Skip to contentSkip to conversation

tpm: Support boot measurements

PATCH Superseded Eddie James 3 participants 17 messages Aug 7, 2023 – Aug 10, 2023 Library code

v10 of 14 Superseded Latest is v14 →

lore
Share & tools
lore
#7 lore raw

PATCH v10 6/10 doc: Add measured boot documentation

In reply to Eddie James

Briefly describe the feature and specify the requirements.

---
 doc/usage/index.rst         |  1 +
 doc/usage/measured_boot.rst | 23 +++++++++++++++++++++++
 2 files changed, 24 insertions(+)
 create mode 100644 doc/usage/measured_boot.rst
doc/usage/index.rst +1 -0
index 388e59f173..64eb362aef 100644
@@ -13,6 +13,7 @@ Use U-Boot
    partitions
    cmdline
    semihosting
+   measured_boot
 
 Shell commands
 --------------
doc/usage/measured_boot.rst new file +23 -0
new file mode 100644
index 0000000000..8357b1f480
@@ -0,0 +1,23 @@
+.. SPDX-License-Identifier: GPL-2.0+
+
+Measured Boot
+=====================
+
+U-Boot can perform a measured boot, the process of hashing various components
+of the boot process, extending the results in the TPM and logging the
+component's measurement in memory for the operating system to consume.
+
+Requirements
+---------------------
+
+* A hardware TPM 2.0 supported by the U-Boot drivers
+* CONFIG_TPM=y
+* CONFIG_MEASURED_BOOT=y
+* Device-tree configuration of the TPM device to specify the memory area
+  for event logging. The TPM device node must either contain a phandle to
+  a reserved memory region or "linux,sml-base" and "linux,sml-size"
+  indicating the address and size of the memory region. An example can be
+  found in arch/sandbox/dts/test.dts
+* The operating system must also be configured to use the memory regions
+  specified in the U-Boot device-tree in order to make use of the event
+  log.
Signature · 1 line
2.39.3
#5 lore raw

PATCH v10 4/10 bootm: Support boot measurement

In reply to Eddie James

Add a configuration option to measure the boot through the bootm function. Add the measurement state to the booti and bootz paths as well.

---
Changes since v8:
- Added a configuration option to select to ignore any existing event log. This would only be selected for systems that know that U-Boot is the first stage bootloader. This is necessary because the reserved memory region may persist through resets and so U-Boot attempts to append to the previous boot's log.

Changes since v6:
- Added comment for bootm_measure
- Fixed line length in bootm_measure

 boot/Kconfig    | 32 +++++++++++++++++++++
 boot/bootm.c    | 74 +++++++++++++++++++++++++++++++++++++++++++++++++
 cmd/booti.c     |  1 +
 cmd/bootm.c     |  2 ++
 cmd/bootz.c     |  1 +
 include/bootm.h | 11 ++++++++
 include/image.h |  1 +
 7 files changed, 122 insertions(+)
boot/Kconfig +32 -0
index a643a3d128..b160b20599 100644
@@ -675,6 +675,38 @@ config LEGACY_IMAGE_FORMAT
 	  loaded. If a board needs the legacy image format support in this
 	  case, enable it here.
 
+config MEASURED_BOOT
+	bool "Measure boot images and configuration to TPM and event log"
+	depends on HASH && TPM_V2
+	help
+	  This option enables measurement of the boot process. Measurement
+	  involves creating cryptographic hashes of the binary images that
+	  are booting and storing them in the TPM. In addition, a log of
+	  these hashes is stored in memory for the OS to verify the booted
+	  images and configuration. Enable this if the OS has configured
+	  some memory area for the event log and you intend to use some
+	  attestation tools on your system.
+
+if MEASURED_BOOT
+	config MEASURE_DEVICETREE
+	bool "Measure the devicetree image"
+	default y if MEASURED_BOOT
+	help
+	  On some platforms, the devicetree is not static as it may contain
+	  random MAC addresses or other such data that changes each boot.
+	  Therefore, it should not be measured into the TPM. In that case,
+	  disable the measurement here.
+
+	config MEASURE_IGNORE_LOG
+	bool "Ignore the existing event log"
+	default n
+	help
+	  On platforms that use an event log memory region that persists
+	  through system resets and are the first stage bootloader, then
+	  this option should be enabled to ignore any existing data in the
+	  event log memory region.
+endif # MEASURED_BOOT
+
 config SUPPORT_RAW_INITRD
 	bool "Enable raw initrd images"
 	help
boot/bootm.c +74 -0

92 lines not on this page · raw message

cmd/booti.c +1 -0
index 6ac39193db..659bb10549 100644
@@ -127,6 +127,7 @@ int do_booti(struct cmd_tbl *cmdtp, int flag, int argc, char *const argv[])
 #ifdef CONFIG_SYS_BOOT_RAMDISK_HIGH
 			      BOOTM_STATE_RAMDISK |
 #endif
+			      BOOTM_STATE_MEASURE |
 			      BOOTM_STATE_OS_PREP | BOOTM_STATE_OS_FAKE_GO |
 			      BOOTM_STATE_OS_GO,
 			      &images, 1);
cmd/bootm.c +2 -0
index 37c2af96e0..0c4a713e02 100644
@@ -161,6 +161,8 @@ int do_bootm(struct cmd_tbl *cmdtp, int flag, int argc, char *const argv[])
 		BOOTM_STATE_OS_GO;
 	if (IS_ENABLED(CONFIG_SYS_BOOT_RAMDISK_HIGH))
 		states |= BOOTM_STATE_RAMDISK;
+	if (IS_ENABLED(CONFIG_MEASURED_BOOT))
+		states |= BOOTM_STATE_MEASURE;
 	if (IS_ENABLED(CONFIG_PPC) || IS_ENABLED(CONFIG_MIPS))
 		states |= BOOTM_STATE_OS_CMDLINE;
 	ret = do_bootm_states(cmdtp, flag, argc, argv, states, &images, 1);
cmd/bootz.c +1 -0
index f1423573d2..87922bfc3c 100644
@@ -81,6 +81,7 @@ int do_bootz(struct cmd_tbl *cmdtp, int flag, int argc, char *const argv[])
 #ifdef CONFIG_SYS_BOOT_RAMDISK_HIGH
 			      BOOTM_STATE_RAMDISK |
 #endif
+			      BOOTM_STATE_MEASURE |
 			      BOOTM_STATE_OS_PREP | BOOTM_STATE_OS_FAKE_GO |
 			      BOOTM_STATE_OS_GO,
 			      &images, 1);
include/bootm.h +11 -0
index 044a4797ed..76e8e38c82 100644
@@ -55,6 +55,17 @@ ulong bootm_disable_interrupts(void);
 int bootm_find_images(int flag, int argc, char *const argv[], ulong start,
 		      ulong size);
 
