Add TPM2 RNG #4117

atreiber94 · 2024-06-12T12:51:50Z

This PR is for #3877 regarding the RNG integration of TPM2. It already includes FFI and Python integration.

For current progress on the entire TPM2 integration beyond RNG check #3877.

There's no need to merge this for 3.5.0 (this API may still possibly be subject to changes within broader TPM2 integration of #3877).

Setup with `swtpm`

To emulate the TPM2, we use swtpm. The following steps are necessary to setup swtpmto work with our implementation:

Install tpm2-tools, libtss2-dev and swtpm: sudo apt install tpm2-tools libtss2-dev swtpm
Configure Botan --with-tpm2 and build
mkdir /tmp/myvtpm
swtpm socket --tpmstate dir=/tmp/myvtpm --tpm2 --ctrl type=tcp,port=2322 --server type=tcp,port=2321 --flags not-need-init (terminal has to stay open)
tpm2_startup -c

By default, Botan's new TPM2_Context will use the TPM2-TSS default strategy for identifying the TCTI. To directly use swtpm or another TCTI, pass the corresponding TCTI name configuration to Botan::TPM2_Context::create.

Python example (because that's what we need)

import botan3 as botan
from binascii import hexlify

tpm2_ctx = botan.TPM2Context("swtpm")
tpm2_rng = botan.RandomNumberGenerator("tpm2", tpm2_ctx)

print(hexlify(tpm2_rng.get(32)).decode("utf-8"))

ToDos

CI Integration
Documentation

Co-Authored-By: René Meusel <[email protected]>

coveralls · 2024-06-12T13:27:13Z

coverage: 91.777% (+0.002%) from 91.775%
when pulling bda3913 on Rohde-Schwarz:feature/tpm2
into 927aab8 on randombit:master.

coveralls · 2024-06-12T15:12:39Z

coverage: 91.77% (-0.005%) from 91.775%
when pulling 9890bfa on Rohde-Schwarz:feature/tpm2
into 927aab8 on randombit:master.

reneme · 2024-06-12T20:35:14Z

src/lib/prov/tpm2/tpm2_rng.cpp

+      BOTAN_ASSERT_NOMSG(digest->size == requested_bytes);
+      out.append({digest->buffer, digest->size});
+
+      Esys_Free(digest);


Might be a good idea to use scoped_cleanup (or some variant of it) for this. Its unlikely, but if either the assert or the append above throw, the digest won't be freed.

reneme · 2024-06-12T20:38:18Z

src/python/botan3.py

+#
+class TPM2Context:
+    def __init__(self, tcti_nameconf: str = None) -> None:
+        tcti = c_char_p(0 if not tcti_nameconf else tcti_nameconf.encode("utf-8"))


Other locations (for instance the constructor of RandomNumberGenerator) use a _ctype_str() helper to transform a string object into a (presumable) const char*. Perhaps we should use that here as well? Let's look into it once more...

reneme · 2024-06-12T20:39:09Z

src/python/botan3.py

+
+        self.__obj = c_void_p(0)
+        rc = _DLL.botan_tpm2_ctx_init(byref(self.__obj), tcti)
+        if rc == -40: # 'Not Implemented'


Any chance we could pull this magic number from the enum? 😅
(there are other places that would benefit from it)

Agreed, though it seems like this is hard to do since we can't access the enum via CDLL. The other options of manually keeping the same enum in Python or providing functions that return respective enum values don't seem particularly worthwhile to me.

Mhh, there seems to be some secret sauce around the translation of parameter/result types from C to Python and back. This blog post provides a little inspiration: https://v4.chriskrycho.com/2015/ctypes-structures-and-dll-exports.html

Perhaps we could actually mirror the error-enum into botan3.py. That's a manual duplication, but should be okay, given that this enum probably doesn't change all the time. And then we could make CDLL do the translation for us.

