean status. Once the VM is reset, and sgx epc virt device will be reseted by reset callback registered by qemu_register_reset(). Since this epc virt device depend on backend, this reset will call backend reset interface to re-mmap epc to guest.

---
 hw/i386/sgx-epc.c | 94 ++++++++++++++++++++++++++++++++++++++++-------
 1 file changed, 81 insertions(+), 13 deletions(-)
hw/i386/sgx-epc.c +81 -13
index 924dea22f0..9880d832d5 100644
@@ -18,6 +18,9 @@
 #include "qapi/visitor.h"
 #include "target/i386/cpu.h"
 #include "exec/address-spaces.h"
+#include "sysemu/reset.h"
+
+uint32_t epc_num;
 
 static Property sgx_epc_properties[] = {
     DEFINE_PROP_UINT64(SGX_EPC_ADDR_PROP, SGXEPCDevice, addr, 0),
@@ -47,12 +50,84 @@ static void sgx_epc_init(Object *obj)
                         NULL, NULL, NULL);
 }
 
+static void sgx_epc_del_subregion(DeviceState *dev)
+{
+    PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
+    SGXEPCState *sgx_epc = &pcms->sgx_epc;
+    SGXEPCDevice *epc = SGX_EPC(dev);
+
+    /* del subregion and related operations */
+    memory_region_del_subregion(&sgx_epc->mr,
+                                host_memory_backend_get_memory(epc->hostmem));
+    host_memory_backend_set_mapped(epc->hostmem, false);
+    g_free(sgx_epc->sections);
+    sgx_epc->sections = NULL;
+
+    /* multiple epc devices, only zero the first time */
+    if (epc_num == sgx_epc->nr_sections) {
+        sgx_epc->size = 0;
+        sgx_epc->nr_sections = 0;
+    }
+}
+
+static void sgx_epc_initialization(DeviceState *dev)
+{
+    PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
+    SGXEPCState *sgx_epc = &pcms->sgx_epc;
+    MemoryDeviceState *md = MEMORY_DEVICE(dev);
+    SGXEPCDevice *epc = SGX_EPC(dev);
+    Error *errp = NULL;
+
+    if (!epc->hostmem) {
+        error_setg(&errp, "'" SGX_EPC_MEMDEV_PROP "' property is not set");
+        return;
+    }
+
+    epc->addr = sgx_epc->base + sgx_epc->size;
+
+    memory_region_add_subregion(&sgx_epc->mr, epc->addr - sgx_epc->base,
+                                host_memory_backend_get_memory(epc->hostmem));
+
+    host_memory_backend_set_mapped(epc->hostmem, true);
+
+    sgx_epc->sections = g_renew(SGXEPCDevice *, sgx_epc->sections,
+                                sgx_epc->nr_sections + 1);
+    sgx_epc->sections[sgx_epc->nr_sections++] = epc;
+
+    sgx_epc->size += memory_device_get_region_size(md, &errp);
+}
+
+static void sgx_epc_reset(void *opaque)
+{
+    DeviceState *dev = opaque;
+    SGXEPCDevice *epc = SGX_EPC(dev);
+    Error *errp = NULL;
+    int fd;
+
+    if (!epc->hostmem) {
+        error_setg(&errp, "'" SGX_EPC_MEMDEV_PROP "' property is not set");
+        return;
+    }
+
+    /* delete subregion and related operations */
+    sgx_epc_del_subregion(dev);
+
+    /* reset sgx backend */
+    fd = memory_region_get_fd(host_memory_backend_get_memory(epc->hostmem));
+    sgx_memory_backend_reset(epc->hostmem, fd, &errp);
+    if (errp) {
+        error_setg(&errp, "failed to call sgx_memory_backend_reset");
+        return;
+    }
+
+    /* re-add subregion and related operations */
+    sgx_epc_initialization(dev);
+}
+
 static void sgx_epc_realize(DeviceState *dev, Error **errp)
 {
     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
     X86MachineState *x86ms = X86_MACHINE(pcms);
-    MemoryDeviceState *md = MEMORY_DEVICE(dev);
-    SGXEPCState *sgx_epc = &pcms->sgx_epc;
     SGXEPCDevice *epc = SGX_EPC(dev);
     const char *path;
 
@@ -71,18 +146,11 @@ static void sgx_epc_realize(DeviceState *dev, Error **errp)
         return;
     }
 
-    epc->addr = sgx_epc->base + sgx_epc->size;
-
-    memory_region_add_subregion(&sgx_epc->mr, epc->addr - sgx_epc->base,
-                                host_memory_backend_get_memory(epc->hostmem));
-
-    host_memory_backend_set_mapped(epc->hostmem, true);
-
-    sgx_epc->sections = g_renew(SGXEPCDevice *, sgx_epc->sections,
-                                sgx_epc->nr_sections + 1);
-    sgx_epc->sections[sgx_epc->nr_sections++] = epc;
+    sgx_epc_initialization(dev);
+    epc_num++;
 
-    sgx_epc->size += memory_device_get_region_size(md, errp);
+    /* register the reset callback for sgx reset */
+    qemu_register_reset(sgx_epc_reset, dev);
 }
 
 static void sgx_epc_unrealize(DeviceState *dev)
#39 lore raw

Re: [PATCH v4 23/33] sgx-epc: Add the reset interface for sgx-epc virt device

Quoted 9 lines · Yang Zhong
hw/i386/sgx-epc.c +9 -0
+static void sgx_epc_del_subregion(DeviceState *dev)
+{
+    PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
+    SGXEPCState *sgx_epc = &pcms->sgx_epc;
+    SGXEPCDevice *epc = SGX_EPC(dev);
+
+    /* del subregion and related operations */
+    memory_region_del_subregion(&sgx_epc->mr,
+                                host_memory_backend_get_memory(epc->hostmem));

hw/i386/sgx-epc.c static void sgx_epc_init(Object *obj) · line 71

Quoted 10 lines
hw/i386/sgx-epc.c +10 -0
+    host_memory_backend_set_mapped(epc->hostmem, false);
+    g_free(sgx_epc->sections);
+    sgx_epc->sections = NULL;
+
+    /* multiple epc devices, only zero the first time */
+    if (epc_num == sgx_epc->nr_sections) {
+        sgx_epc->size = 0;
+        sgx_epc->nr_sections = 0;
+    }
+}

Can it just invoke a method on the memory devices, without reusing the code in sgx_epc_realize? In particular why is it necessary to update sgx_epc->size and sgx_epc->nr_sections?

Paolo

#40 lore raw

Re: [PATCH v4 23/33] sgx-epc: Add the reset interface for sgx-epc virt device

In reply to Yang Zhong
Quoted 16 lines · Yang Zhong

If the VM is reset, we need make sure sgx virt epc in clean status. Once the VM is reset, and sgx epc virt device will be reseted by reset callback registered by qemu_register_reset(). Since this epc virt device depend on backend, this reset will call backend reset interface to re-mmap epc to guest.

---
 hw/i386/sgx-epc.c | 94 ++++++++++++++++++++++++++++++++++++++++-------
 1 file changed, 81 insertions(+), 13 deletions(-)
hw/i386/sgx-epc.c +0 -0
index 924dea22f0..9880d832d5 100644

hw/i386/sgx-epc.c line 23

Quoted 6 lines
hw/i386/sgx-epc.c +3 -0
@@ -18,6 +18,9 @@
 #include "qapi/visitor.h"
 #include "target/i386/cpu.h"
 #include "exec/address-spaces.h"
+#include "sysemu/reset.h"
+
+uint32_t epc_num;

Missing 'static' qualifier.

author
#41 lore raw

Re: [PATCH v4 23/33] sgx-epc: Add the reset interface for sgx-epc virt device

Quoted 12 lines · Philippe Mathieu-Daudé
Quoted 11 lines · Yang Zhong

If the VM is reset, we need make sure sgx virt epc in clean status. Once the VM is reset, and sgx epc virt device will be reseted by reset callback registered by qemu_register_reset(). Since this epc virt device depend on backend, this reset will call backend reset interface to re-mmap epc to guest.

---
 hw/i386/sgx-epc.c | 94 ++++++++++++++++++++++++++++++++++++++++-------
 1 file changed, 81 insertions(+), 13 deletions(-)
Quoted 13 lines

hw/i386/sgx-epc.c line 23

Quoted 11 lines
hw/i386/sgx-epc.c +3 -0
index 924dea22f0..9880d832d5 100644
@@ -18,6 +18,9 @@
 #include "qapi/visitor.h"
 #include "target/i386/cpu.h"
 #include "exec/address-spaces.h"
+#include "sysemu/reset.h"
+
+uint32_t epc_num;

Missing 'static' qualifier.

Philippe, thanks! This Qemu SGX reset solution will be replaced by kernel patchset made by Paolo(https://www.spinics.net/lists/kvm/msg253930.html).

Yang

author
#42 lore raw

PATCH v4 24/33 sgx-epc: Avoid bios reset during sgx epc initialization

In reply to Yang Zhong

Since bios do the reset when qemu boot up, and sgx epc will be reset by the registered reset callback function. Like this, the sgx epc will do two times initialization. This patch will check protected mode from cr0 register, and will bypass reset operation from bios. The reset callback will only accept reset operation from guest.

---
 hw/i386/sgx-epc.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)
hw/i386/sgx-epc.c +19 -0
index 9880d832d5..70075db37c 100644
@@ -19,6 +19,7 @@
 #include "target/i386/cpu.h"
 #include "exec/address-spaces.h"
 #include "sysemu/reset.h"
+#include "sysemu/hw_accel.h"
 
 uint32_t epc_num;
 
@@ -97,6 +98,21 @@ static void sgx_epc_initialization(DeviceState *dev)
     sgx_epc->size += memory_device_get_region_size(md, &errp);
 }
 
+static bool check_reset_from_guest(void)
+{
+    CPUState *cs = first_cpu;
+    X86CPU *cpu = X86_CPU(cs);
+    CPUX86State *env = &cpu->env;
+
+    cpu_synchronize_state(cs);
+
+    if (env->cr[0] & CR0_PE_MASK) {
+        return true;
+    }
+
+    return false;
+}
+
 static void sgx_epc_reset(void *opaque)
 {
     DeviceState *dev = opaque;
@@ -104,6 +120,9 @@ static void sgx_epc_reset(void *opaque)
     Error *errp = NULL;
     int fd;
 
+    if (!check_reset_from_guest())
+        return;
+
     if (!epc->hostmem) {
         error_setg(&errp, "'" SGX_EPC_MEMDEV_PROP "' property is not set");
         return;
author
#43 lore raw

PATCH v4 25/33 hostmem-epc: Make prealloc consistent with qemu cmdline during reset

In reply to Yang Zhong

If qemu cmdline set the prealloc property for sgx epc and VM do the reset the prealloc property will be different with cmdline settings. This patch can make sure same prealloc property setting with cmdline.

---
 backends/hostmem-epc.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
backends/hostmem-epc.c +10 -0
index 3bd1535d82..9836358841 100644
@@ -60,6 +60,8 @@ void sgx_memory_backend_reset(HostMemoryBackend *backend, int fd,
                               Error **errp)
 {
     MemoryRegion *mr = &backend->mr;
+    void *ptr;
+    uint64_t sz;
 
     mr->enabled = false;
 
@@ -69,6 +71,14 @@ void sgx_memory_backend_reset(HostMemoryBackend *backend, int fd,
     }
 
     sgx_epc_backend_memory_alloc(backend, errp);
+
+    ptr = memory_region_get_ram_ptr(&backend->mr);
+    sz = memory_region_size(&backend->mr);
+
+    if (backend->prealloc) {
+        os_mem_prealloc(memory_region_get_fd(&backend->mr), ptr, sz,
+                        backend->prealloc_threads, errp);
+    }
 }
 
 static void sgx_epc_backend_instance_init(Object *obj)
author
#44 lore raw

PATCH v4 26/33 qmp: Add query-sgx command

In reply to Yang Zhong

This QMP query command can be used by some userspaces to retrieve the SGX information when SGX is enabled on Intel platform.

v1-->v2:
- "Since: 5.1" to "Since: 6.1", and grammar error(Eric Blake).