+/*
+ * Measure the boot images. Measurement is the process of hashing some binary
+ * data and storing it into secure memory, i.e. TPM PCRs. In addition, each
+ * measurement is logged into the platform event log such that the operating
+ * system can access it and perform attestation of the boot.
+ *
+ * @images:	The structure containing the various images to boot (linux,
+ *		initrd, dts, etc.)
+ */
+int bootm_measure(struct bootm_headers *images);
+
 int do_bootm_states(struct cmd_tbl *cmdtp, int flag, int argc,
 		    char *const argv[], int states, struct bootm_headers *images,
 		    int boot_progress);
include/image.h +1 -0
index 01a6787d21..ea492cc2b2 100644
@@ -408,6 +408,7 @@ struct bootm_headers {
 #define BOOTM_STATE_OS_FAKE_GO	0x00000200	/* 'Almost' run the OS */
 #define BOOTM_STATE_OS_GO	0x00000400
 #define BOOTM_STATE_PRE_LOAD	0x00000800
+#define BOOTM_STATE_MEASURE	0x00001000
 	int		state;
 
 #if defined(CONFIG_LMB) && !defined(USE_HOSTCC)
Signature · 1 line
2.39.3
#8 lore raw

PATCH v10 7/10 efi_loader: fix EFI_ENTRY point on get_active_pcr_banks

In reply to Eddie James

From: Ilias Apalodimas <[email protected]>

---
 lib/efi_loader/efi_tcg2.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
lib/efi_loader/efi_tcg2.c +2 -2
index 5f0f4b5dd2..829bae7436 100644
@@ -718,16 +718,16 @@ efi_tcg2_get_active_pcr_banks(struct efi_tcg2_protocol *this,
 	struct udevice *dev;
 	efi_status_t ret;
 
+	EFI_ENTRY("%p, %p", this, active_pcr_banks);
+
 	if (!this || !active_pcr_banks) {
 		ret = EFI_INVALID_PARAMETER;
 		goto out;
 	}
-
 	ret = tcg2_platform_get_tpm2(&dev);
 	if (ret != EFI_SUCCESS)
 		goto out;
 
-	EFI_ENTRY("%p, %p", this, active_pcr_banks);
 	ret = tcg2_get_active_pcr_banks(dev, active_pcr_banks);
 
 out:
Signature · 1 line
2.39.3
#0 lore raw

v10 cover tpm: Support boot measurements

This series adds support for measuring the boot images more generically than the existing EFI support. Several EFI functions have been moved to the TPM layer. The series includes optional measurement from the bootm command.
A new test case has been added for the bootm measurement to test the new path, and the sandbox TPM2 driver has been updated to support this use case.

Changes since v9:
- Rebase and add Ilias' fixes (thanks!)

Changes since v8:
- Fix a sandbox driver off-by-one error in checking the property type.
- Fix log parsing again - any data corruption seen while replaying the event log was failing the entire measurement.
- Added an option to ignore the existing log and a configuration option for systems to select that for the bootm measurement. This would only be selected for systems that know that U-Boot is the first stage bootloader. This is necessary because the reserved memory region may persist through resets and so U-Boot attempts to append to the previous boot's log.

Changes since v7:
- Change name of tcg2_init_log and add more documentation
- Add a check, when parsing the event log header, to ensure that the previous stage bootloader used all the active PCRs.
- Change name of tcg2_log_find_end
- Fix the greater than or equal to check to exit the log parsing
- Make sure log_position is 0 if there is any error discovering the log
- Return errors parsing the log if the data is corrupt so that we don't end up with half a log

Changes since v6:
- Added comment for bootm_measure
- Fixed line length in bootm_measure
- Added Linaro copyright for all the EFI moved code
- Changed tcg2_init_log (and by extension, tcg2_measurement_init) to copy any discovered event log to the user's log if passed in.

Changes since v5:
- Re-ordered the patches to put the sandbox TPM driver patch second
- Remove unused platform_get_eventlog in efi_tcg2.c
- First look for tpm_event_log_* properties instead of linux,sml-*
- Fix efi_tcg2.c compilation
- Select SHA* configs
- Remove the !SANDBOX dependency for EFI TCG2
- Only compile in the measurement u-boot command when CONFIG_MEASURED_BOOT is enabled

Changes since v4:
- Remove tcg2_measure_event function and check for NULL data in tcg2_measure_data
- Use tpm_auto_startup
- Fix efi_tcg2.c compilation for removing tcg2_pcr_read function
- Change PCR indexes for initrd and dtb
- Drop u8 casting in measurement test
- Use bullets in documentation

Changes since v3:
- Reordered headers
- Refactored more of EFI code into common code
Removed digest_info structure and instead used the common alg_to_mask
and alg_to_len
Improved event log parsing in common code to get it equivalent to EFI
Common code now extends PCR if previous bootloader stage couldn't No need to allocate memory in the common code, so EFI copies the discovered buffer like it did before
Rename efi measure_event function

Changes since v2:
- Add documentation.
- Changed reserved memory address to the top of the RAM for sandbox dts.
- Add measure state to booti and bootz.
- Skip measurement for EFI images that should be measured

Changes since v1:
- Refactor TPM layer functions to allow EFI system to use them, and remove duplicate EFI functions.
- Add test case
- Drop #ifdefs for bootm
- Add devicetree measurement config option
- Update sandbox TPM driver

Eddie James (6):
tpm: Fix spelling for tpmu_ha union
tpm: sandbox: Update for needed TPM2 capabilities
tpm: Support boot measurements
bootm: Support boot measurement
test: Add sandbox TPM boot measurement
doc: Add measured boot documentation

