Overview

This repository hosts source code implementing the Trusted Computing Group's (TCG) TPM2 Software Stack (TSS). This stack consists of the following layers from top to bottom:

[1] We currently deviate from the specification by increasing the value of TPM2_NUM_PCR_BANKS from 3 to 16 to ensure compatibility with TPM2 implementations that have enabled a larger than typical number of PCR banks. This is expected to be included in a future revision of the specification.

Build and Installation Instructions:

Instructions to build and install tpm2-tss are available in the INSTALL file.

Getting in Touch:

If you're looking to discuss the source code in this project or get some questions answered you should join the TPM2 mailing list:

• https://lore.kernel.org/tpm2/
• To subscribe write an email to tpm2+subscribe@lists.linux.dev see also here
• The old list https://lists.linuxfoundation.org/mailman/listinfo/tpm2 was decomissioned by Linux

We also have an IRC channel set up on FreeNode called #tpm2.0-tss. You can also try Gitter

You can join a weekly online call at TPM.dev, where we are discussing the tpm2-tss stack, the tpm2-pkcs11 project and other Linux TPM2 & TSS2-Software.

In case you want to contribute to the project, please also have a look at the Contribution Guidelines.

Documentation

The doxygen documentation can either be built by oneself (see the INSTALL file) or browsed directly on tpm2-tss.readthedocs.io.

Test Suite

This repository contains a test suite intended to exercise the TCTI, SAPI and ESAPI code. This test suite is not intended to test a TPM implementation, so this test suite should only be run against a TPM simulator. If this test suite is executed against a TPM other than the software simulator it may cause damage to the TPM (NV storage wear out, etc.). You have been warned.

Simulator

The TPM library specification contains reference code sufficient to construct a software TPM 2.0 simulator. This code was provided by Microsoft and they provide a binary download for Windows here.

There are two implementations that enable building and running this code on Linux. Issues building or running the simulator should be reported to respective project.

Software TPM

The Software TPM is an open-source TPM emulator with different front-end interfaces such as socket and character device. Its code is hosted on GitHub and building is faciliated by the GNU Autotools. The TCTI module for using this simulator is called swtpm.

Since tpm2-tss v3.0 swtpm is the default simulator used by this project.

IBM's Software Simulator

IBM has also repackaged this code with a few Makefiles so that the Microsoft code can be built and run on Linux systems. The Linux version of the Microsoft TPM 2.0 simulator can be obtained on SourceForge. Once you've downloaded and successfully built and execute the simulator it will, by default, be accepting connections on the localhost, TCP ports 2321 and 2322. The TCTI module for using this simulator is called mssim.

Testing

To test the various TCTI, SAPI and ESAPI api calls, unit and integration tests can be run by configuring the build to enable unit testing and running the "check" build target. It is recommended to use a simulator for testing, and the simulator will be automatically launched by the tests. Please review the dependency list in INSTALL for dependencies when building the test suite.

$ ./configure --enable-unit --enable-integration
$ make -j$(nproc) check

This will generate a file called "test-suite.log" in the root of the build directory.

Please report failures in a Github 'issue' with a full log of the test run.

NOTE: The unit and integration tests can be enabled independently. The --enable-unit option controls unit tests, and --enable-integration controls the integration tests.

Running tests on physical TPM device

To run integration tests on a physical TPM device, including a TPM hardware or a software TPM implemented in platform firmware the configure script provides two options. The first option is called --with-device and it is used to point to the TPM device interface exposed by the OS, for example:

  $ ./configure  --with-device=/dev/tpm0

The second option, --with-devicetests, enables a "class" of test. There are three classes:

1. destructive - these tests can affect TPM capability or lifespan
2. mandatory - these tests check all the functionality that is mandatory per the TCG specification (default).
3. optional - these tests are for functionality that is optional per the TCG specification.

For example to enable both mandatory and optional test cases during configure one needs to set this flag as follows:

  $ ./configure --with-devicetests="mandatory,optional"

Tht default value for the flag is "mandatory" Any combination of the three is valid. The two flags are only valid when the integration tests are enabled with --enable-integration flag.

After that the following command is used to run the test on the configured TPM device:

  $ sudo make check-device

or

  $ sudo make check -j 1

Note: The tests can not be run in paralel.

Running valgrind check

The unit and integration tests can be run under the valgrind tool, which performs additional checks on the library and test code, such as memory leak checks etc. The following command is used to run the tests under valgrind:

$ make check-valgrind

This command will enable all valgrind "tools" and kick off as many test as they support. It is possible to enable different valgrind tools (checks) in more granularity. This can be controlled by invoking different tools separately using check-valgrind-<tool>, for instance:

  $ make check-valgrind-memcheck

or

  $ make check-valgrind-drd

Currently the the following tools are supported:

memcheck - Performs memory related checks. This is the default tool. helgrind - Performs synchronization errors checks. drd - Performs thread related checks. sgcheck - Performs stack overrun related checks.

Note that the valgring tool can also be invoked manually using the standard libtool:

  $ libtool exec valgrind --tool=memcheck --leak-check=full \
    test/integration/esys-auto-session-flags.int

This allows for more control on what checks are performed.

Logging

While investigating issues it might be helpful to enable extra debug/trace output. It can be enabled separately for different components. The description how to do this can be found in the logging file.

Fuzzing

All system API function calls can be tested using a fuzzing library. The description how to do this can be found in the fuzzing file.

Architecture/Block Diagram

SAPI library, TAB/RM, and Test Code Block Diagram:

Project Layout

|-- doc     : various bits of documentation\
|-- include : header files installed in $(includedir)\
|   +-- tss2      : all public headers for this project\
|-- lib     : data files used by the build or installed into $(libdir)\
|-- m4      : autoconf support macros\
|-- man     : man pages\
|-- script  : scripts used by the build or CI\
|-- src     : all source files\
|   |-- tss2-esys : enhanced system API (ESAPI) implementation\
|   |   +-- api   : ESAPI TPM API implementation\
|   |-- tss2-mu   : TPM2 type marshaling/unmarshaling (MU) API implementation\
|   |-- tss2-sys  : system API (SAPI) implementation\
|   |   +-- api   : SAPI public API implementation\
|   |-- tss2-tcti : TCTI implementations for device and mssim\
|   +-- util      : Internal utility library (e.g. logging framework)\
+-- test    : test code\
    |-- integration : integration test harness and test cases\
    |-- tpmclient   : monolithic, legacy test application\
    +-- unit        : unit tests