---
 monitor/qmp-cmds.c         |  6 ++++++
 qapi/misc.json             | 42 ++++++++++++++++++++++++++++++++++++++
 tests/qtest/qmp-cmd-test.c |  1 +
 3 files changed, 49 insertions(+)
monitor/qmp-cmds.c +6 -0
index f7d64a6457..d63d59149f 100644
@@ -351,3 +351,9 @@ void qmp_display_reload(DisplayReloadOptions *arg, Error **errp)
         abort();
     }
 }
+
+SGXInfo *qmp_query_sgx(Error **errp)
+{
+    error_setg(errp, QERR_FEATURE_DISABLED, "query-sgx");
+    return NULL;
+}
qapi/misc.json +42 -0

45 lines not on this page · raw message

tests/qtest/qmp-cmd-test.c +1 -0
index c98b78d033..b75f3364f3 100644
@@ -100,6 +100,7 @@ static bool query_is_ignored(const char *cmd)
         /* Success depends on Host or Hypervisor SEV support */
         "query-sev",
         "query-sev-capabilities",
+        "query-sgx",
         NULL
     };
     int i;
author
#45 lore raw

PATCH v4 27/33 hmp: Add 'info sgx' command

In reply to Yang Zhong

The command can be used to show the SGX information in the monitor when SGX is enabled on intel platform.

---
 hmp-commands-info.hx  | 15 +++++++++++++++
 include/monitor/hmp.h |  1 +
 monitor/hmp-cmds.c    |  6 ++++++
 3 files changed, 22 insertions(+)
hmp-commands-info.hx +15 -0
index 27206ac049..4c966e8a6b 100644
@@ -877,3 +877,18 @@ SRST
   ``info dirty_rate``
     Display the vcpu dirty rate information.
 ERST
+
+#if defined(TARGET_I386)
+    {
+        .name       = "sgx",
+        .args_type  = "",
+        .params     = "",
+        .help       = "show intel SGX information",
+        .cmd        = hmp_info_sgx,
+    },
+#endif
+
+SRST
+  ``info sgx``
+    Show intel SGX information.
+ERST
include/monitor/hmp.h +1 -0
index 3baa1058e2..eba5eda996 100644
@@ -131,5 +131,6 @@ void hmp_replay_delete_break(Monitor *mon, const QDict *qdict);
 void hmp_replay_seek(Monitor *mon, const QDict *qdict);
 void hmp_info_dirty_rate(Monitor *mon, const QDict *qdict);
 void hmp_calc_dirty_rate(Monitor *mon, const QDict *qdict);
+void hmp_info_sgx(Monitor *mon, const QDict *qdict);
 
 #endif
monitor/hmp-cmds.c +6 -0
index 0942027208..9ebc86190f 100644
@@ -2230,3 +2230,9 @@ void hmp_info_memory_size_summary(Monitor *mon, const QDict *qdict)
     }
     hmp_handle_error(mon, err);
 }
+
+void hmp_info_sgx(Monitor *mon, const QDict *qdict)
+{
+    error_setg(errp, QERR_FEATURE_DISABLED, "query-sgx");
+    return NULL;
+}
author
#46 lore raw

PATCH v4 28/33 i386: Add sgx_get_info() interface

In reply to Yang Zhong

Add the sgx_get_info() interface for hmp and QMP usage, which will get the SGX info from this API.

---
 hw/i386/sgx.c             | 20 ++++++++++++++++++++
 include/hw/i386/pc.h      |  1 +
 include/hw/i386/sgx-epc.h |  1 +
 monitor/hmp-cmds.c        | 20 ++++++++++++++++++--
 monitor/qmp-cmds.c        | 12 ++++++++++--
 stubs/meson.build         |  1 +
 stubs/sgx-stub.c          |  7 +++++++
 7 files changed, 58 insertions(+), 4 deletions(-)
 create mode 100644 stubs/sgx-stub.c
hw/i386/sgx.c +20 -0
index 5f988c6368..beff7e3cf0 100644
@@ -18,6 +18,26 @@
 #include "qapi/error.h"
 #include "exec/address-spaces.h"
 
+SGXInfo *sgx_get_info(void)
+{
+    SGXInfo *info = NULL;
+    MachineState *ms = MACHINE(qdev_get_machine());
+    X86MachineState *x86ms = X86_MACHINE(qdev_get_machine());
+
+    if (x86ms->sgx_epc_list) {
+        PCMachineState *pcms = PC_MACHINE(ms);
+        SGXEPCState *sgx_epc = &pcms->sgx_epc;
+        info = g_new0(SGXInfo, 1);
+
+        info->sgx = true;
+        info->sgx1 = true;
+        info->sgx2 = true;
+        info->flc = true;
+        info->section_size = sgx_epc->size;
+    }
+    return info;
+}
+
 int sgx_epc_get_section(int section_nr, uint64_t *addr, uint64_t *size)
 {
     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
include/hw/i386/pc.h +1 -0
index c725dfef5b..de351ec48a 100644
@@ -206,6 +206,7 @@ extern const size_t pc_compat_6_0_len;
 void pc_machine_init_sgx_epc(PCMachineState *pcms);
 void sgx_memory_backend_reset(HostMemoryBackend *backend, int fd,
                               Error **errp);
+SGXInfo *sgx_get_info(void);
 
 extern GlobalProperty pc_compat_5_2[];
 extern const size_t pc_compat_5_2_len;
include/hw/i386/sgx-epc.h +1 -0
index 155e8fad3e..8f9c7becb9 100644
@@ -13,6 +13,7 @@
 #define QEMU_SGX_EPC_H
 
 #include "sysemu/hostmem.h"
+#include "qapi/qapi-types-misc.h"
 
 #define TYPE_SGX_EPC "sgx-epc"
 #define SGX_EPC(obj) \
monitor/hmp-cmds.c +18 -2
index 9ebc86190f..053d210535 100644
@@ -2233,6 +2233,22 @@ void hmp_info_memory_size_summary(Monitor *mon, const QDict *qdict)
 
 void hmp_info_sgx(Monitor *mon, const QDict *qdict)
 {
-    error_setg(errp, QERR_FEATURE_DISABLED, "query-sgx");
-    return NULL;
+    SGXInfo *info = qmp_query_sgx(NULL);
+
+    if (info && info->sgx) {
+        monitor_printf(mon, "SGX support: %s\n",
+                       info->sgx ? "enabled" : "disabled");
+        monitor_printf(mon, "SGX1 support: %s\n",
+                       info->sgx1 ? "enabled" : "disabled");
+        monitor_printf(mon, "SGX2 support: %s\n",
+                       info->sgx2 ? "enabled" : "disabled");
+        monitor_printf(mon, "FLC support: %s\n",
+                       info->flc ? "enabled" : "disabled");
+        monitor_printf(mon, "size: %" PRIu64 "\n",
+                       info->section_size);
+    } else {
+        monitor_printf(mon, "SGX is not enabled\n");
+    }
+
+    qapi_free_SGXInfo(info);
 }
monitor/qmp-cmds.c +10 -2
index d63d59149f..48f7708ffe 100644
@@ -40,6 +40,7 @@
 #include "qapi/qmp/qerror.h"
 #include "hw/mem/memory-device.h"
 #include "hw/acpi/acpi_dev_interface.h"
+#include "hw/i386/pc.h"
 
 NameInfo *qmp_query_name(Error **errp)
 {
@@ -354,6 +355,13 @@ void qmp_display_reload(DisplayReloadOptions *arg, Error **errp)
 
 SGXInfo *qmp_query_sgx(Error **errp)
 {
-    error_setg(errp, QERR_FEATURE_DISABLED, "query-sgx");
-    return NULL;
+    SGXInfo *info;
+
+    info = sgx_get_info();
+    if (!info) {
+        error_setg(errp, "SGX features are not available");
+        return NULL;
+    }
+
+    return info;
 }
stubs/meson.build +1 -0
index d3fa8646b3..d6788c3614 100644
@@ -57,3 +57,4 @@ if have_system
 else
   stub_ss.add(files('qdev.c'))
 endif
+stub_ss.add(files('sgx-stub.c'))
stubs/sgx-stub.c new file +7 -0
new file mode 100644
index 0000000000..c2b59a88fd
@@ -0,0 +1,7 @@
+#include "qemu/osdep.h"
+#include "hw/i386/pc.h"
+
+SGXInfo *sgx_get_info(void)
+{
+    return NULL;
+}
author
#47 lore raw

PATCH v4 29/33 bitops: Support 32 and 64 bit mask macro

In reply to Yang Zhong

The Qemu should enable bit mask macro like Linux did in the kernel, the GENMASK(h, l) and GENMASK_ULL(h, l) will set the bit to 1 from l to h bit in the 32 bit or 64 bit long type.

---
 include/qemu/bitops.h | 7 +++++++
 1 file changed, 7 insertions(+)
include/qemu/bitops.h +7 -0
index 110c56e099..130d746ba3 100644
@@ -18,6 +18,7 @@
 
 #define BITS_PER_BYTE           CHAR_BIT
 #define BITS_PER_LONG           (sizeof (unsigned long) * BITS_PER_BYTE)
+#define BITS_PER_LONG_LONG       64
 
 #define BIT(nr)                 (1UL << (nr))
 #define BIT_ULL(nr)             (1ULL << (nr))
@@ -28,6 +29,12 @@
 #define MAKE_64BIT_MASK(shift, length) \
     (((~0ULL) >> (64 - (length))) << (shift))
 
+#define GENMASK(h, l) \
+    (((~0UL) << (l)) & (~0UL >> (BITS_PER_LONG - 1 - (h))))
+
+#define GENMASK_ULL(h, l) \
+    (((~0ULL) << (l)) & (~0ULL >> (BITS_PER_LONG_LONG - 1 - (h))))
+
 /**
  * set_bit - Set a bit in memory
  * @nr: the bit to set
author
#48 lore raw

PATCH v4 30/33 qmp: Add the qmp_query_sgx_capabilities()

In reply to Yang Zhong

Libvirt can use qmp_query_sgx_capabilities() to get the host sgx capabilities.

v1-->v2:
- Changed the blurb error and "Since: 5.1" to "Since: 6.1"(Eric Blake).