Ilias Apalodimas (4):
efi_loader: fix EFI_ENTRY point on get_active_pcr_banks
test: use a non system PCR for testing PCR extend test/py: only run 'tpm2 autostart' to init the tpm fix armv7 compilation warning

 arch/sandbox/dts/sandbox.dtsi  |   13 +
 arch/sandbox/dts/test.dts      |   13 +
 boot/Kconfig                   |   32 +
 boot/bootm.c                   |   74 +++
 cmd/booti.c                    |    1 +
 cmd/bootm.c                    |    2 +
 cmd/bootz.c                    |    1 +
 configs/sandbox_defconfig      |    1 +
 doc/usage/index.rst            |    1 +
 doc/usage/measured_boot.rst    |   23 +
 drivers/tpm/tpm2_tis_sandbox.c |  100 ++-
 include/bootm.h                |   11 +
 include/efi_tcg2.h             |   44 --
 include/image.h                |    1 +
 include/test/suites.h          |    1 +
 include/tpm-v2.h               |  263 +++++++-
 lib/Kconfig                    |    4 +
 lib/efi_loader/Kconfig         |    2 -
 lib/efi_loader/efi_tcg2.c      | 1056 +++-----------------------------
 lib/tpm-v2.c                   |  814 ++++++++++++++++++++++++
 test/boot/Makefile             |    1 +
 test/boot/measurement.c        |   66 ++
 test/cmd_ut.c                  |    4 +
 test/py/tests/test_tpm2.py     |   24 +-
 24 files changed, 1484 insertions(+), 1068 deletions(-)
 create mode 100644 doc/usage/measured_boot.rst
 create mode 100644 test/boot/measurement.c
Signature · 1 line
2.39.3
#2 lore raw

PATCH v10 2/10 tpm: sandbox: Update for needed TPM2 capabilities

The driver needs to support getting the PCRs in the capabilities command. Fix various other things and support the max number of PCRs for TPM2.
Remove the !SANDBOX dependency for EFI TCG2 as well.

---
Changes since v8:
- Use >= for checking the property against TPM2_PROPERTIES_OFFSET

Changes since v5:
- Remove the !SANDBOX dependency for EFI TCG2

 drivers/tpm/tpm2_tis_sandbox.c | 100 ++++++++++++++++++++++++---------
 lib/efi_loader/Kconfig         |   2 -
 2 files changed, 72 insertions(+), 30 deletions(-)
drivers/tpm/tpm2_tis_sandbox.c +72 -28
index e4004cfcca..d15a28d9fc 100644
@@ -22,11 +22,6 @@ enum tpm2_hierarchy {
 	TPM2_HIERARCHY_NB,
 };
 
-/* Subset of supported capabilities */
-enum tpm2_capability {
-	TPM_CAP_TPM_PROPERTIES = 0x6,
-};
-
 /* Subset of supported properties */
 #define TPM2_PROPERTIES_OFFSET 0x0000020E
 
@@ -38,7 +33,8 @@ enum tpm2_cap_tpm_property {
 	TPM2_PROPERTY_NB,
 };
 
-#define SANDBOX_TPM_PCR_NB 1
+#define SANDBOX_TPM_PCR_NB TPM2_MAX_PCRS
+#define SANDBOX_TPM_PCR_SELECT_MAX	((SANDBOX_TPM_PCR_NB + 7) / 8)
 
 /*
  * Information about our TPM emulation. This is preserved in the sandbox
@@ -433,7 +429,7 @@ static int sandbox_tpm2_xfer(struct udevice *dev, const u8 *sendbuf,
 	int i, j;
 
 	/* TPM2_GetProperty */
-	u32 capability, property, property_count;
+	u32 capability, property, property_count, val;
 
 	/* TPM2_PCR_Read/Extend variables */
 	int pcr_index = 0;
@@ -542,19 +538,32 @@ static int sandbox_tpm2_xfer(struct udevice *dev, const u8 *sendbuf,
 	case TPM2_CC_GET_CAPABILITY:
 		capability = get_unaligned_be32(sent);
 		sent += sizeof(capability);
-		if (capability != TPM_CAP_TPM_PROPERTIES) {
-			printf("Sandbox TPM only support TPM_CAPABILITIES\n");
-			return TPM2_RC_HANDLE;
-		}
-
 		property = get_unaligned_be32(sent);
 		sent += sizeof(property);
-		property -= TPM2_PROPERTIES_OFFSET;
-
 		property_count = get_unaligned_be32(sent);
 		sent += sizeof(property_count);
-		if (!property_count ||
-		    property + property_count > TPM2_PROPERTY_NB) {
+
+		switch (capability) {
+		case TPM2_CAP_PCRS:
+			break;
+		case TPM2_CAP_TPM_PROPERTIES:
+			if (!property_count) {
+				rc = TPM2_RC_HANDLE;
+				return sandbox_tpm2_fill_buf(recv, recv_len,
+							     tag, rc);
+			}
+
+			if (property >= TPM2_PROPERTIES_OFFSET &&
+			    ((property - TPM2_PROPERTIES_OFFSET) +
+			     property_count > TPM2_PROPERTY_NB)) {
+				rc = TPM2_RC_HANDLE;
+				return sandbox_tpm2_fill_buf(recv, recv_len,
+							     tag, rc);
+			}
+			break;
+		default:
+			printf("Sandbox TPM2 only supports TPM2_CAP_PCRS or "
+			       "TPM2_CAP_TPM_PROPERTIES\n");
 			rc = TPM2_RC_HANDLE;
 			return sandbox_tpm2_fill_buf(recv, recv_len, tag, rc);
 		}
@@ -578,18 +587,53 @@ static int sandbox_tpm2_xfer(struct udevice *dev, const u8 *sendbuf,
 		put_unaligned_be32(capability, recv);
 		recv += sizeof(capability);
 
