There are currently 4 tests in ssl-opt.sh with either -C "resend" or -S
"resend", that is, asserting that no retransmission will occur. They sometimes
fail on loaded CI machines as one side doesn't send a message fast enough,
causing the other side to retransmit, causing the test to fail.
(For the "reconnect" test there was an other issue causing random failures,
fixed in a previous commit, but even after that fix the test would still
sometimes randomly fail, even if much more rarely.)
While it's a hard problem to fix in a general and perfect way, in practice the
probability of failures can be drastically reduced by making the timeout
values much larger.
For some tests, where retransmissions are actually expected, this would have
the negative effect of increasing the average running time of the test, as
each side would wait for longer before it starts retransmission, so we have a
trade-off between average running time and probability of spurious failures.
But for tests where retransmission is not expected, there is no such trade-off
as the expected running time of the test (assuming the code is correct most of
the time) is not impacted by the timeout value. So the only negative effect of
increasing the timeout value is on the worst-case running time on the test,
which is much less important, as test should only fail quite rarely.
This commit addresses the easy case of tests that don't expect retransmission
by increasing the value of their timeout range to 10s-20s. This value
corresponds to the value used for tests that assert `-S "autoreduction"` which
are in the same case and where the current value seems acceptable so far.
It also represents an increase, compared to the values before this commit, of
a factor 20 for the "reconnect" tests which were frequently observed to fail
in the CI, and of a factor 10 for the first two "DTLS proxy" tests, which were
observed to fail much less frequently, so hopefully the new values are enough
to reduce the probability of spurious failures to an acceptable level.
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
The server must check client reachability (we chose to do that by checking a
cookie) before destroying the existing association (RFC 6347 section 4.2.8).
Let's make sure we do, by having a proxy-in-the-middle inject a ClientHello -
the server should notice, but not destroy the connection.
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
Section 4.2.8 of RFC 6347 describes how to handle the case of a DTLS client
establishing a new connection using the same UDP quartet as an already active
connection, which we implement under the compile option
MBEDTLS_SSL_DLTS_CLIENT_PORT_REUSE. Relevant excerpts:
[the server] MUST NOT destroy the existing
association until the client has demonstrated reachability either by
completing a cookie exchange or by completing a complete handshake
including delivering a verifiable Finished message.
[...]
The reachability requirement prevents
off-path/blind attackers from destroying associations merely by
sending forged ClientHellos.
Our code chooses to use a cookie exchange for establishing reachability, but
unfortunately that check was effectively removed in a recent refactoring,
which changed what value ssl_handle_possible_reconnect() needs to return in
order for ssl_get_next_record() (introduced in that refactoring) to take the
proper action. Unfortunately, in addition to changing the value, the
refactoring also changed a return statement to an assignment to the ret
variable, causing the function to reach the code for a valid cookie, which
immediately destroys the existing association, effectively bypassing the
cookie verification.
This commit fixes that by immediately returning after sending a
HelloVerifyRequest when a ClientHello without a valid cookie is found. It also
updates the description of the function to reflect the new return value
convention (the refactoring updated the code but not the documentation).
The commit that changed the return value convention (and introduced the bug)
is 2fddd3765e, whose commit message explains the
change.
Note: this bug also indirectly caused the ssl-opt.sh test case "DTLS client
reconnect from same port: reconnect" to occasionally fail due to a race
condition between the reception of the ClientHello carrying a valid cookie and
the closure of the connection by the server after noticing the ClientHello
didn't carry a valid cookie after it incorrectly destroyed the previous
connection, that could cause that ClientHello to be invisible to the server
(if that message reaches the server just before it does `net_close()`). A
welcome side effect of this commit is to remove that race condition, as the
new connection will immediately start with a ClientHello carrying a valid
cookie in the SSL input buffer, so the server will not call `net_close()` and
not risk discarding a better ClientHello that arrived in the meantime.
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
All the content that is currently in ChangeLog.md is the result of
merging Mbed Crypto, and that content is already present in
ChangeLog. Since we aren't using ChangeLog.md yet, remove it for now.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Conflicts:
* .github/issue_template.md, .github/pull_request_template.md:
The changes in Mbed Crypto are not relevant to Mbed TLS. Keep the
mbedtls versions.
- "Default" should only be used for tests that actually use the defaults (ie,
not passing options on the command line, except maybe debug/dtls)
- All tests in the "Encrypt then MAC" group should start with that string as a
common prefix
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
Merge the latest state of the target branch (mbedtls/development) into the
pull request to merge mbed-crypto into mbedtls.
Conflicts:
* ChangeLog: add/add conflict. Resolve by using the usual section order.
In Windows cmd, `del foo` succeeds if `foo` doesn't exist, but
`del *.ext` fails if `*.ext` doesn't match any files. Therefore we use
the idiom `if exist *.ext del *.ext` to delete files matching
wildcards.