---
 hw/i386/sgx.c              | 66 ++++++++++++++++++++++++++++++++++++++
 include/hw/i386/pc.h       |  1 +
 monitor/qmp-cmds.c         |  5 +++
 qapi/misc.json             | 19 +++++++++++
 stubs/sgx-stub.c           |  5 +++
 tests/qtest/qmp-cmd-test.c |  1 +
 6 files changed, 97 insertions(+)
hw/i386/sgx.c +66 -0
index beff7e3cf0..993e885980 100644
@@ -17,6 +17,72 @@
 #include "monitor/qdev.h"
 #include "qapi/error.h"
 #include "exec/address-spaces.h"
+#include "sysemu/hw_accel.h"
+
+#define SGX_MAX_EPC_SECTIONS            8
+#define SGX_CPUID_EPC_INVALID           0x0
+
+/* A valid EPC section. */
+#define SGX_CPUID_EPC_SECTION           0x1
+#define SGX_CPUID_EPC_MASK              GENMASK(3, 0)
+
+static uint64_t sgx_calc_section_metric(uint64_t low, uint64_t high)
+{
+    return (low & GENMASK_ULL(31, 12)) +
+           ((high & GENMASK_ULL(19, 0)) << 32);
+}
+
+static uint64_t sgx_calc_host_epc_section_size(void)
+{
+    uint32_t i, type;
+    uint32_t eax, ebx, ecx, edx;
+    uint64_t size = 0;
+
+    for (i = 0; i < SGX_MAX_EPC_SECTIONS; i++) {
+        host_cpuid(0x12, i + 2, &eax, &ebx, &ecx, &edx);
+
+        type = eax & SGX_CPUID_EPC_MASK;
+        if (type == SGX_CPUID_EPC_INVALID) {
+            break;
+        }
+
+        if (type != SGX_CPUID_EPC_SECTION) {
+            break;
+        }
+
+        size += sgx_calc_section_metric(ecx, edx);
+    }
+
+    return size;
+}
+
+SGXInfo *sgx_get_capabilities(Error **errp)
+{
+    SGXInfo *info = NULL;
+    uint32_t eax, ebx, ecx, edx;
+
+    int fd = qemu_open_old("/dev/sgx_vepc", O_RDWR);
+    if (fd < 0) {
+        error_setg(errp, "SGX is not enabled in KVM");
+        return NULL;
+    }
+
+    info = g_new0(SGXInfo, 1);
+    host_cpuid(0x7, 0, &eax, &ebx, &ecx, &edx);
+
+    info->sgx = ebx & (1U << 2) ? true : false;
+    info->flc = ecx & (1U << 30) ? true : false;
+
+    host_cpuid(0x12, 0, &eax, &ebx, &ecx, &edx);
+    info->sgx1 = eax & (1U << 0) ? true : false;
+    info->sgx2 = eax & (1U << 1) ? true : false;
+
+    info->section_size = sgx_calc_host_epc_section_size();
+
+    close(fd);
+
+    return info;
+}
 
 SGXInfo *sgx_get_info(void)
 {
include/hw/i386/pc.h +1 -0
index de351ec48a..6f687e62bc 100644
@@ -207,6 +207,7 @@ void pc_machine_init_sgx_epc(PCMachineState *pcms);
 void sgx_memory_backend_reset(HostMemoryBackend *backend, int fd,
                               Error **errp);
 SGXInfo *sgx_get_info(void);
+SGXInfo *sgx_get_capabilities(Error **errp);
 
 extern GlobalProperty pc_compat_5_2[];
 extern const size_t pc_compat_5_2_len;
monitor/qmp-cmds.c +5 -0
index 48f7708ffe..f1360e9f4e 100644
@@ -365,3 +365,8 @@ SGXInfo *qmp_query_sgx(Error **errp)
 
     return info;
 }
+
+SGXInfo *qmp_query_sgx_capabilities(Error **errp)
+{
+    return sgx_get_capabilities(errp);
+}
qapi/misc.json +19 -0
index 83b45a1460..8d1a78d138 100644
@@ -561,3 +561,22 @@
 #
 ##
 { 'command': 'query-sgx', 'returns': 'SGXInfo' }
+
+
+##
+# @query-sgx-capabilities:
+#
+# Returns information from host SGX capabilities
+#
+# Returns: @SGXInfo
+#
+# Since: 6.1
+#
+# Example:
+#
+# -> { "execute": "query-sgx-capabilities" }
+# <- { "return": { "sgx": true, "sgx1" : true, "sgx2" : true,
+#                  "flc": true, "section-size" : 0 } }
+#
+##
+{ 'command': 'query-sgx-capabilities', 'returns': 'SGXInfo' }
stubs/sgx-stub.c +5 -0
index c2b59a88fd..1dedf3f3db 100644
@@ -5,3 +5,8 @@ SGXInfo *sgx_get_info(void)
 {
     return NULL;
 }
+
+SGXInfo *sgx_get_capabilities(Error **errp)
+{
+    return NULL;
+}
tests/qtest/qmp-cmd-test.c +1 -0
index b75f3364f3..1af2f74c28 100644
@@ -101,6 +101,7 @@ static bool query_is_ignored(const char *cmd)
         "query-sev",
         "query-sev-capabilities",
         "query-sgx",
+        "query-sgx-capabilities",
         NULL
     };
     int i;
author
#49 lore raw

PATCH v4 31/33 Kconfig: Add CONFIG_SGX support

In reply to Yang Zhong

Add new CONFIG_SGX for sgx support in the Qemu, and the Kconfig default enable sgx in the i386 platform.

---
 backends/meson.build                     |  2 +-
 configs/devices/i386-softmmu/default.mak |  1 +
 hw/i386/Kconfig                          |  5 +++++
 hw/i386/meson.build                      |  4 ++--
 hw/i386/sgx-stub.c                       | 13 +++++++++++++
 5 files changed, 22 insertions(+), 3 deletions(-)
 create mode 100644 hw/i386/sgx-stub.c
backends/meson.build +1 -1
index 46fd16b269..6e68945528 100644
@@ -16,6 +16,6 @@ softmmu_ss.add(when: ['CONFIG_VHOST_USER', 'CONFIG_VIRTIO'], if_true: files('vho
 softmmu_ss.add(when: 'CONFIG_VIRTIO_CRYPTO', if_true: files('cryptodev-vhost.c'))
 softmmu_ss.add(when: ['CONFIG_VIRTIO_CRYPTO', 'CONFIG_VHOST_CRYPTO'], if_true: files('cryptodev-vhost-user.c'))
 softmmu_ss.add(when: 'CONFIG_GIO', if_true: [files('dbus-vmstate.c'), gio])
-softmmu_ss.add(when: 'CONFIG_LINUX', if_true: files('hostmem-epc.c'))
+softmmu_ss.add(when: 'CONFIG_SGX', if_true: files('hostmem-epc.c'))
 
 subdir('tpm')
configs/devices/i386-softmmu/default.mak +1 -0
index 84d1a2487c..598c6646df 100644
@@ -22,6 +22,7 @@
 #CONFIG_TPM_CRB=n
 #CONFIG_TPM_TIS_ISA=n
 #CONFIG_VTD=n
+#CONFIG_SGX=n
 
 # Boards:
 #
hw/i386/Kconfig +5 -0
index bad6cf5b4e..9072f9d030 100644
@@ -6,6 +6,10 @@ config SEV
     select X86_FW_OVMF
     depends on KVM
 
+config SGX
+    bool
+    depends on KVM
+
 config PC
     bool
     imply APPLESMC
@@ -21,6 +25,7 @@ config PC
     imply PVPANIC_ISA
     imply QXL
     imply SEV
+    imply SGX
     imply SGA
     imply TEST_DEVICES
     imply TPM_CRB
hw/i386/meson.build +2 -2
index fefce9e4ba..c502965219 100644
@@ -5,8 +5,6 @@ i386_ss.add(files(
   'e820_memory_layout.c',
   'multiboot.c',
   'x86.c',
-  'sgx-epc.c',
-  'sgx.c'
 ))
 
 i386_ss.add(when: 'CONFIG_X86_IOMMU', if_true: files('x86-iommu.c'),
@@ -18,6 +16,8 @@ i386_ss.add(when: 'CONFIG_Q35', if_true: files('pc_q35.c'))
 i386_ss.add(when: 'CONFIG_VMMOUSE', if_true: files('vmmouse.c'))
 i386_ss.add(when: 'CONFIG_VMPORT', if_true: files('vmport.c'))
 i386_ss.add(when: 'CONFIG_VTD', if_true: files('intel_iommu.c'))
+i386_ss.add(when: 'CONFIG_SGX', if_true: files('sgx-epc.c','sgx.c'),
+                                if_false: files('sgx-stub.c'))
 
 i386_ss.add(when: 'CONFIG_ACPI', if_true: files('acpi-common.c'))
 i386_ss.add(when: 'CONFIG_ACPI_HW_REDUCED', if_true: files('generic_event_device_x86.c'))
hw/i386/sgx-stub.c new file +13 -0
new file mode 100644
index 0000000000..edf17c3309
@@ -0,0 +1,13 @@
+#include "qemu/osdep.h"
+#include "hw/i386/pc.h"
+#include "hw/i386/sgx-epc.h"
+
+void pc_machine_init_sgx_epc(PCMachineState *pcms)
+{
+    return;
+}
+
+int sgx_epc_get_section(int section_nr, uint64_t *addr, uint64_t *size)
+{
+    return 1;
+}
author
#50 lore raw

PATCH v4 32/33 sgx-epc: Add the fill_device_info() callback support

In reply to Yang Zhong

Since there is no fill_device_info() callback support, and when we execute "info memory-devices" command in the monitor, the segfault will be found.

This patch will add this callback support and "info memory-devices" will show sgx epc memory exposed to guest. The result as below:

qemu) info memory-devices
Memory device [sgx-epc]: ""
memaddr: 0x180000000
size: 29360128
memdev: /objects/mem1
Memory device [sgx-epc]: ""
memaddr: 0x181c00000
size: 10485760
memdev: /objects/mem2

v3-->v4:
- Removed the epc device 'id' from compound property(Paolo).

v1-->v2:
- "Since: 5.1" to "Since: 6.1"(Eric Blake).

---
 hw/i386/sgx-epc.c  | 11 ++++++++++-
 monitor/hmp-cmds.c | 10 ++++++++++
 qapi/machine.json  | 26 +++++++++++++++++++++++++-
 3 files changed, 45 insertions(+), 2 deletions(-)
hw/i386/sgx-epc.c +10 -1
index 70075db37c..7fc3dd81b9 100644
@@ -214,7 +214,16 @@ static MemoryRegion *sgx_epc_md_get_memory_region(MemoryDeviceState *md,
 static void sgx_epc_md_fill_device_info(const MemoryDeviceState *md,
                                         MemoryDeviceInfo *info)
 {
-    /* TODO */
+    SgxEPCDeviceInfo *se = g_new0(SgxEPCDeviceInfo, 1);
+    SGXEPCDevice *epc = SGX_EPC(md);
+
+    se->memaddr = epc->addr;
+    se->size = object_property_get_uint(OBJECT(epc), SGX_EPC_SIZE_PROP,
+                                        NULL);
+    se->memdev = object_get_canonical_path(OBJECT(epc->hostmem));
+
+    info->u.sgx_epc.data = se;
+    info->type = MEMORY_DEVICE_INFO_KIND_SGX_EPC;
 }
 
 static void sgx_epc_class_init(ObjectClass *oc, void *data)
monitor/hmp-cmds.c +10 -0
index 053d210535..7dfc98c380 100644
@@ -1823,6 +1823,7 @@ void hmp_info_memory_devices(Monitor *mon, const QDict *qdict)
     VirtioMEMDeviceInfo *vmi;
     MemoryDeviceInfo *value;
     PCDIMMDeviceInfo *di;
+    SgxEPCDeviceInfo *se;
 
     for (info = info_list; info; info = info->next) {
         value = info->value;
@@ -1870,6 +1871,15 @@ void hmp_info_memory_devices(Monitor *mon, const QDict *qdict)
                                vmi->block_size);
                 monitor_printf(mon, "  memdev: %s\n", vmi->memdev);
                 break;
+            case MEMORY_DEVICE_INFO_KIND_SGX_EPC:
+                se = value->u.sgx_epc.data;
+                monitor_printf(mon, "Memory device [%s]: \"%s\"\n",
+                               MemoryDeviceInfoKind_str(value->type),
+                               se->id ? se->id : "");
+                monitor_printf(mon, "  memaddr: 0x%" PRIx64 "\n", se->memaddr);
+                monitor_printf(mon, "  size: %" PRIu64 "\n", se->size);
+                monitor_printf(mon, "  memdev: %s\n", se->memdev);
+                break;
             default:
                 g_assert_not_reached();
             }
qapi/machine.json +25 -1
index 3bbe5b590e..b060635942 100644
@@ -1220,6 +1220,29 @@
   'data': { 'sgx-epc': ['SgxEPC'] }
 }
 