-		/* Give the number of properties that follow */
-		put_unaligned_be32(property_count, recv);
-		recv += sizeof(property_count);
-
-		/* Fill with the properties */
-		for (i = 0; i < property_count; i++) {
-			put_unaligned_be32(TPM2_PROPERTIES_OFFSET + property +
-					   i, recv);
-			recv += sizeof(property);
-			put_unaligned_be32(tpm->properties[property + i],
-					   recv);
-			recv += sizeof(property);
+		switch (capability) {
+		case TPM2_CAP_PCRS:
+			/* Give the number of algorithms supported - just SHA256 */
+			put_unaligned_be32(1, recv);
+			recv += sizeof(u32);
+
+			/* Give SHA256 algorithm */
+			put_unaligned_be16(TPM2_ALG_SHA256, recv);
+			recv += sizeof(u16);
+
+			/* Select the PCRs supported */
+			*recv = SANDBOX_TPM_PCR_SELECT_MAX;
+			recv++;
+
+			/* Activate all the PCR bits */
+			for (i = 0; i < SANDBOX_TPM_PCR_SELECT_MAX; ++i) {
+				*recv = 0xff;
+				recv++;
+			}
+			break;
+		case TPM2_CAP_TPM_PROPERTIES:
+			/* Give the number of properties that follow */
+			put_unaligned_be32(property_count, recv);
+			recv += sizeof(property_count);
+
+			/* Fill with the properties */
+			for (i = 0; i < property_count; i++) {
+				put_unaligned_be32(property + i, recv);
+				recv += sizeof(property);
+				if (property >= TPM2_PROPERTIES_OFFSET) {
+					val = tpm->properties[(property -
+						TPM2_PROPERTIES_OFFSET) + i];
+				} else {
+					switch (property) {
+					case TPM2_PT_PCR_COUNT:
+						val = SANDBOX_TPM_PCR_NB;
+						break;
+					default:
+						val = 0xffffffff;
+						break;
+					}
+				}
+
+				put_unaligned_be32(val, recv);
+				recv += sizeof(property);
+			}
+			break;
 		}
 
 		/* Add trailing \0 */
lib/efi_loader/Kconfig +0 -2
index c5835e6ef6..605719d2b6 100644
@@ -333,8 +333,6 @@ config EFI_TCG2_PROTOCOL
 	bool "EFI_TCG2_PROTOCOL support"
 	default y
 	depends on TPM_V2
-	# Sandbox TPM currently fails on GetCapabilities needed for TCG2
-	depends on !SANDBOX
 	select SHA1
 	select SHA256
 	select SHA384
Signature · 1 line
2.39.3
#12 lore raw

PATCH v10 9/10 test/py: only run 'tpm2 autostart' to init the tpm

In reply to Eddie James

From: Ilias Apalodimas <[email protected]>

commit xxxxxxxx ("")
replaced the forced and sandbox tpm2 initialization running 'tpm2 autostart' instead of the startup tpm sequence. The difference is that the new function handles the internal tpm_init state internally and doesn't return an error when trying to initialize the tpm multiple times. Replace the remaining instances

---
 test/py/tests/test_tpm2.py | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)
test/py/tests/test_tpm2.py +2 -6
index 8cd3046285..5bbb2d2069 100644
@@ -61,7 +61,7 @@ def test_tpm2_init(u_boot_console):
     skip_test = u_boot_console.config.env.get('env__tpm_device_test_skip', False)
     if skip_test:
         pytest.skip('skip TPM device test')
-    u_boot_console.run_command('tpm2 init')
+    u_boot_console.run_command('tpm2 autostart')
     output = u_boot_console.run_command('echo $?')
     assert output.endswith('0')
 
@@ -97,11 +97,7 @@ def test_tpm2_sandbox_self_test_full(u_boot_console):
     """
     if is_sandbox(u_boot_console):
         u_boot_console.restart_uboot()
-        u_boot_console.run_command('tpm2 init')
-        output = u_boot_console.run_command('echo $?')
-        assert output.endswith('0')
-
-        u_boot_console.run_command('tpm2 startup TPM2_SU_CLEAR')
+        u_boot_console.run_command('tpm2 autostart')
         output = u_boot_console.run_command('echo $?')
         assert output.endswith('0')
 
Signature · 1 line
2.39.3
#3 lore raw

PATCH v10 3/10 tpm: Support boot measurements

In reply to Eddie James

Add TPM2 functions to support boot measurement. This includes starting up the TPM, initializing/appending the event log, and measuring the U-Boot version. Much of the code was used in the EFI subsystem, so remove it there and use the common functions.

---
Changes since v8:
- Fix log parsing again - any data corruption seen while replaying the event log was failing the entire measurement.
- Added an option to ignore the existing log. This should only be used for systems that know that U-Boot is the first stage bootloader. This is necessary because the reserved memory region may persist through resets and so U-Boot attempts to append to the previous boot's log.

Changes since v7:
- Change name of tcg2_init_log and add more documentation
- Add a check, when parsing the event log header, to ensure that the previous stage bootloader used all the active PCRs.
- Change name of tcg2_log_find_end
- Fix the greater than or equal to check to exit the log parsing
- Make sure log_position is 0 if there is any error discovering the log
- Return errors parsing the log if the data is corrupt so that we don't end up with half a log

Changes since v6:
- Added Linaro copyright for all the EFI moved code
- Changed tcg2_init_log (and by extension, tcg2_measurement_init) to copy any discovered event log to the user's log if passed in.

Changes since v5:
- Remove unused platform_get_eventlog in efi_tcg2.c
- First look for tpm_event_log_* properties instead of linux,sml-*
- Fix efi_tcg2.c compilation
- Select SHA* configs

Changes since v4:
- Remove tcg2_measure_event function and check for NULL data in tcg2_measure_data
- Use tpm_auto_startup
- Fix efi_tcg2.c compilation for removing tcg2_pcr_read function

Changes since v3:
- Reordered headers
- Refactored more of EFI code into common code
Removed digest_info structure and instead used the common alg_to_mask
and alg_to_len
Improved event log parsing in common code to get it equivalent to EFI
Common code now extends PCR if previous bootloader stage couldn't No need to allocate memory in the common code, so EFI copies the discovered buffer like it did before
Rename efi measure_event function