Any of this should certainly not go into this pull request, though.

randombit · 2024-06-12T22:43:18Z

src/lib/ffi/ffi_tpm2.cpp

+         }
+      }();
+
+      ctx->ctx = Botan::TPM2_Context::create(std::move(tcti));


If an exception is throwing here due to for instance no hardware being available at runtime, the caller will get BOTAN_FFI_ERROR_EXCEPTION_THROWN which is not particularly helpful. Maybe catch this and then translate an error to something more specific?

Maybe even just reusing BOTAN_FFI_ERROR_NOT_IMPLEMENTED if the hardware is unavailable? IDK

Good point. I don't think reusing BOTAN_FFI_ERROR_NOT_IMPLEMENTED is all that helpful since the function would be implemented and TPM2 usage desired in that case (as the build is configured with --with-tpm2).

The exception that would be thrown here is TPM2_Error with error type ErrorType::TPMError. I added a corresponding FFI error type such that the caller will now get that error translated within ffi_guard_thunk instead of BOTAN_FFI_ERROR_EXCEPTION_THROWN. That way, one can distinguish between a runtime TPM error when configured with TPM and a not implemented error when TPM is not configured.

src/lib/prov/tpm2/tpm2.h

src/lib/prov/tpm2/tpm2.cpp

randombit · 2024-06-12T22:54:53Z

src/lib/prov/tpm2/tpm2.h

+
+#include <tss2/tss2_esys.h>
+#include <tss2/tss2_rc.h>
+#include <tss2/tss2_tctildr.h>


I'm guessing that most/all uses of TPM will just have a shared_ptr<TPM2_Context> that they invoke various functions from the library with. That being the case, if we could hide these few uses of the TPM2 types in this header, we likely could avoid pulling the TPM library includes into anything a user might include.

src/lib/prov/tpm2/tpm2.h

src/lib/prov/tpm2/tpm2_rng.cpp

atreiber94 · 2024-06-13T14:47:03Z

Thanks for the comments! I applied review suggestions except the parts that require re-structuring tpm2.h, which I will pick up later on. As mentioned, we do not intend to merge this soon (not for 3.5.0)

coveralls · 2024-06-13T15:12:43Z

coverage: 91.768% (-0.007%) from 91.775%
when pulling ed26467 on Rohde-Schwarz:feature/tpm2
into 927aab8 on randombit:master.

reneme · 2024-06-14T09:49:47Z

doc/api_ref/ffi.rst

@@ -177,6 +177,10 @@ The following enum values are defined in the FFI header:
   calling :cpp:func:`botan_hash_destroy` on a ``botan_rng_t`` object will cause
   this error.

+.. cpp:enumerator:: BOTAN_FFI_TPM_ERROR = -78


Suggested change

.. cpp:enumerator:: BOTAN_FFI_TPM_ERROR = -78

.. cpp:enumerator:: BOTAN_FFI_TPM2_ERROR = -78

... we've been specific basically everywhere. I think we should always state TPM2.

See #4117 (comment) - this FFI error translates the ErrorType::TPMError, which is also the error type for TPM2_Error. Or should we just introduce a new TPM2 ErrorType and keep them separated entirely?

src/lib/prov/tpm2/tpm2.cpp

src/lib/prov/tpm2/tpm2.h

coveralls · 2024-06-14T10:23:39Z

coverage: 91.763% (-0.01%) from 91.775%
when pulling 2ab6008 on Rohde-Schwarz:feature/tpm2
into 927aab8 on randombit:master.

coveralls · 2024-06-14T12:38:43Z

coverage: 91.772% (-0.003%) from 91.775%
when pulling 70be246 on Rohde-Schwarz:feature/tpm2
into 927aab8 on randombit:master.

src/lib/prov/tpm2/tpm2.h

src/lib/prov/tpm2/tpm2_rng.cpp

…ntext_object() and add a helper function to safely cast to ESYS_CONTEXT*.

coveralls · 2024-07-01T09:46:39Z

coverage: 91.731% (-0.04%) from 91.775%
when pulling 6e2af38 on Rohde-Schwarz:feature/tpm2
into 927aab8 on randombit:master.