+##
+# @SgxEPCDeviceInfo:
+#
+# Sgx EPC state information
+#
+# @id: device's ID
+#
+# @memaddr: physical address in memory, where device is mapped
+#
+# @size: size of memory that the device provides
+#
+# @memdev: memory backend linked with device
+#
+# Since: 6.1
+##
+{ 'struct': 'SgxEPCDeviceInfo',
+  'data': { '*id': 'str',
+            'memaddr': 'size',
+            'size': 'size',
+            'memdev': 'str'
+          }
+}
+
 ##
 # @MemoryDeviceInfo:
 #
@@ -1234,7 +1257,8 @@
   'data': { 'dimm': 'PCDIMMDeviceInfo',
             'nvdimm': 'PCDIMMDeviceInfo',
             'virtio-pmem': 'VirtioPMEMDeviceInfo',
-            'virtio-mem': 'VirtioMEMDeviceInfo'
+            'virtio-mem': 'VirtioMEMDeviceInfo',
+            'sgx-epc': 'SgxEPCDeviceInfo'
           }
 }
 
author
#51 lore raw

PATCH v4 33/33 doc: Add the SGX doc

In reply to Yang Zhong

From: Sean Christopherson <[email protected]>

---
 docs/intel-sgx.txt | 167 +++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 167 insertions(+)
 create mode 100644 docs/intel-sgx.txt
docs/intel-sgx.txt new file +167 -0
new file mode 100644
index 0000000000..f934d92f37
@@ -0,0 +1,167 @@
+===============================
+Software Guard eXtensions (SGX)
+===============================
+
+Overview
+========
+
+Intel Software Guard eXtensions (SGX) is a set of instructions and mechanisms
+for memory accesses in order to provide security accesses for sensitive
+applications and data. SGX allows an application to use it's pariticular
+address space as an *enclave*, which is a protected area provides confidentiality
+and integrity even in the presence of privileged malware. Accesses to the
+enclave memory area from any software not resident in the enclave are prevented,
+including those from privileged software.
+
+Virtual SGX
+===========
+
+SGX feature is exposed to guest via SGX CPUID. Looking at SGX CPUID, we can
+report the same CPUID info to guest as on host for most of SGX CPUID. With
+reporting the same CPUID guest is able to use full capacity of SGX, and KVM
+doesn't need to emulate those info.
+
+The guest's EPC base and size are determined by Qemu, and KVM needs Qemu to
+notify such info to it before it can initialize SGX for guest.
+
+Virtual EPC
+-----------
+
+By default, Qemu does not assign EPC to a VM, i.e. fully enabling SGX in a VM
+requires explicit allocation of EPC to the VM. Similar to other specialized
+memory types, e.g. hugetlbfs, EPC is exposed as a memory backend.
+
+SGX EPC is enumerated through CPUID, i.e. EPC "devices" need to be realized
+prior to realizing the vCPUs themselves, which occurs long before generic
+devices are parsed and realized.  This limitation means that EPC does not
+require -maxmem as EPC is not treated as {cold,hot}plugged memory.
+
+Qemu does not artificially restrict the number of EPC sections exposed to a
+guest, e.g. Qemu will happily allow you to create 64 1M EPC sections. Be aware
+that some kernels may not recognize all EPC sections, e.g. the Linux SGX driver
+is hardwired to support only 8 EPC sections.
+
+The following Qemu snippet creates two EPC sections, with 64M pre-allocated
+to the VM and an additional 28M mapped but not allocated:
+
+ -object memory-backend-epc,id=mem1,size=64M,prealloc=on \
+ -object memory-backend-epc,id=mem2,size=28M \
+ -M sgx-epc.0.memdev=mem1,sgx-epc.1.memdev=mem2
+
+Note:
+
+The size and location of the virtual EPC are far less restricted compared
+to physical EPC. Because physical EPC is protected via range registers,
+the size of the physical EPC must be a power of two (though software sees
+a subset of the full EPC, e.g. 92M or 128M) and the EPC must be naturally
+aligned.  KVM SGX's virtual EPC is purely a software construct and only
+requires the size and location to be page aligned. Qemu enforces the EPC
+size is a multiple of 4k and will ensure the base of the EPC is 4k aligned.
+To simplify the implementation, EPC is always located above 4g in the guest
+physical address space.
+
+Migration
+---------
+
+Qemu/KVM doesn't prevent live migrating SGX VMs, although from hardware's
+perspective, SGX doesn't support live migration, since both EPC and the SGX
+key hierarchy are bound to the physical platform. However live migration
+can be supported in the sense if guest software stack can support recreating
+enclaves when it suffers sudden lose of EPC; and if guest enclaves can detect
+SGX keys being changed, and handle gracefully. For instance, when ERESUME fails
+with #PF.SGX, guest software can gracefully detect it and recreate enclaves;
+and when enclave fails to unseal sensitive information from outside, it can
+detect such error and sensitive information can be provisioned to it again.
+
+CPUID
+-----
+
+Due to its myriad dependencies, SGX is currently not listed as supported
+in any of Qemu's built-in CPU configuration. To expose SGX (and SGX Launch
+Control) to a guest, you must either use `-cpu host` to pass-through the
+host CPU model, or explicitly enable SGX when using a built-in CPU model,
+e.g. via `-cpu <model>,+sgx` or `-cpu <model>,+sgx,+sgxlc`.
+
+All SGX sub-features enumerated through CPUID, e.g. SGX2, MISCSELECT,
+ATTRIBUTES, etc... can be restricted via CPUID flags. Be aware that enforcing
+restriction of MISCSELECT, ATTRIBUTES and XFRM requires intercepting ECREATE,
+i.e. may marginally reduce SGX performance in the guest. All SGX sub-features
+controlled via -cpu are prefixed with "sgx", e.g.:
+
+$ qemu-system-x86_64 -cpu help | xargs printf "%s\n" | grep sgx
+  sgx
+  sgx-debug
+  sgx-encls-c
+  sgx-enclv
+  sgx-exinfo
+  sgx-kss
+  sgx-mode64
+  sgx-provisionkey
+  sgx-tokenkey
+  sgx1
+  sgx2
+  sgxlc
+
+The following Qemu snippet passes through the host CPU (and host physical
+address width) but restricts access to the provision and EINIT token keys:
+
+ -cpu host,host-phys-bits,-sgx-provisionkey,-sgx-tokenkey
+
+Note:
+
+SGX sub-features cannot be emulated, i.e. sub-features that are not present
+in hardware cannot be forced on via '-cpu'.
+
+Virtualize SGX Launch Control
+-----------------------------
+
+Qemu SGX support for Launch Control (LC) is passive, in the sense that it
+does not actively change the LC configuration.  Qemu SGX provides the user
+the ability to set/clear the CPUID flag (and by extension the associated
+IA32_FEATURE_CONTROL MSR bit in fw_cfg) and saves/restores the LE Hash MSRs
+when getting/putting guest state, but Qemu does not add new controls to
+directly modify the LC configuration.  Similar to hardware behavior, locking
+the LC configuration to a non-Intel value is left to guest firmware.  Unlike
+host bios setting for SGX launch control(LC), there is no special bios setting
+for SGX guest by our design. If host is in locked mode, we can still allow
+creating VM with SGX.
+
+Feature Control
+---------------
+
+Qemu SGX updates the `etc/msr_feature_control` fw_cfg entry to set the SGX
+(bit 18) and SGX LC (bit 17) flags based on their respective CPUID support,
+i.e. existing guest firmware will automatically set SGX and SGX LC accordingly,
+assuming said firmware supports fw_cfg.msr_feature_control.
+
+Launch a guest
+==============
+
+To launch a SGX guest
+${QEMU} \
+   -cpu host,+sgx-provisionkey \
+   -object memory-backend-epc,id=mem1,size=64M,prealloc=on \
+   -object memory-backend-epc,id=mem2,size=28M \
+   -M sgx-epc.0.memdev=mem1,sgx-epc.1.memdev=mem2
+
+Utilizing SGX in the guest requires a kernel/OS with SGX support.
+
+The support can be determined in guest by:
+$ grep sgx /proc/cpuinfo
+
+Check the SGX epc info in the Guest:
+$ dmesg | grep sgx
+[    1.242142] sgx: EPC section 0x180000000-0x181bfffff
+[    1.242319] sgx: EPC section 0x181c00000-0x1837fffff
+
+References
+==========
+
+SGX Homepage:
+https://software.intel.com/sgx
+
+SGX SDK:
+https://github.com/intel/linux-sgx.git
+
+SGX SPEC:
+Intel SDM Volume 3
#52 lore raw
In reply to Yang Zhong
Quoted 87 lines · Yang Zhong

Since Sean Christopherson has left Intel and i am responsible for Qemu SGX upstream work. His @intel.com address will be bouncing and his new email( [email protected]) is also in CC lists.

This series is Qemu SGX virtualization implementation rebased on latest Qemu release. The numa support for SGX will be sent in another patchset once this basic SGX patchset are merged.

You can find Qemu repo here:

https://github.com/intel/qemu-sgx.git upstream

If you want to try SGX, you can directly install the linux release(at least 5.13.0-rc1+) since kvm SGX has been merged into linux release.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git

To simplify, you'd better install linux on host and guest, which can support SGX on host and guest kernel. And to me, use below reference command to boot SGX guest:

     #qemu-system-x86_64 \
         ...... \
         -cpu host,+sgx-provisionkey \
         -object memory-backend-epc,id=mem1,size=64M,prealloc=on \
         -object memory-backend-epc,id=mem2,size=28M \
         -M sgx-epc.0.memdev=mem1,sgx-epc.1.memdev=mem2

Overview
========

Intel Software Guard eXtensions (SGX) is a set of instructions and mechanisms for memory accesses in order to provide security accesses for sensitive applications and data. SGX allows an application to use it's pariticular address space as an *enclave*, which is a protected area provides confidentiality and integrity even in the presence of privileged malware. Accesses to the enclave memory area from any software not resident in the enclave are prevented, including those from privileged software.

SGX virtaulization
==================

The KVM SGX creates one new misc device, sgx_vepc, and Qemu will open '/dev/sgx_vepc' device node to mmap() host EPC memory to guest. The Qemu also adds 'sgx-epc' device to expose EPC sections to guest through CPUID and ACPI table. The Qemu SGX also supports multiple virtual EPC sections to guest, we just put them together physically contiguous for the sake of simplicity. The kernel SGX NUMA has been merged into Linux tip tree, we will support this function in the next phase.

Although the current host SGX subsystem can not support SGX2 feature, the KVM/Qemu implementation still expose this feature to guest. Guest SGX2 support doesn't have interaction with host kernel SGX driver, the SGX guest can normally use those new instructions.

As for SGX virtualization detailed infomation, please reference docs/intel-sgx.txt docuement(patch 33).

Changelog:
=========

(Changelog here is for global changes, please see each patch's changelog for changes made to specific patch.)

v3-->v4:
- Rebased the sgx patches into latest Qemu release.
- Moved sgx compound property setter/getter from MachineState to X86MachineState(Paolo).
- Re-defined struct SgxEPC, removed 'id' property and added struct SgxEPCList for sgx-epc.0.{memdev}(Paolo).
- Removed g_malloc0(), and changed the 'SGXEPCState *sgx_epc' to 'SGXEPCState sgx_epc' in struct PCMachineState(Paolo).
- Changed the SGX compound property cmdline from sgx-epc.{memdev}.0 to sgx-epc.0.{memdev}(Paolo).
- Removed the signature from the 'git format-patch' command(Jarkko).

v2-->v3:
- Rebased the sgx patches into latest Qemu release.
- Implemented the compound property for SGX, ref patch5, the command from '-sgx-epc' to '-M'(Paolo).
- Moved the sgx common code from sgx-epc.c to sgx.c. The sgx-epc.c is only responsible for virtual epc device.
- Removed the previous patch13(linux-headers: Add placeholder for KVM_CAP_SGX_ATTRIBUTE) because [email protected] updated Linux headers to 5.13-rc4 with commit 278f064e452.
- Updated the patch1 because ram_flags were changed by David Hildenbra.
- Added one patch24, which avoid reset operation caused by bios reset.
- Added one patch25, which make prealloc property consistent with Qemu cmdline during VM reset.

Quoted 7 lines

v1-->v2:
- Rebased the sgx patches into latest Qemu release.
- Unified the "share" and "protected" arguments with ram_flags in the memory_region_init_ram_from_fd()(Paolo).
- Added the new MemoryBackendEpcProperties and related documents(Eric Blake).
- Changed the KVM_CAP_SGX_ATTRIBUTE from 195 to 196(Kai).
- Changed the version and some grammar issues(Eric Blake).