Changes since v1:
- Refactor TPM layer functions to allow EFI system to use them, and remove duplicate EFI functions

 include/efi_tcg2.h        |   44 --
 include/tpm-v2.h          |  259 +++++++++
 lib/Kconfig               |    4 +
 lib/efi_loader/efi_tcg2.c | 1054 +++----------------------------------
 lib/tpm-v2.c              |  814 ++++++++++++++++++++++++++++
 5 files changed, 1154 insertions(+), 1021 deletions(-)
include/efi_tcg2.h +0 -44
index b1c3abd097..b21c5cb3dd 100644
@@ -129,50 +129,6 @@ struct efi_tcg2_boot_service_capability {
 #define BOOT_SERVICE_CAPABILITY_MIN \
 	offsetof(struct efi_tcg2_boot_service_capability, number_of_pcr_banks)
 
-#define TCG_EFI_SPEC_ID_EVENT_SIGNATURE_03 "Spec ID Event03"
-#define TCG_EFI_SPEC_ID_EVENT_SPEC_VERSION_MAJOR_TPM2 2
-#define TCG_EFI_SPEC_ID_EVENT_SPEC_VERSION_MINOR_TPM2 0
-#define TCG_EFI_SPEC_ID_EVENT_SPEC_VERSION_ERRATA_TPM2 2
-
-/**
- *  struct TCG_EfiSpecIdEventAlgorithmSize - hashing algorithm information
- *
- *  @algorithm_id:	algorithm defined in enum tpm2_algorithms
- *  @digest_size:	size of the algorithm
- */
-struct tcg_efi_spec_id_event_algorithm_size {
-	u16      algorithm_id;
-	u16      digest_size;
-} __packed;
-
-/**
- * struct TCG_EfiSpecIDEventStruct - content of the event log header
- *
- * @signature:			signature, set to Spec ID Event03
- * @platform_class:		class defined in TCG ACPI Specification
- *				Client  Common Header.
- * @spec_version_minor:		minor version
- * @spec_version_major:		major version
- * @spec_version_errata:	major version
- * @uintn_size:			size of the efi_uintn_t fields used in various
- *				data structures used in this specification.
- *				0x01 indicates u32  and 0x02  indicates u64
- * @number_of_algorithms:	hashing algorithms used in this event log
- * @digest_sizes:		array of number_of_algorithms pairs
- *				1st member defines the algorithm id
- *				2nd member defines the algorithm size
- */
-struct tcg_efi_spec_id_event {
-	u8 signature[16];
-	u32 platform_class;
-	u8 spec_version_minor;
-	u8 spec_version_major;
-	u8 spec_errata;
-	u8 uintn_size;
-	u32 number_of_algorithms;
-	struct tcg_efi_spec_id_event_algorithm_size digest_sizes[];
-} __packed;
-
 /**
  * struct tdEFI_TCG2_FINAL_EVENTS_TABLE - log entries after Get Event Log
  * @version:		version number for this structure
include/tpm-v2.h +259 -0

283 lines not on this page · raw message

lib/Kconfig +4 -0
index c8b3ec1ec9..cd37cba511 100644
@@ -423,6 +423,10 @@ config TPM
 	bool "Trusted Platform Module (TPM) Support"
 	depends on DM
 	imply DM_RNG
+	select SHA1
+	select SHA256
+	select SHA384
+	select SHA512
 	help
 	  This enables support for TPMs which can be used to provide security
 	  features for your board. The TPM can be connected via LPC or I2C
lib/efi_loader/efi_tcg2.c +77 -977

1,320 lines not on this page · raw message

lib/tpm-v2.c +814 -0

836 lines not on this page · raw message

Signature · 1 line
2.39.3
#1 lore raw

PATCH v10 1/10 tpm: Fix spelling for tpmu_ha union

In reply to Eddie James

tmpu -> tpmu

---
 include/tpm-v2.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
include/tpm-v2.h +2 -2
index 2b6980e441..6684033deb 100644
@@ -169,7 +169,7 @@ struct tcg_pcr_event {
 /**
  * Definition of TPMU_HA Union
  */
-union tmpu_ha {
+union tpmu_ha {
 	u8 sha1[TPM2_SHA1_DIGEST_SIZE];
 	u8 sha256[TPM2_SHA256_DIGEST_SIZE];
 	u8 sm3_256[TPM2_SM3_256_DIGEST_SIZE];
@@ -185,7 +185,7 @@ union tmpu_ha {
  */
 struct tpmt_ha {
 	u16 hash_alg;
-	union tmpu_ha digest;
+	union tpmu_ha digest;
 } __packed;
 
 /**
Signature · 1 line
2.39.3
#11 lore raw

PATCH v10 8/10 test: use a non system PCR for testing PCR extend

In reply to Eddie James

From: Ilias Apalodimas <[email protected]>

We currently use PCR 0 for testing the PCR read/extend functionality in our selftests. How ever those PCRs are defined by the TCG spec for platform use. For example if the tests run *after* the efi subsystem initialization, which extends PCRs 0 & 7 it will give a false positive.

So let's switch over to a PCR which is more suitable and is defined for OS use. It's worth noting that we are using PCR10 here, since PCR9 is used internally by U-Boot if we choose to measure the loaded DTB

---
 test/py/tests/test_tpm2.py | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)
test/py/tests/test_tpm2.py +8 -8
index fce689cd99..8cd3046285 100644
@@ -236,7 +236,7 @@ def test_tpm2_dam_parameters(u_boot_console):
 def test_tpm2_pcr_read(u_boot_console):
     """Execute a TPM2_PCR_Read command.
 
-    Perform a PCR read of the 0th PCR. Must be zero.
+    Perform a PCR read of the 10th PCR. Must be zero.
     """
     if is_sandbox(u_boot_console):
         tpm2_sandbox_init(u_boot_console)
@@ -244,7 +244,7 @@ def test_tpm2_pcr_read(u_boot_console):
     force_init(u_boot_console)
     ram = u_boot_utils.find_ram_base(u_boot_console)
 
-    read_pcr = u_boot_console.run_command('tpm2 pcr_read 0 0x%x' % ram)
+    read_pcr = u_boot_console.run_command('tpm2 pcr_read 10 0x%x' % ram)
     output = u_boot_console.run_command('echo $?')
     assert output.endswith('0')
 
@@ -254,7 +254,7 @@ def test_tpm2_pcr_read(u_boot_console):
     updates = int(re.findall(r'\d+', str)[0])
 
     # Check the output value
-    assert 'PCR #0 content' in read_pcr
+    assert 'PCR #10 content' in read_pcr
     assert '00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00' in read_pcr
 