atreiber94 and others added 4 commits June 10, 2024 16:47

Add TPM2 Context and tpm2_rng

6109b2b

Co-Authored-By: René Meusel <[email protected]>

Add FFI support for TPM2 RNG

5250131

Co-Authored-By: René Meusel <[email protected]>

Add TPM2 RNG support for Python

e613f07

Co-Authored-By: René Meusel <[email protected]>

Include TCTI configuration into TPM2_Context

c205858

Co-Authored-By: René Meusel <[email protected]>

atreiber94 added the enhancement Enhancement or new feature label Jun 12, 2024

atreiber94 added this to the Botan 3.6.0 milestone Jun 12, 2024

atreiber94 self-assigned this Jun 12, 2024

atreiber94 mentioned this pull request Jun 12, 2024

Future TPM support #3877

Open

Fix CI

bda3913

atreiber94 added 5 commits June 12, 2024 15:28

Fix Python linting issues

78f1bee

Add CI support for TPM2/swtpm

1cd900b

Fix TPM2_Context vs TPm2Context in botan3.py

c6a3a02

Extend CI for TPM2

d1244e3

Fix Memory Leak in C FFI Test

9890bfa

atreiber94 marked this pull request as ready for review June 12, 2024 15:23

reneme reviewed Jun 12, 2024

View reviewed changes

randombit self-requested a review June 12, 2024 22:35

randombit reviewed Jun 12, 2024

View reviewed changes

Apply review suggestions except tpm2.h restructuring

ed26467

Restructure tpm2.h to avoid external includes in public headers

2ab6008

reneme reviewed Jun 14, 2024

View reviewed changes

Address comments regarding tpm2.h restructuring

70be246

randombit reviewed Jun 20, 2024

View reviewed changes

src/lib/prov/tpm2/tpm2.h Outdated Show resolved Hide resolved

randombit reviewed Jun 29, 2024

View reviewed changes

src/lib/prov/tpm2/tpm2_rng.cpp Outdated Show resolved Hide resolved

Review comments: Rename TPM2_Context::get() to TPM2_Context::inner_co…

6e2af38

…ntext_object() and add a helper function to safely cast to ESYS_CONTEXT*.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add TPM2 RNG #4117

Add TPM2 RNG #4117

atreiber94 commented Jun 12, 2024 •

edited

Loading

coveralls commented Jun 12, 2024

coveralls commented Jun 12, 2024

reneme Jun 12, 2024

reneme Jun 12, 2024

reneme Jun 12, 2024

atreiber94 Jun 13, 2024

reneme Jun 13, 2024

randombit Jun 12, 2024

randombit Jun 12, 2024

atreiber94 Jun 13, 2024

randombit Jun 12, 2024

atreiber94 commented Jun 13, 2024 •

edited

Loading

coveralls commented Jun 13, 2024

reneme Jun 14, 2024

atreiber94 Jun 14, 2024

coveralls commented Jun 14, 2024

coveralls commented Jun 14, 2024

coveralls commented Jul 1, 2024

	.. cpp:enumerator:: BOTAN_FFI_TPM_ERROR = -78
	.. cpp:enumerator:: BOTAN_FFI_TPM2_ERROR = -78

Add TPM2 RNG #4117

Are you sure you want to change the base?

Add TPM2 RNG #4117

Conversation

atreiber94 commented Jun 12, 2024 • edited Loading

Setup with swtpm

Python example (because that's what we need)

ToDos

coveralls commented Jun 12, 2024

coveralls commented Jun 12, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

atreiber94 commented Jun 13, 2024 • edited Loading

coveralls commented Jun 13, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

coveralls commented Jun 14, 2024

coveralls commented Jun 14, 2024

coveralls commented Jul 1, 2024

atreiber94 commented Jun 12, 2024 •

edited

Loading

Setup with `swtpm`

atreiber94 commented Jun 13, 2024 •

edited

Loading