Looks good, I will queue it for 6.2.

Thanks for your patience with the compound machine properties support.

Paolo

Quoted 87 lines

Sean Christopherson (21):
memory: Add RAM_PROTECTED flag to skip IOMMU mappings
hostmem: Add hostmem-epc as a backend for SGX EPC
i386: Add 'sgx-epc' device to expose EPC sections to guest
vl: Add sgx compound properties to expose SGX EPC sections to guest
i386: Add primary SGX CPUID and MSR defines
i386: Add SGX CPUID leaf FEAT_SGX_12_0_EAX
i386: Add SGX CPUID leaf FEAT_SGX_12_0_EBX
i386: Add SGX CPUID leaf FEAT_SGX_12_1_EAX
i386: Add get/set/migrate support for SGX_LEPUBKEYHASH MSRs
i386: Add feature control MSR dependency when SGX is enabled
i386: Update SGX CPUID info according to hardware/KVM/user input
i386: kvm: Add support for exposing PROVISIONKEY to guest
i386: Propagate SGX CPUID sub-leafs to KVM
Adjust min CPUID level to 0x12 when SGX is enabled hw/i386/fw_cfg: Set SGX bits in feature control fw_cfg accordingly hw/i386/pc: Account for SGX EPC sections when calculating device
memory
i386/pc: Add e820 entry for SGX EPC section(s)
i386: acpi: Add SGX EPC entry to ACPI tables
q35: Add support for SGX EPC
i440fx: Add support for SGX EPC
doc: Add the SGX doc

Yang Zhong (12):
qom: Add memory-backend-epc ObjectOptions support
hostmem-epc: Add the reset interface for EPC backend reset
sgx-epc: Add the reset interface for sgx-epc virt device
sgx-epc: Avoid bios reset during sgx epc initialization
hostmem-epc: Make prealloc consistent with qemu cmdline during reset
qmp: Add query-sgx command
hmp: Add 'info sgx' command
i386: Add sgx_get_info() interface
bitops: Support 32 and 64 bit mask macro
qmp: Add the qmp_query_sgx_capabilities()
Kconfig: Add CONFIG_SGX support
sgx-epc: Add the fill_device_info() callback support

  backends/hostmem-epc.c                   | 118 ++++++++++
  backends/meson.build                     |   1 +
  configs/devices/i386-softmmu/default.mak |   1 +
  docs/intel-sgx.txt                       | 167 ++++++++++++++
  hmp-commands-info.hx                     |  15 ++
  hw/i386/Kconfig                          |   5 +
  hw/i386/acpi-build.c                     |  22 ++
  hw/i386/fw_cfg.c                         |  10 +-
  hw/i386/meson.build                      |   2 +
  hw/i386/pc.c                             |  15 +-
  hw/i386/pc_piix.c                        |   4 +
  hw/i386/pc_q35.c                         |   3 +
  hw/i386/sgx-epc.c                        | 265 +++++++++++++++++++++++
  hw/i386/sgx-stub.c                       |  13 ++
  hw/i386/sgx.c                            | 170 +++++++++++++++
  hw/i386/x86.c                            |  29 +++
  hw/vfio/common.c                         |   1 +
  include/exec/memory.h                    |  15 +-
  include/hw/i386/pc.h                     |  10 +
  include/hw/i386/sgx-epc.h                |  68 ++++++
  include/hw/i386/x86.h                    |   1 +
  include/monitor/hmp.h                    |   1 +
  include/qemu/bitops.h                    |   7 +
  monitor/hmp-cmds.c                       |  32 +++
  monitor/qmp-cmds.c                       |  19 ++
  qapi/machine.json                        |  52 ++++-
  qapi/misc.json                           |  61 ++++++
  qapi/qom.json                            |  19 ++
  qemu-options.hx                          |  10 +-
  softmmu/memory.c                         |   5 +
  softmmu/physmem.c                        |   3 +-
  stubs/meson.build                        |   1 +
  stubs/sgx-stub.c                         |  12 +
  target/i386/cpu.c                        | 168 +++++++++++++-
  target/i386/cpu.h                        |  16 ++
  target/i386/kvm/kvm.c                    |  75 +++++++
  target/i386/kvm/kvm_i386.h               |   2 +
  target/i386/machine.c                    |  20 ++
  tests/qtest/qmp-cmd-test.c               |   2 +
  39 files changed, 1430 insertions(+), 10 deletions(-)

create mode 100644 backends/hostmem-epc.c
create mode 100644 docs/intel-sgx.txt
create mode 100644 hw/i386/sgx-epc.c
create mode 100644 hw/i386/sgx-stub.c
create mode 100644 hw/i386/sgx.c
create mode 100644 include/hw/i386/sgx-epc.h
create mode 100644 stubs/sgx-stub.c

author
#53 lore raw
Quoted 96 lines · Paolo Bonzini
Quoted 87 lines · Yang Zhong

Since Sean Christopherson has left Intel and i am responsible for Qemu SGX upstream work. His @intel.com address will be bouncing and his new email( [email protected]) is also in CC lists.

This series is Qemu SGX virtualization implementation rebased on latest Qemu release. The numa support for SGX will be sent in another patchset once this basic SGX patchset are merged.

You can find Qemu repo here:

https://github.com/intel/qemu-sgx.git upstream

If you want to try SGX, you can directly install the linux release(at least 5.13.0-rc1+) since kvm SGX has been merged into linux release.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git

To simplify, you'd better install linux on host and guest, which can support SGX on host and guest kernel. And to me, use below reference command to boot SGX guest:

    #qemu-system-x86_64 \
        ...... \
        -cpu host,+sgx-provisionkey \
        -object memory-backend-epc,id=mem1,size=64M,prealloc=on \
        -object memory-backend-epc,id=mem2,size=28M \
        -M sgx-epc.0.memdev=mem1,sgx-epc.1.memdev=mem2

Overview
========

Intel Software Guard eXtensions (SGX) is a set of instructions and mechanisms for memory accesses in order to provide security accesses for sensitive applications and data. SGX allows an application to use it's pariticular address space as an *enclave*, which is a protected area provides confidentiality and integrity even in the presence of privileged malware. Accesses to the enclave memory area from any software not resident in the enclave are prevented, including those from privileged software.

SGX virtaulization
==================

The KVM SGX creates one new misc device, sgx_vepc, and Qemu will open '/dev/sgx_vepc' device node to mmap() host EPC memory to guest. The Qemu also adds 'sgx-epc' device to expose EPC sections to guest through CPUID and ACPI table. The Qemu SGX also supports multiple virtual EPC sections to guest, we just put them together physically contiguous for the sake of simplicity. The kernel SGX NUMA has been merged into Linux tip tree, we will support this function in the next phase.

Although the current host SGX subsystem can not support SGX2 feature, the KVM/Qemu implementation still expose this feature to guest. Guest SGX2 support doesn't have interaction with host kernel SGX driver, the SGX guest can normally use those new instructions.

As for SGX virtualization detailed infomation, please reference docs/intel-sgx.txt docuement(patch 33).

Changelog:
=========

(Changelog here is for global changes, please see each patch's changelog for changes made to specific patch.)

v3-->v4:
- Rebased the sgx patches into latest Qemu release.
- Moved sgx compound property setter/getter from MachineState to X86MachineState(Paolo).
- Re-defined struct SgxEPC, removed 'id' property and added struct SgxEPCList for

     sgx-epc.0.{memdev}(Paolo).
   - Removed g_malloc0(), and changed the 'SGXEPCState *sgx_epc' to 'SGXEPCState sgx_epc'
     in struct PCMachineState(Paolo).
   - Changed the SGX compound property cmdline from sgx-epc.{memdev}.0 to
     sgx-epc.0.{memdev}(Paolo).

- Removed the signature from the 'git format-patch' command(Jarkko).

v2-->v3:
- Rebased the sgx patches into latest Qemu release.
- Implemented the compound property for SGX, ref patch5, the command from '-sgx-epc'

     to '-M'(Paolo).

- Moved the sgx common code from sgx-epc.c to sgx.c. The sgx-epc.c is only responsible for virtual epc device.
- Removed the previous patch13(linux-headers: Add placeholder for KVM_CAP_SGX_ATTRIBUTE) because [email protected] updated Linux headers to 5.13-rc4 with commit 278f064e452.
- Updated the patch1 because ram_flags were changed by David Hildenbra.
- Added one patch24, which avoid reset operation caused by bios reset.
- Added one patch25, which make prealloc property consistent with Qemu cmdline during VM reset.

Quoted 7 lines

v1-->v2:
- Rebased the sgx patches into latest Qemu release.
- Unified the "share" and "protected" arguments with ram_flags in the memory_region_init_ram_from_fd()(Paolo).
- Added the new MemoryBackendEpcProperties and related documents(Eric Blake).
- Changed the KVM_CAP_SGX_ATTRIBUTE from 195 to 196(Kai).
- Changed the version and some grammar issues(Eric Blake).

Quoted 4 lines

Looks good, I will queue it for 6.2.

Thanks for your patience with the compound machine properties support.

Paolo, thanks! Once this patchset is merged, i will send SGX NUMA patchset, thanks!

Regards,

Yang

Quoted 89 lines

Paolo

Quoted 87 lines

Sean Christopherson (21):
memory: Add RAM_PROTECTED flag to skip IOMMU mappings
hostmem: Add hostmem-epc as a backend for SGX EPC
i386: Add 'sgx-epc' device to expose EPC sections to guest
vl: Add sgx compound properties to expose SGX EPC sections to guest
i386: Add primary SGX CPUID and MSR defines
i386: Add SGX CPUID leaf FEAT_SGX_12_0_EAX
i386: Add SGX CPUID leaf FEAT_SGX_12_0_EBX
i386: Add SGX CPUID leaf FEAT_SGX_12_1_EAX
i386: Add get/set/migrate support for SGX_LEPUBKEYHASH MSRs
i386: Add feature control MSR dependency when SGX is enabled
i386: Update SGX CPUID info according to hardware/KVM/user input
i386: kvm: Add support for exposing PROVISIONKEY to guest
i386: Propagate SGX CPUID sub-leafs to KVM
Adjust min CPUID level to 0x12 when SGX is enabled hw/i386/fw_cfg: Set SGX bits in feature control fw_cfg accordingly hw/i386/pc: Account for SGX EPC sections when calculating device
memory
i386/pc: Add e820 entry for SGX EPC section(s)
i386: acpi: Add SGX EPC entry to ACPI tables
q35: Add support for SGX EPC
i440fx: Add support for SGX EPC
doc: Add the SGX doc