 @pytest.mark.buildconfigspec('cmd_tpm_v2')
@@ -272,19 +272,19 @@ def test_tpm2_pcr_extend(u_boot_console):
     force_init(u_boot_console)
     ram = u_boot_utils.find_ram_base(u_boot_console)
 
-    read_pcr = u_boot_console.run_command('tpm2 pcr_read 0 0x%x' % (ram + 0x20))
+    read_pcr = u_boot_console.run_command('tpm2 pcr_read 10 0x%x' % (ram + 0x20))
     output = u_boot_console.run_command('echo $?')
     assert output.endswith('0')
     str = re.findall(r'\d+ known updates', read_pcr)[0]
     updates = int(re.findall(r'\d+', str)[0])
 
-    u_boot_console.run_command('tpm2 pcr_extend 0 0x%x' % ram)
+    u_boot_console.run_command('tpm2 pcr_extend 10 0x%x' % ram)
     output = u_boot_console.run_command('echo $?')
     assert output.endswith('0')
 
     # Read the value back into a different place so we can still use 'ram' as
     # our zero bytes
-    read_pcr = u_boot_console.run_command('tpm2 pcr_read 0 0x%x' % (ram + 0x20))
+    read_pcr = u_boot_console.run_command('tpm2 pcr_read 10 0x%x' % (ram + 0x20))
     output = u_boot_console.run_command('echo $?')
     assert output.endswith('0')
     assert 'f5 a5 fd 42 d1 6a 20 30 27 98 ef 6e d3 09 97 9b' in read_pcr
@@ -294,11 +294,11 @@ def test_tpm2_pcr_extend(u_boot_console):
     new_updates = int(re.findall(r'\d+', str)[0])
     assert (updates + 1) == new_updates
 
-    u_boot_console.run_command('tpm2 pcr_extend 0 0x%x' % ram)
+    u_boot_console.run_command('tpm2 pcr_extend 10 0x%x' % ram)
     output = u_boot_console.run_command('echo $?')
     assert output.endswith('0')
 
-    read_pcr = u_boot_console.run_command('tpm2 pcr_read 0 0x%x' % (ram + 0x20))
+    read_pcr = u_boot_console.run_command('tpm2 pcr_read 10 0x%x' % (ram + 0x20))
     output = u_boot_console.run_command('echo $?')
     assert output.endswith('0')
     assert '7a 05 01 f5 95 7b df 9c b3 a8 ff 49 66 f0 22 65' in read_pcr
Signature · 1 line
2.39.3
#14 lore raw

PATCH v10 10/10 fix armv7 compilation warning

In reply to Eddie James

From: Ilias Apalodimas <[email protected]>

---
 lib/tpm-v2.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
lib/tpm-v2.c +1 -1
index d22e21985b..bd0fb078dc 100644
@@ -671,7 +671,7 @@ __weak int tcg2_platform_get_log(struct udevice *dev, void **addr, u32 *size)
 	} else {
 		struct ofnode_phandle_args args;
 		phys_addr_t a;
-		phys_size_t s;
+		fdt_size_t s;
 
 		if (dev_read_phandle_with_args(dev, "memory-region", NULL, 0,
 					       0, &args))
Signature · 1 line
2.39.3
#6 lore raw

PATCH v10 5/10 test: Add sandbox TPM boot measurement

In reply to Eddie James

Use the sandbox TPM driver to measure some boot images in a unit test case.

---
Changes since v5:
- Only compile in the measurement u-boot command when CONFIG_MEASURED_BOOT is enabled.

 arch/sandbox/dts/sandbox.dtsi | 13 +++++++
 arch/sandbox/dts/test.dts     | 13 +++++++
 configs/sandbox_defconfig     |  1 +
 include/test/suites.h         |  1 +
 test/boot/Makefile            |  1 +
 test/boot/measurement.c       | 66 +++++++++++++++++++++++++++++++++++
 test/cmd_ut.c                 |  4 +++
 7 files changed, 99 insertions(+)
 create mode 100644 test/boot/measurement.c
arch/sandbox/dts/sandbox.dtsi +13 -0
index 30a305c4d2..ed39d20c6a 100644
@@ -4,11 +4,23 @@
  * and sandbox64 builds.
  */
 
+#include <config.h>
 #include <dt-bindings/input/input.h>
 
 #define USB_CLASS_HUB			9
 
 / {
+	reserved-memory {
+		#address-cells = <1>;
+		#size-cells = <1>;
+		ranges;
+
+		event_log: tcg_event_log {
+			no-map;
+			reg = <(CFG_SYS_SDRAM_SIZE - 0x2000) 0x2000>;
+		};
+	};
+
 	binman {
 	};
 
@@ -336,6 +348,7 @@
 
 	tpm2 {
 		compatible = "sandbox,tpm2";
+		memory-region = <&event_log>;
 	};
 
 	triangle {
arch/sandbox/dts/test.dts +13 -0
index ff9f9222e6..0bac073178 100644
@@ -9,6 +9,7 @@
 
 /dts-v1/;
 
+#include <config.h>
 #include <dt-bindings/gpio/gpio.h>
 #include <dt-bindings/gpio/sandbox-gpio.h>
 #include <dt-bindings/input/input.h>
@@ -66,6 +67,17 @@
 		osd0 = "/osd";
 	};
 
+	reserved-memory {
+		#address-cells = <1>;
+		#size-cells = <1>;
+		ranges;
+
+		event_log: tcg_event_log {
+			no-map;
+			reg = <(CFG_SYS_SDRAM_SIZE - 0x2000) 0x2000>;
+		};
+	};
+
 	binman: binman {
 	};
 
@@ -1365,6 +1377,7 @@
 
 	tpm2 {
 		compatible = "sandbox,tpm2";
+		memory-region = <&event_log>;
 	};
 
 	tpm {
configs/sandbox_defconfig +1 -0
index 1ec44d5b33..85ef821296 100644
@@ -344,3 +344,4 @@ CONFIG_TEST_FDTDEC=y
 CONFIG_UNIT_TEST=y
 CONFIG_UT_TIME=y
 CONFIG_UT_DM=y
+CONFIG_MEASURED_BOOT=y
include/test/suites.h +1 -0
index 1c7dc65966..48ed549c13 100644
@@ -45,6 +45,7 @@ int do_ut_font(struct cmd_tbl *cmdtp, int flag, int argc, char *const argv[]);
 int do_ut_lib(struct cmd_tbl *cmdtp, int flag, int argc, char *const argv[]);
 int do_ut_loadm(struct cmd_tbl *cmdtp, int flag, int argc, char *const argv[]);
 int do_ut_log(struct cmd_tbl *cmdtp, int flag, int argc, char * const argv[]);
+int do_ut_measurement(struct cmd_tbl *cmdtp, int flag, int argc, char * const argv[]);
 int do_ut_mem(struct cmd_tbl *cmdtp, int flag, int argc, char *const argv[]);
 int do_ut_optee(struct cmd_tbl *cmdtp, int flag, int argc, char *const argv[]);
 int do_ut_overlay(struct cmd_tbl *cmdtp, int flag, int argc,
test/boot/Makefile +1 -0
index 22ed61c8fa..2dbb032a7e 100644
@@ -4,6 +4,7 @@
 
 obj-$(CONFIG_BOOTSTD) += bootdev.o bootstd_common.o bootflow.o bootmeth.o
 obj-$(CONFIG_FIT) += image.o
+obj-$(CONFIG_MEASURED_BOOT) += measurement.o
 
 obj-$(CONFIG_EXPO) += expo.o
 
test/boot/measurement.c new file +66 -0

66 lines not on this page · raw message

test/cmd_ut.c +4 -0
index 0cb514490b..23198a5064 100644
@@ -99,6 +99,10 @@ static struct cmd_tbl cmd_ut_sub[] = {
 #if CONFIG_IS_ENABLED(UT_UNICODE) && !defined(API_BUILD)
 	U_BOOT_CMD_MKENT(unicode, CONFIG_SYS_MAXARGS, 1, do_ut_unicode, "", ""),
 #endif
+#ifdef CONFIG_MEASURED_BOOT
+	U_BOOT_CMD_MKENT(measurement, CONFIG_SYS_MAXARGS, 1, do_ut_measurement,
+			 "", ""),
+#endif
 #ifdef CONFIG_SANDBOX
 	U_BOOT_CMD_MKENT(compression, CONFIG_SYS_MAXARGS, 1, do_ut_compression,
 			 "", ""),
Signature · 1 line
2.39.3
#15 lore raw

Re: [PATCH v10 10/10] fix armv7 compilation warning

In reply to Eddie James

Hi Eddie,

Quoted 13 lines · Eddie James

From: Ilias Apalodimas <[email protected]>

---
 lib/tpm-v2.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
lib/tpm-v2.c +0 -0
index d22e21985b..bd0fb078dc 100644

lib/tpm-v2.c __weak int tcg2_platform_get_log(struct udevice *dev, void **addr, u32 *size) · line 674

Quoted 5 lines
lib/tpm-v2.c +1 -1
@@ -671,7 +671,7 @@ __weak int tcg2_platform_get_log(struct udevice *dev, void **addr, u32 *size)
        } else {
                struct ofnode_phandle_args args;
                phys_addr_t a;
-               phys_size_t s;
+               fdt_size_t s;

It's been a while since I fixed this, but iirc this had to be squashed to your changes?

Cheers
/Ilias

Quoted 6 lines
lib/tpm-v2.c +0 -1
                if (dev_read_phandle_with_args(dev, "memory-region", NULL, 0,
                                               0, &args))
--

2.39.3

#9 lore raw

Re: [PATCH v10 07/10] efi_loader: fix EFI_ENTRY point on get_active_pcr_banks

In reply to Eddie James

Hi Eddie,

Quoted 2 lines · Eddie James

From: Ilias Apalodimas <[email protected]>

We need a commit message for that. Something along the lines of efi_tcg2_get_active_pcr_banks() doesnt immediately call the EFI_ENTRY() wrappers once it enters the function. Move the call a few lines above and cover the error cases properly as well

Thanks
/Ilias

Quoted 32 lines
---
 lib/efi_loader/efi_tcg2.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
lib/efi_loader/efi_tcg2.c +2 -3
index 5f0f4b5dd2..829bae7436 100644
@@ -718,16 +718,16 @@ efi_tcg2_get_active_pcr_banks(struct efi_tcg2_protocol *this,
        struct udevice *dev;
        efi_status_t ret;
 
+       EFI_ENTRY("%p, %p", this, active_pcr_banks);
+
        if (!this || !active_pcr_banks) {
                ret = EFI_INVALID_PARAMETER;
                goto out;
        }
-
        ret = tcg2_platform_get_tpm2(&dev);
        if (ret != EFI_SUCCESS)
                goto out;
 
-       EFI_ENTRY("%p, %p", this, active_pcr_banks);
        ret = tcg2_get_active_pcr_banks(dev, active_pcr_banks);
 
 out:
--

2.39.3

#13 lore raw

Re: [PATCH v10 09/10] test/py: only run 'tpm2 autostart' to init the tpm

In reply to Eddie James
Quoted 11 lines · Eddie James

From: Ilias Apalodimas <[email protected]>

commit xxxxxxxx ("")
replaced the forced and sandbox tpm2 initialization running 'tpm2 autostart' instead of the startup tpm sequence. The difference is that the new function handles the internal tpm_init state internally and doesn't return an error when trying to initialize the tpm multiple times. Replace the remaining instances

We don't need this anymore, commit 789ed2784256 ("test/py: replace 'tpm2 init, startup, selftest' sequences") should be enough. In any case, I'll queue this on the CI without this and make sure.