Yang Zhong (12):
qom: Add memory-backend-epc ObjectOptions support
hostmem-epc: Add the reset interface for EPC backend reset
sgx-epc: Add the reset interface for sgx-epc virt device
sgx-epc: Avoid bios reset during sgx epc initialization
hostmem-epc: Make prealloc consistent with qemu cmdline during reset
qmp: Add query-sgx command
hmp: Add 'info sgx' command
i386: Add sgx_get_info() interface
bitops: Support 32 and 64 bit mask macro
qmp: Add the qmp_query_sgx_capabilities()
Kconfig: Add CONFIG_SGX support
sgx-epc: Add the fill_device_info() callback support

 backends/hostmem-epc.c                   | 118 ++++++++++
 backends/meson.build                     |   1 +
 configs/devices/i386-softmmu/default.mak |   1 +
 docs/intel-sgx.txt                       | 167 ++++++++++++++
 hmp-commands-info.hx                     |  15 ++
 hw/i386/Kconfig                          |   5 +
 hw/i386/acpi-build.c                     |  22 ++
 hw/i386/fw_cfg.c                         |  10 +-
 hw/i386/meson.build                      |   2 +
 hw/i386/pc.c                             |  15 +-
 hw/i386/pc_piix.c                        |   4 +
 hw/i386/pc_q35.c                         |   3 +
 hw/i386/sgx-epc.c                        | 265 +++++++++++++++++++++++
 hw/i386/sgx-stub.c                       |  13 ++
 hw/i386/sgx.c                            | 170 +++++++++++++++
 hw/i386/x86.c                            |  29 +++
 hw/vfio/common.c                         |   1 +
 include/exec/memory.h                    |  15 +-
 include/hw/i386/pc.h                     |  10 +
 include/hw/i386/sgx-epc.h                |  68 ++++++
 include/hw/i386/x86.h                    |   1 +
 include/monitor/hmp.h                    |   1 +
 include/qemu/bitops.h                    |   7 +
 monitor/hmp-cmds.c                       |  32 +++
 monitor/qmp-cmds.c                       |  19 ++
 qapi/machine.json                        |  52 ++++-
 qapi/misc.json                           |  61 ++++++
 qapi/qom.json                            |  19 ++
 qemu-options.hx                          |  10 +-
 softmmu/memory.c                         |   5 +
 softmmu/physmem.c                        |   3 +-
 stubs/meson.build                        |   1 +
 stubs/sgx-stub.c                         |  12 +
 target/i386/cpu.c                        | 168 +++++++++++++-
 target/i386/cpu.h                        |  16 ++
 target/i386/kvm/kvm.c                    |  75 +++++++
 target/i386/kvm/kvm_i386.h               |   2 +
 target/i386/machine.c                    |  20 ++
 tests/qtest/qmp-cmd-test.c               |   2 +
 39 files changed, 1430 insertions(+), 10 deletions(-)
 create mode 100644 backends/hostmem-epc.c
 create mode 100644 docs/intel-sgx.txt
 create mode 100644 hw/i386/sgx-epc.c
 create mode 100644 hw/i386/sgx-stub.c
 create mode 100644 hw/i386/sgx.c
 create mode 100644 include/hw/i386/sgx-epc.h
 create mode 100644 stubs/sgx-stub.c
#54 lore raw
In reply to Yang Zhong

Hi,

the monitor patches did not pass the test-hmp qtest, and also they should be in target/i386/monitor.c (see other commands that were implemented there for SEV). However, I've sent a pull request with the rest.

Thanks,

Paolo

Quoted 182 lines · Yang Zhong

Since Sean Christopherson has left Intel and i am responsible for Qemu SGX upstream work. His @intel.com address will be bouncing and his new email( [email protected]) is also in CC lists.

This series is Qemu SGX virtualization implementation rebased on latest Qemu release. The numa support for SGX will be sent in another patchset once this basic SGX patchset are merged.

You can find Qemu repo here:

https://github.com/intel/qemu-sgx.git upstream

If you want to try SGX, you can directly install the linux release(at least 5.13.0-rc1+) since kvm SGX has been merged into linux release.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git

To simplify, you'd better install linux on host and guest, which can support SGX on host and guest kernel. And to me, use below reference command to boot SGX guest:

    #qemu-system-x86_64 \
        ...... \
        -cpu host,+sgx-provisionkey \
        -object memory-backend-epc,id=mem1,size=64M,prealloc=on \
        -object memory-backend-epc,id=mem2,size=28M \
        -M sgx-epc.0.memdev=mem1,sgx-epc.1.memdev=mem2

Overview
========

Intel Software Guard eXtensions (SGX) is a set of instructions and mechanisms for memory accesses in order to provide security accesses for sensitive applications and data. SGX allows an application to use it's pariticular address space as an *enclave*, which is a protected area provides confidentiality and integrity even in the presence of privileged malware. Accesses to the enclave memory area from any software not resident in the enclave are prevented, including those from privileged software.

SGX virtaulization
==================

The KVM SGX creates one new misc device, sgx_vepc, and Qemu will open '/dev/sgx_vepc' device node to mmap() host EPC memory to guest. The Qemu also adds 'sgx-epc' device to expose EPC sections to guest through CPUID and ACPI table. The Qemu SGX also supports multiple virtual EPC sections to guest, we just put them together physically contiguous for the sake of simplicity. The kernel SGX NUMA has been merged into Linux tip tree, we will support this function in the next phase.

Although the current host SGX subsystem can not support SGX2 feature, the KVM/Qemu implementation still expose this feature to guest. Guest SGX2 support doesn't have interaction with host kernel SGX driver, the SGX guest can normally use those new instructions.

As for SGX virtualization detailed infomation, please reference docs/intel-sgx.txt docuement(patch 33).

Changelog:
=========

(Changelog here is for global changes, please see each patch's changelog for changes made to specific patch.)

v3-->v4:
- Rebased the sgx patches into latest Qemu release.
- Moved sgx compound property setter/getter from MachineState to X86MachineState(Paolo).
- Re-defined struct SgxEPC, removed 'id' property and added struct SgxEPCList for

     sgx-epc.0.{memdev}(Paolo).
   - Removed g_malloc0(), and changed the 'SGXEPCState *sgx_epc' to 'SGXEPCState sgx_epc'
     in struct PCMachineState(Paolo).
   - Changed the SGX compound property cmdline from sgx-epc.{memdev}.0 to
     sgx-epc.0.{memdev}(Paolo).

- Removed the signature from the 'git format-patch' command(Jarkko).

v2-->v3:
- Rebased the sgx patches into latest Qemu release.
- Implemented the compound property for SGX, ref patch5, the command from '-sgx-epc'

     to '-M'(Paolo).

- Moved the sgx common code from sgx-epc.c to sgx.c. The sgx-epc.c is only responsible for virtual epc device.
- Removed the previous patch13(linux-headers: Add placeholder for KVM_CAP_SGX_ATTRIBUTE) because [email protected] updated Linux headers to 5.13-rc4 with commit 278f064e452.
- Updated the patch1 because ram_flags were changed by David Hildenbra.
- Added one patch24, which avoid reset operation caused by bios reset.
- Added one patch25, which make prealloc property consistent with Qemu cmdline during VM reset.

v1-->v2:
- Rebased the sgx patches into latest Qemu release.
- Unified the "share" and "protected" arguments with ram_flags in the memory_region_init_ram_from_fd()(Paolo).
- Added the new MemoryBackendEpcProperties and related documents(Eric Blake).
- Changed the KVM_CAP_SGX_ATTRIBUTE from 195 to 196(Kai).
- Changed the version and some grammar issues(Eric Blake).

Sean Christopherson (21):
memory: Add RAM_PROTECTED flag to skip IOMMU mappings
hostmem: Add hostmem-epc as a backend for SGX EPC
i386: Add 'sgx-epc' device to expose EPC sections to guest
vl: Add sgx compound properties to expose SGX EPC sections to guest
i386: Add primary SGX CPUID and MSR defines
i386: Add SGX CPUID leaf FEAT_SGX_12_0_EAX
i386: Add SGX CPUID leaf FEAT_SGX_12_0_EBX
i386: Add SGX CPUID leaf FEAT_SGX_12_1_EAX
i386: Add get/set/migrate support for SGX_LEPUBKEYHASH MSRs
i386: Add feature control MSR dependency when SGX is enabled
i386: Update SGX CPUID info according to hardware/KVM/user input
i386: kvm: Add support for exposing PROVISIONKEY to guest
i386: Propagate SGX CPUID sub-leafs to KVM
Adjust min CPUID level to 0x12 when SGX is enabled hw/i386/fw_cfg: Set SGX bits in feature control fw_cfg accordingly hw/i386/pc: Account for SGX EPC sections when calculating device
memory
i386/pc: Add e820 entry for SGX EPC section(s)
i386: acpi: Add SGX EPC entry to ACPI tables
q35: Add support for SGX EPC
i440fx: Add support for SGX EPC
doc: Add the SGX doc

Yang Zhong (12):
qom: Add memory-backend-epc ObjectOptions support
hostmem-epc: Add the reset interface for EPC backend reset
sgx-epc: Add the reset interface for sgx-epc virt device
sgx-epc: Avoid bios reset during sgx epc initialization
hostmem-epc: Make prealloc consistent with qemu cmdline during reset
qmp: Add query-sgx command
hmp: Add 'info sgx' command
i386: Add sgx_get_info() interface
bitops: Support 32 and 64 bit mask macro
qmp: Add the qmp_query_sgx_capabilities()
Kconfig: Add CONFIG_SGX support
sgx-epc: Add the fill_device_info() callback support

 backends/hostmem-epc.c                   | 118 ++++++++++
 backends/meson.build                     |   1 +
 configs/devices/i386-softmmu/default.mak |   1 +
 docs/intel-sgx.txt                       | 167 ++++++++++++++
 hmp-commands-info.hx                     |  15 ++
 hw/i386/Kconfig                          |   5 +
 hw/i386/acpi-build.c                     |  22 ++
 hw/i386/fw_cfg.c                         |  10 +-
 hw/i386/meson.build                      |   2 +
 hw/i386/pc.c                             |  15 +-
 hw/i386/pc_piix.c                        |   4 +
 hw/i386/pc_q35.c                         |   3 +
 hw/i386/sgx-epc.c                        | 265 +++++++++++++++++++++++
 hw/i386/sgx-stub.c                       |  13 ++
 hw/i386/sgx.c                            | 170 +++++++++++++++
 hw/i386/x86.c                            |  29 +++
 hw/vfio/common.c                         |   1 +
 include/exec/memory.h                    |  15 +-
 include/hw/i386/pc.h                     |  10 +
 include/hw/i386/sgx-epc.h                |  68 ++++++
 include/hw/i386/x86.h                    |   1 +
 include/monitor/hmp.h                    |   1 +
 include/qemu/bitops.h                    |   7 +
 monitor/hmp-cmds.c                       |  32 +++
 monitor/qmp-cmds.c                       |  19 ++
 qapi/machine.json                        |  52 ++++-
 qapi/misc.json                           |  61 ++++++
 qapi/qom.json                            |  19 ++
 qemu-options.hx                          |  10 +-
 softmmu/memory.c                         |   5 +
 softmmu/physmem.c                        |   3 +-
 stubs/meson.build                        |   1 +
 stubs/sgx-stub.c                         |  12 +
 target/i386/cpu.c                        | 168 +++++++++++++-
 target/i386/cpu.h                        |  16 ++
 target/i386/kvm/kvm.c                    |  75 +++++++
 target/i386/kvm/kvm_i386.h               |   2 +
 target/i386/machine.c                    |  20 ++
 tests/qtest/qmp-cmd-test.c               |   2 +
 39 files changed, 1430 insertions(+), 10 deletions(-)
 create mode 100644 backends/hostmem-epc.c
 create mode 100644 docs/intel-sgx.txt
 create mode 100644 hw/i386/sgx-epc.c
 create mode 100644 hw/i386/sgx-stub.c
 create mode 100644 hw/i386/sgx.c
 create mode 100644 include/hw/i386/sgx-epc.h
 create mode 100644 stubs/sgx-stub.c
author
#55 lore raw
Quoted 7 lines · Paolo Bonzini

Hi,

the monitor patches did not pass the test-hmp qtest, and also they should be in target/i386/monitor.c (see other commands that were implemented there for SEV). However, I've sent a pull request with the rest.

Paolo, thanks for pulling those patches! In fact, the first POC implemented the monitor to target/i386/monitor.c, but in that time, there were lots of 'pragma GCC poison' errors during the build, so i had to move to common monitor. Let me check this again, thanks!

Yang

Quoted 187 lines

Thanks,

Paolo

Quoted 182 lines · Yang Zhong

Since Sean Christopherson has left Intel and i am responsible for Qemu SGX upstream work. His @intel.com address will be bouncing and his new email( [email protected]) is also in CC lists.

This series is Qemu SGX virtualization implementation rebased on latest Qemu release. The numa support for SGX will be sent in another patchset once this basic SGX patchset are merged.