Cheers
/Ilias

Quoted 33 lines
---
 test/py/tests/test_tpm2.py | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)
test/py/tests/test_tpm2.py +2 -7
index 8cd3046285..5bbb2d2069 100644
@@ -61,7 +61,7 @@ def test_tpm2_init(u_boot_console):
     skip_test = u_boot_console.config.env.get('env__tpm_device_test_skip', False)
     if skip_test:
         pytest.skip('skip TPM device test')
-    u_boot_console.run_command('tpm2 init')
+    u_boot_console.run_command('tpm2 autostart')
     output = u_boot_console.run_command('echo $?')
     assert output.endswith('0')
 
@@ -97,11 +97,7 @@ def test_tpm2_sandbox_self_test_full(u_boot_console):
     """
     if is_sandbox(u_boot_console):
         u_boot_console.restart_uboot()
-        u_boot_console.run_command('tpm2 init')
-        output = u_boot_console.run_command('echo $?')
-        assert output.endswith('0')
-
-        u_boot_console.run_command('tpm2 startup TPM2_SU_CLEAR')
+        u_boot_console.run_command('tpm2 autostart')
         output = u_boot_console.run_command('echo $?')
         assert output.endswith('0')
 
--

2.39.3

#16 lore raw

Re: [PATCH v10 10/10] fix armv7 compilation warning

In reply to Ilias Apalodimas
Quoted 10 lines · Ilias Apalodimas

Hi Eddie,

Quoted 7 lines · Eddie James

From: Ilias Apalodimas <[email protected]>

---
  lib/tpm-v2.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 12 lines

lib/tpm-v2.c __weak int tcg2_platform_get_log(struct udevice *dev, void **addr, u32 *size) · line 674

Quoted 10 lines
lib/tpm-v2.c +1 -1
index d22e21985b..bd0fb078dc 100644
@@ -671,7 +671,7 @@ __weak int tcg2_platform_get_log(struct udevice *dev, void **addr, u32 *size)
         } else {
                 struct ofnode_phandle_args args;
                 phys_addr_t a;
-               phys_size_t s;
+               fdt_size_t s;

It's been a while since I fixed this, but iirc this had to be squashed to your changes?

Oh, ok, I can do that.

Thanks,

Eddie

Quoted 8 lines

Cheers
/Ilias

Quoted 5 lines
lib/tpm-v2.c +0 -1
                 if (dev_read_phandle_with_args(dev, "memory-region", NULL, 0,
                                                0, &args))
--

2.39.3

#10 lore raw

Re: [PATCH v10 07/10] efi_loader: fix EFI_ENTRY point on get_active_pcr_banks

In reply to Ilias Apalodimas
Quoted 8 lines · Ilias Apalodimas

Hi Eddie,

Quoted 1 line · Eddie James

From: Ilias Apalodimas <[email protected]>

We need a commit message for that. Something along the lines of efi_tcg2_get_active_pcr_banks() doesnt immediately call the EFI_ENTRY() wrappers once it enters the function. Move the call a few lines above and cover the error cases properly as well

Oops, yep. I'll fix this too.

Thanks,

Eddie

Quoted 34 lines

Thanks
/Ilias

Quoted 31 lines
---
  lib/efi_loader/efi_tcg2.c | 4 ++--
  1 file changed, 2 insertions(+), 2 deletions(-)
lib/efi_loader/efi_tcg2.c +2 -3
index 5f0f4b5dd2..829bae7436 100644
@@ -718,16 +718,16 @@ efi_tcg2_get_active_pcr_banks(struct efi_tcg2_protocol *this,
         struct udevice *dev;
         efi_status_t ret;
 
+       EFI_ENTRY("%p, %p", this, active_pcr_banks);
+
         if (!this || !active_pcr_banks) {
                 ret = EFI_INVALID_PARAMETER;
                 goto out;
         }
-
         ret = tcg2_platform_get_tpm2(&dev);
         if (ret != EFI_SUCCESS)
                 goto out;
 
-       EFI_ENTRY("%p, %p", this, active_pcr_banks);
         ret = tcg2_get_active_pcr_banks(dev, active_pcr_banks);
 
  out:
--

2.39.3

#4 lore raw
In reply to Eddie James

Hi Eddie,

Quoted 58 lines · Eddie James

Add TPM2 functions to support boot measurement. This includes starting up the TPM, initializing/appending the event log, and measuring the U-Boot version. Much of the code was used in the EFI subsystem, so remove it there and use the common functions.

---
Changes since v8:
- Fix log parsing again - any data corruption seen while replaying the event log was failing the entire measurement.
- Added an option to ignore the existing log. This should only be used for systems that know that U-Boot is the first stage bootloader. This is necessary because the reserved memory region may persist through resets and so U-Boot attempts to append to the previous boot's log.

Changes since v7:
- Change name of tcg2_init_log and add more documentation
- Add a check, when parsing the event log header, to ensure that the previous stage bootloader used all the active PCRs.
- Change name of tcg2_log_find_end
- Fix the greater than or equal to check to exit the log parsing
- Make sure log_position is 0 if there is any error discovering the log
- Return errors parsing the log if the data is corrupt so that we don't end up with half a log

Changes since v6:
- Added Linaro copyright for all the EFI moved code
- Changed tcg2_init_log (and by extension, tcg2_measurement_init) to copy any discovered event log to the user's log if passed in.

Changes since v5:
- Remove unused platform_get_eventlog in efi_tcg2.c
- First look for tpm_event_log_* properties instead of linux,sml-*
- Fix efi_tcg2.c compilation
- Select SHA* configs

Changes since v4:
- Remove tcg2_measure_event function and check for NULL data in tcg2_measure_data
- Use tpm_auto_startup
- Fix efi_tcg2.c compilation for removing tcg2_pcr_read function

Changes since v3:
- Reordered headers
- Refactored more of EFI code into common code
Removed digest_info structure and instead used the common alg_to_mask
and alg_to_len
Improved event log parsing in common code to get it equivalent to EFI
Common code now extends PCR if previous bootloader stage couldn't No need to allocate memory in the common code, so EFI copies the discovered buffer like it did before
Rename efi measure_event function

Changes since v1:
- Refactor TPM layer functions to allow EFI system to use them, and remove duplicate EFI functions

Quoted 6 lines
 include/efi_tcg2.h        |   44 --
 include/tpm-v2.h          |  259 +++++++++
 lib/Kconfig               |    4 +
 lib/efi_loader/efi_tcg2.c | 1054 +++----------------------------------
 lib/tpm-v2.c              |  814 ++++++++++++++++++++++++++++
 5 files changed, 1154 insertions(+), 1021 deletions(-)

This is adding a lot more code to the TPM implementation and I see you have added more tests. But do we have enough unit testing in test/dm/tpm.c for these new functions?

Regards,
Simon