You can find Qemu repo here:

https://github.com/intel/qemu-sgx.git upstream

If you want to try SGX, you can directly install the linux release(at least 5.13.0-rc1+) since kvm SGX has been merged into linux release.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git

To simplify, you'd better install linux on host and guest, which can support SGX on host and guest kernel. And to me, use below reference command to boot SGX guest:

    #qemu-system-x86_64 \
        ...... \
        -cpu host,+sgx-provisionkey \
        -object memory-backend-epc,id=mem1,size=64M,prealloc=on \
        -object memory-backend-epc,id=mem2,size=28M \
        -M sgx-epc.0.memdev=mem1,sgx-epc.1.memdev=mem2

Overview
========

Intel Software Guard eXtensions (SGX) is a set of instructions and mechanisms for memory accesses in order to provide security accesses for sensitive applications and data. SGX allows an application to use it's pariticular address space as an *enclave*, which is a protected area provides confidentiality and integrity even in the presence of privileged malware. Accesses to the enclave memory area from any software not resident in the enclave are prevented, including those from privileged software.

SGX virtaulization
==================

The KVM SGX creates one new misc device, sgx_vepc, and Qemu will open '/dev/sgx_vepc' device node to mmap() host EPC memory to guest. The Qemu also adds 'sgx-epc' device to expose EPC sections to guest through CPUID and ACPI table. The Qemu SGX also supports multiple virtual EPC sections to guest, we just put them together physically contiguous for the sake of simplicity. The kernel SGX NUMA has been merged into Linux tip tree, we will support this function in the next phase.

Although the current host SGX subsystem can not support SGX2 feature, the KVM/Qemu implementation still expose this feature to guest. Guest SGX2 support doesn't have interaction with host kernel SGX driver, the SGX guest can normally use those new instructions.

As for SGX virtualization detailed infomation, please reference docs/intel-sgx.txt docuement(patch 33).

Changelog:
=========

(Changelog here is for global changes, please see each patch's changelog for changes made to specific patch.)

v3-->v4:
- Rebased the sgx patches into latest Qemu release.
- Moved sgx compound property setter/getter from MachineState to X86MachineState(Paolo).
- Re-defined struct SgxEPC, removed 'id' property and added struct SgxEPCList for

     sgx-epc.0.{memdev}(Paolo).
   - Removed g_malloc0(), and changed the 'SGXEPCState *sgx_epc' to 'SGXEPCState sgx_epc'
     in struct PCMachineState(Paolo).
   - Changed the SGX compound property cmdline from sgx-epc.{memdev}.0 to
     sgx-epc.0.{memdev}(Paolo).

- Removed the signature from the 'git format-patch' command(Jarkko).

v2-->v3:
- Rebased the sgx patches into latest Qemu release.
- Implemented the compound property for SGX, ref patch5, the command from '-sgx-epc'

     to '-M'(Paolo).

- Moved the sgx common code from sgx-epc.c to sgx.c. The sgx-epc.c is only responsible for virtual epc device.
- Removed the previous patch13(linux-headers: Add placeholder for KVM_CAP_SGX_ATTRIBUTE) because [email protected] updated Linux headers to 5.13-rc4 with commit 278f064e452.
- Updated the patch1 because ram_flags were changed by David Hildenbra.
- Added one patch24, which avoid reset operation caused by bios reset.
- Added one patch25, which make prealloc property consistent with Qemu cmdline during VM reset.

v1-->v2:
- Rebased the sgx patches into latest Qemu release.
- Unified the "share" and "protected" arguments with ram_flags in the memory_region_init_ram_from_fd()(Paolo).
- Added the new MemoryBackendEpcProperties and related documents(Eric Blake).
- Changed the KVM_CAP_SGX_ATTRIBUTE from 195 to 196(Kai).
- Changed the version and some grammar issues(Eric Blake).

Sean Christopherson (21):
memory: Add RAM_PROTECTED flag to skip IOMMU mappings
hostmem: Add hostmem-epc as a backend for SGX EPC
i386: Add 'sgx-epc' device to expose EPC sections to guest
vl: Add sgx compound properties to expose SGX EPC sections to guest
i386: Add primary SGX CPUID and MSR defines
i386: Add SGX CPUID leaf FEAT_SGX_12_0_EAX
i386: Add SGX CPUID leaf FEAT_SGX_12_0_EBX
i386: Add SGX CPUID leaf FEAT_SGX_12_1_EAX
i386: Add get/set/migrate support for SGX_LEPUBKEYHASH MSRs
i386: Add feature control MSR dependency when SGX is enabled
i386: Update SGX CPUID info according to hardware/KVM/user input
i386: kvm: Add support for exposing PROVISIONKEY to guest
i386: Propagate SGX CPUID sub-leafs to KVM
Adjust min CPUID level to 0x12 when SGX is enabled hw/i386/fw_cfg: Set SGX bits in feature control fw_cfg accordingly hw/i386/pc: Account for SGX EPC sections when calculating device
memory
i386/pc: Add e820 entry for SGX EPC section(s)
i386: acpi: Add SGX EPC entry to ACPI tables
q35: Add support for SGX EPC
i440fx: Add support for SGX EPC
doc: Add the SGX doc

Yang Zhong (12):
qom: Add memory-backend-epc ObjectOptions support
hostmem-epc: Add the reset interface for EPC backend reset
sgx-epc: Add the reset interface for sgx-epc virt device
sgx-epc: Avoid bios reset during sgx epc initialization
hostmem-epc: Make prealloc consistent with qemu cmdline during reset
qmp: Add query-sgx command
hmp: Add 'info sgx' command
i386: Add sgx_get_info() interface
bitops: Support 32 and 64 bit mask macro
qmp: Add the qmp_query_sgx_capabilities()
Kconfig: Add CONFIG_SGX support
sgx-epc: Add the fill_device_info() callback support

 backends/hostmem-epc.c                   | 118 ++++++++++
 backends/meson.build                     |   1 +
 configs/devices/i386-softmmu/default.mak |   1 +
 docs/intel-sgx.txt                       | 167 ++++++++++++++
 hmp-commands-info.hx                     |  15 ++
 hw/i386/Kconfig                          |   5 +
 hw/i386/acpi-build.c                     |  22 ++
 hw/i386/fw_cfg.c                         |  10 +-
 hw/i386/meson.build                      |   2 +
 hw/i386/pc.c                             |  15 +-
 hw/i386/pc_piix.c                        |   4 +
 hw/i386/pc_q35.c                         |   3 +
 hw/i386/sgx-epc.c                        | 265 +++++++++++++++++++++++
 hw/i386/sgx-stub.c                       |  13 ++
 hw/i386/sgx.c                            | 170 +++++++++++++++
 hw/i386/x86.c                            |  29 +++
 hw/vfio/common.c                         |   1 +
 include/exec/memory.h                    |  15 +-
 include/hw/i386/pc.h                     |  10 +
 include/hw/i386/sgx-epc.h                |  68 ++++++
 include/hw/i386/x86.h                    |   1 +
 include/monitor/hmp.h                    |   1 +
 include/qemu/bitops.h                    |   7 +
 monitor/hmp-cmds.c                       |  32 +++
 monitor/qmp-cmds.c                       |  19 ++
 qapi/machine.json                        |  52 ++++-
 qapi/misc.json                           |  61 ++++++
 qapi/qom.json                            |  19 ++
 qemu-options.hx                          |  10 +-
 softmmu/memory.c                         |   5 +
 softmmu/physmem.c                        |   3 +-
 stubs/meson.build                        |   1 +
 stubs/sgx-stub.c                         |  12 +
 target/i386/cpu.c                        | 168 +++++++++++++-
 target/i386/cpu.h                        |  16 ++
 target/i386/kvm/kvm.c                    |  75 +++++++
 target/i386/kvm/kvm_i386.h               |   2 +
 target/i386/machine.c                    |  20 ++
 tests/qtest/qmp-cmd-test.c               |   2 +
 39 files changed, 1430 insertions(+), 10 deletions(-)
 create mode 100644 backends/hostmem-epc.c
 create mode 100644 docs/intel-sgx.txt
 create mode 100644 hw/i386/sgx-epc.c
 create mode 100644 hw/i386/sgx-stub.c
 create mode 100644 hw/i386/sgx.c
 create mode 100644 include/hw/i386/sgx-epc.h
 create mode 100644 stubs/sgx-stub.c
author
#56 lore raw
In reply to Paolo Bonzini
Quoted 7 lines · Paolo Bonzini

Hi,

the monitor patches did not pass the test-hmp qtest, and also they should be in target/i386/monitor.c (see other commands that were implemented there for SEV). However, I've sent a pull request with the rest.

Paolo, I have moved hmp and qmp codes to target/i386/monitor.c and also fixed the issue with test tool(tests/qtest/test-hmp). This issue is caused by 'machine none' test in the hmp, the previous patches only covered qmp test in 'machine none' with Libvirt.

So, the next issue:
1) re-send all sgx basic patches(including monitors patches) to you?
2) only send monitor patches in the next phase when the basic sgx patches are merged?

Regards,

Yang

Quoted 187 lines

Thanks,

Paolo

Quoted 182 lines · Yang Zhong

Since Sean Christopherson has left Intel and i am responsible for Qemu SGX upstream work. His @intel.com address will be bouncing and his new email( [email protected]) is also in CC lists.

This series is Qemu SGX virtualization implementation rebased on latest Qemu release. The numa support for SGX will be sent in another patchset once this basic SGX patchset are merged.

You can find Qemu repo here:

https://github.com/intel/qemu-sgx.git upstream

If you want to try SGX, you can directly install the linux release(at least 5.13.0-rc1+) since kvm SGX has been merged into linux release.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git

To simplify, you'd better install linux on host and guest, which can support SGX on host and guest kernel. And to me, use below reference command to boot SGX guest:

    #qemu-system-x86_64 \
        ...... \
        -cpu host,+sgx-provisionkey \
        -object memory-backend-epc,id=mem1,size=64M,prealloc=on \
        -object memory-backend-epc,id=mem2,size=28M \
        -M sgx-epc.0.memdev=mem1,sgx-epc.1.memdev=mem2

Overview
========

Intel Software Guard eXtensions (SGX) is a set of instructions and mechanisms for memory accesses in order to provide security accesses for sensitive applications and data. SGX allows an application to use it's pariticular address space as an *enclave*, which is a protected area provides confidentiality and integrity even in the presence of privileged malware. Accesses to the enclave memory area from any software not resident in the enclave are prevented, including those from privileged software.

SGX virtaulization
==================

The KVM SGX creates one new misc device, sgx_vepc, and Qemu will open '/dev/sgx_vepc' device node to mmap() host EPC memory to guest. The Qemu also adds 'sgx-epc' device to expose EPC sections to guest through CPUID and ACPI table. The Qemu SGX also supports multiple virtual EPC sections to guest, we just put them together physically contiguous for the sake of simplicity. The kernel SGX NUMA has been merged into Linux tip tree, we will support this function in the next phase.

Although the current host SGX subsystem can not support SGX2 feature, the KVM/Qemu implementation still expose this feature to guest. Guest SGX2 support doesn't have interaction with host kernel SGX driver, the SGX guest can normally use those new instructions.

As for SGX virtualization detailed infomation, please reference docs/intel-sgx.txt docuement(patch 33).

Changelog:
=========

(Changelog here is for global changes, please see each patch's changelog for changes made to specific patch.)

v3-->v4:
- Rebased the sgx patches into latest Qemu release.
- Moved sgx compound property setter/getter from MachineState to X86MachineState(Paolo).
- Re-defined struct SgxEPC, removed 'id' property and added struct SgxEPCList for

     sgx-epc.0.{memdev}(Paolo).
   - Removed g_malloc0(), and changed the 'SGXEPCState *sgx_epc' to 'SGXEPCState sgx_epc'
     in struct PCMachineState(Paolo).
   - Changed the SGX compound property cmdline from sgx-epc.{memdev}.0 to
     sgx-epc.0.{memdev}(Paolo).

- Removed the signature from the 'git format-patch' command(Jarkko).

v2-->v3:
- Rebased the sgx patches into latest Qemu release.
- Implemented the compound property for SGX, ref patch5, the command from '-sgx-epc'

     to '-M'(Paolo).

- Moved the sgx common code from sgx-epc.c to sgx.c. The sgx-epc.c is only responsible for virtual epc device.
- Removed the previous patch13(linux-headers: Add placeholder for KVM_CAP_SGX_ATTRIBUTE) because [email protected] updated Linux headers to 5.13-rc4 with commit 278f064e452.
- Updated the patch1 because ram_flags were changed by David Hildenbra.
- Added one patch24, which avoid reset operation caused by bios reset.
- Added one patch25, which make prealloc property consistent with Qemu cmdline during VM reset.

v1-->v2:
- Rebased the sgx patches into latest Qemu release.
- Unified the "share" and "protected" arguments with ram_flags in the memory_region_init_ram_from_fd()(Paolo).
- Added the new MemoryBackendEpcProperties and related documents(Eric Blake).
- Changed the KVM_CAP_SGX_ATTRIBUTE from 195 to 196(Kai).
- Changed the version and some grammar issues(Eric Blake).

Sean Christopherson (21):
memory: Add RAM_PROTECTED flag to skip IOMMU mappings
hostmem: Add hostmem-epc as a backend for SGX EPC
i386: Add 'sgx-epc' device to expose EPC sections to guest
vl: Add sgx compound properties to expose SGX EPC sections to guest
i386: Add primary SGX CPUID and MSR defines
i386: Add SGX CPUID leaf FEAT_SGX_12_0_EAX
i386: Add SGX CPUID leaf FEAT_SGX_12_0_EBX
i386: Add SGX CPUID leaf FEAT_SGX_12_1_EAX
i386: Add get/set/migrate support for SGX_LEPUBKEYHASH MSRs
i386: Add feature control MSR dependency when SGX is enabled
i386: Update SGX CPUID info according to hardware/KVM/user input
i386: kvm: Add support for exposing PROVISIONKEY to guest
i386: Propagate SGX CPUID sub-leafs to KVM
Adjust min CPUID level to 0x12 when SGX is enabled hw/i386/fw_cfg: Set SGX bits in feature control fw_cfg accordingly hw/i386/pc: Account for SGX EPC sections when calculating device
memory
i386/pc: Add e820 entry for SGX EPC section(s)
i386: acpi: Add SGX EPC entry to ACPI tables
q35: Add support for SGX EPC
i440fx: Add support for SGX EPC
doc: Add the SGX doc

Yang Zhong (12):
qom: Add memory-backend-epc ObjectOptions support
hostmem-epc: Add the reset interface for EPC backend reset
sgx-epc: Add the reset interface for sgx-epc virt device
sgx-epc: Avoid bios reset during sgx epc initialization
hostmem-epc: Make prealloc consistent with qemu cmdline during reset
qmp: Add query-sgx command
hmp: Add 'info sgx' command
i386: Add sgx_get_info() interface
bitops: Support 32 and 64 bit mask macro
qmp: Add the qmp_query_sgx_capabilities()
Kconfig: Add CONFIG_SGX support
sgx-epc: Add the fill_device_info() callback support

 backends/hostmem-epc.c                   | 118 ++++++++++
 backends/meson.build                     |   1 +
 configs/devices/i386-softmmu/default.mak |   1 +
 docs/intel-sgx.txt                       | 167 ++++++++++++++
 hmp-commands-info.hx                     |  15 ++
 hw/i386/Kconfig                          |   5 +
 hw/i386/acpi-build.c                     |  22 ++
 hw/i386/fw_cfg.c                         |  10 +-
 hw/i386/meson.build                      |   2 +
 hw/i386/pc.c                             |  15 +-
 hw/i386/pc_piix.c                        |   4 +
 hw/i386/pc_q35.c                         |   3 +
 hw/i386/sgx-epc.c                        | 265 +++++++++++++++++++++++
 hw/i386/sgx-stub.c                       |  13 ++
 hw/i386/sgx.c                            | 170 +++++++++++++++
 hw/i386/x86.c                            |  29 +++
 hw/vfio/common.c                         |   1 +
 include/exec/memory.h                    |  15 +-
 include/hw/i386/pc.h                     |  10 +
 include/hw/i386/sgx-epc.h                |  68 ++++++
 include/hw/i386/x86.h                    |   1 +
 include/monitor/hmp.h                    |   1 +
 include/qemu/bitops.h                    |   7 +
 monitor/hmp-cmds.c                       |  32 +++
 monitor/qmp-cmds.c                       |  19 ++
 qapi/machine.json                        |  52 ++++-
 qapi/misc.json                           |  61 ++++++
 qapi/qom.json                            |  19 ++
 qemu-options.hx                          |  10 +-
 softmmu/memory.c                         |   5 +
 softmmu/physmem.c                        |   3 +-
 stubs/meson.build                        |   1 +
 stubs/sgx-stub.c                         |  12 +
 target/i386/cpu.c                        | 168 +++++++++++++-
 target/i386/cpu.h                        |  16 ++
 target/i386/kvm/kvm.c                    |  75 +++++++
 target/i386/kvm/kvm_i386.h               |   2 +
 target/i386/machine.c                    |  20 ++
 tests/qtest/qmp-cmd-test.c               |   2 +
 39 files changed, 1430 insertions(+), 10 deletions(-)
 create mode 100644 backends/hostmem-epc.c
 create mode 100644 docs/intel-sgx.txt
 create mode 100644 hw/i386/sgx-epc.c
 create mode 100644 hw/i386/sgx-stub.c
 create mode 100644 hw/i386/sgx.c
 create mode 100644 include/hw/i386/sgx-epc.h
 create mode 100644 stubs/sgx-stub.c
#57 lore raw
Quoted 14 lines · Yang Zhong
Quoted 7 lines · Paolo Bonzini

Hi,

the monitor patches did not pass the test-hmp qtest, and also they should be in target/i386/monitor.c (see other commands that were implemented there for SEV). However, I've sent a pull request with the rest.

Paolo, I have moved hmp and qmp codes to target/i386/monitor.c and also fixed the issue with test tool(tests/qtest/test-hmp). This issue is caused by 'machine none' test in the hmp, the previous patches only covered qmp test in 'machine none' with Libvirt.

Quoted 41 lines

So, the next issue:
1) re-send all sgx basic patches(including monitors patches) to you?
2) only send monitor patches in the next phase when the basic sgx patches are merged?

Regards,

Yang

Quoted 31 lines

Thanks,

Paolo

Quoted 26 lines · Yang Zhong

Since Sean Christopherson has left Intel and i am responsible for Qemu SGX upstream work. His @intel.com address will be bouncing and his new email( [email protected]) is also in CC lists.

This series is Qemu SGX virtualization implementation rebased on latest Qemu release. The numa support for SGX will be sent in another patchset once this basic SGX patchset are merged.

You can find Qemu repo here:

https://github.com/intel/qemu-sgx.git upstream

If you want to try SGX, you can directly install the linux release(at least 5.13.0-rc1+) since kvm SGX has been merged into linux release.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git

To simplify, you'd better install linux on host and guest, which can support SGX on host and guest kernel. And to me, use below reference command to boot SGX guest:

    #qemu-system-x86_64 \
        ...... \
        -cpu host,+sgx-provisionkey \
        -object memory-backend-epc,id=mem1,size=64M,prealloc=on \
        -object memory-backend-epc,id=mem2,size=28M \

I would call these just "memory-backend-sgx".

Quoted 13 lines
Quoted 13 lines
Quoted 13 lines
        -M sgx-epc.0.memdev=mem1,sgx-epc.1.memdev=mem2

Overview
========

Intel Software Guard eXtensions (SGX) is a set of instructions and mechanisms for memory accesses in order to provide security accesses for sensitive applications and data. SGX allows an application to use it's pariticular address space as an *enclave*, which is a protected area provides confidentiality and integrity even in the presence of privileged malware. Accesses to the enclave memory area from any software not resident in the enclave are prevented, including those from privileged software.

Quoted 1 line
Quoted 1 line
Quoted 1 line

SGX virtaulization

          ~~~~~~~~~~~~~~
	  virtualization

I'm using these patches now instead of "real" hardwave. It seems that 9th gen cores are now Linux compatible SGX. Maybe some ucode update has changed this because I don't recall this working before.

/Jarkko

#58 lore raw
In reply to Yang Zhong
Quoted 10 lines · Yang Zhong

Paolo, I have moved hmp and qmp codes to target/i386/monitor.c and also fixed the issue with test tool(tests/qtest/test-hmp). This issue is caused by 'machine none' test in the hmp, the previous patches only covered qmp test in 'machine none' with Libvirt.

So, the next issue:
1) re-send all sgx basic patches(including monitors patches) to you?
2) only send monitor patches in the next phase when the basic sgx patches are merged?

You can post them now, including

Based-on: <[email protected]>

in the cover letter.

Paolo

#59 lore raw
In reply to Yang Zhong
Quoted 19 lines · Yang Zhong

This series is Qemu SGX virtualization implementation rebased on latest Qemu release. The numa support for SGX will be sent in another patchset once this basic SGX patchset are merged.

 backends/hostmem-epc.c                   | 118 ++++++++++
 backends/meson.build                     |   1 +
 configs/devices/i386-softmmu/default.mak |   1 +
 docs/intel-sgx.txt                       | 167 ++++++++++++++
 hmp-commands-info.hx                     |  15 ++
 hw/i386/Kconfig                          |   5 +
 hw/i386/acpi-build.c                     |  22 ++
 hw/i386/fw_cfg.c                         |  10 +-
 hw/i386/meson.build                      |   2 +
 hw/i386/pc.c                             |  15 +-
 hw/i386/pc_piix.c                        |   4 +
 hw/i386/pc_q35.c                         |   3 +
 hw/i386/sgx-epc.c                        | 265 +++++++++++++++++++++++
 hw/i386/sgx-stub.c                       |  13 ++
 hw/i386/sgx.c                            | 170 +++++++++++++++
Quoted 8 lines
 hw/i386/x86.c                            |  29 +++
 hw/vfio/common.c                         |   1 +
 include/exec/memory.h                    |  15 +-
 include/hw/i386/pc.h                     |  10 +
 include/hw/i386/sgx-epc.h                |  68 ++++++
 include/hw/i386/x86.h                    |   1 +
 include/monitor/hmp.h                    |   1 +
 include/qemu/bitops.h                    |   7 +

Consider using scripts/git.orderfile to ease reviewers workflow.

author
#60 lore raw
Quoted 30 lines · Philippe Mathieu-Daudé
Quoted 19 lines · Yang Zhong

This series is Qemu SGX virtualization implementation rebased on latest Qemu release. The numa support for SGX will be sent in another patchset once this basic SGX patchset are merged.

 backends/hostmem-epc.c                   | 118 ++++++++++
 backends/meson.build                     |   1 +
 configs/devices/i386-softmmu/default.mak |   1 +
 docs/intel-sgx.txt                       | 167 ++++++++++++++
 hmp-commands-info.hx                     |  15 ++
 hw/i386/Kconfig                          |   5 +
 hw/i386/acpi-build.c                     |  22 ++
 hw/i386/fw_cfg.c                         |  10 +-
 hw/i386/meson.build                      |   2 +
 hw/i386/pc.c                             |  15 +-
 hw/i386/pc_piix.c                        |   4 +
 hw/i386/pc_q35.c                         |   3 +
 hw/i386/sgx-epc.c                        | 265 +++++++++++++++++++++++
 hw/i386/sgx-stub.c                       |  13 ++
 hw/i386/sgx.c                            | 170 +++++++++++++++
Quoted 8 lines
 hw/i386/x86.c                            |  29 +++
 hw/vfio/common.c                         |   1 +
 include/exec/memory.h                    |  15 +-
 include/hw/i386/pc.h                     |  10 +
 include/hw/i386/sgx-epc.h                |  68 ++++++
 include/hw/i386/x86.h                    |   1 +
 include/monitor/hmp.h                    |   1 +
 include/qemu/bitops.h                    |   7 +
Quoted 1 line

Consider using scripts/git.orderfile to ease reviewers workflow.

Thanks! I never noticed this issue before, thanks!

  My ref command:
  git format-patch -32 --no-signature --cover-letter -O scripts/git.orderfile --subject-prefix="PATCH"

Yang