We've accidentally used `if exist $(some_list) del $(some_list)`, and
that's wrong, because it's only syntactically correct if
`$(some_list)` contains exactly one element. As long as `$(some_list)`
contains actual file names and not wildcards, just use `del $(some_list)`.
Add dependencies on !MBEDTLS_SHA512_NO_SHA384 to X.509 and SSL unit
tests that use SHA-384 (identified based on having a description that
contains "SHA384" or "SHA-384").
Rename identifiers containing double-underscore (`__`) to avoid `__`.
The reason to avoid double-underscore is that all identifiers
containing double-underscore are reserved in C++. Rename all such
identifiers that appear in any public header, including ssl_internal.h
which is in principle private but in practice is installed with the
public headers.
This commit makes check-names.sh pass.
```
perl -i -pe 's/\bMBEDTLS_SSL__ECP_RESTARTABLE\b/MBEDTLS_SSL_ECP_RESTARTABLE_ENABLED/g; s/\bMBEDTLS_KEY_EXCHANGE_(_\w+)_(_\w+)\b/MBEDTLS_KEY_EXCHANGE${1}${2}/g' include/mbedtls/*.h library/*.c programs/*/*.c scripts/data_files/rename-1.3-2.0.txt tests/suites/*.function
```
Add a changelog entry for an already-released version to indicate when
the crypto submodule became mandatory.
Add a changelog entry for the removal of the crypto submodule.
Remove code guarded by `USE_CRYPTO_SUBMODULE`. It's dead now that
crypto can no longer be a submodule.
In `library/Makefile`:
* Replace `$(CRYPTO_INCLUDE)` with the single include directory
`-I../include`.
* Remove references to `$(OBJS_CRYPTO)` when it's in addition to the
local objects (`*.o`) since `$(OBJS_CRYPTO)` is now a subset of the
local objects.
* Merge modules that were duplicated between the mbedtls and the
mbed-crypto repositories back into the single list for `OBJS_CRYPTO`.
Look for changes that remove X509/SSL functionality.
```
git diff 'HEAD^{/^Merge}~1' HEAD --diff-filter=M -- . ':!library/error.c' ':!library/version_features.c' ':!programs/test/query_config.c' ':!visualc' ':!*.pdf' ':!*.der' | grep -E "^-.*MBEDTLS_(ERR_)?(PKCS11|X509|NET|SSL)_"
```
All of these removals are in `config.h` or `check_config.h`. Selectively revert the differences in these two files.
```
git diff 'HEAD^{/^Merge}~1' 'HEAD^{/^Merge}' include/mbedtls/config.h include/mbedtls/check_config.h | git apply -p1 -R
```
* `include/mbedtls/check_config.h`:
* ARIA for GCM: don't remove it.
* `MBEDTLS_PSA_CRYPTO_SE_C`: don't remove it.
* `MBEDTLS_SHA512_NO_SHA384`: don't remove it.
* `MBEDTLS_SSL_DTLS_CONNECTION_ID`: restore it.
* `include/mbedtls/config.h`:
* warning about `MBEDTLS_ECDSA_SIGN_ALT`: don't remove it.
* `MBEDTLS_CTR_DRBG_USE_128_BIT_KEY` full paragraph: don't remove it.
* `MBEDTLS_PSA_CRYPTO_KEY_FILE_ID_ENCODES_OWNER`: don't remove it.
* `MBEDTLS_SHA512_SMALLER`…: don't remove it.
* `MBEDTLS_SSL_RECORD_CHECKING`: restore it.
* `MBEDTLS_SSL_CONTEXT_SERIALIZATION`: restore it.
* `MBEDTLS_USE_PSA_CRYPTO` note: don't restore the tls version.
* `MBEDTLS_USE_PSA_CRYPTO` warning: restore the tls version.
* `MBEDTLS_CMAC_C`: restore it to being disabled by default. It's a minor API change in Mbed TLS because it changes the layout of `mbedtls_cipher_context_t`.
* `MBEDTLS_CTR_DRBG_C`: don't restore the older version of the description from tls.
* `MBEDTLS_GCM_C`: don't restore the older description from tls.
* `MBEDTLS_PSA_CRYPTO_C`: don't restore `crypto/`.
* `MBEDTLS_PSA_CRYPTO_SE_C`: don't remove it.
* `MBEDTLS_PSA_CRYPTO_STORAGE_C`: don't restore `crypto/`. Don't disable it by default.
* `MBEDTLS_PSA_ITS_FILE_C`: don't restore, like for ``MBEDTLS_PSA_CRYPTO_STORAGE_C``.
* `MBEDTLS_CTR_DRBG_USE_128_BIT_KEY` single line: don't restore it since there is now a full paragraph in the proper section above.
* `MBEDTLS_SSL_CID_IN_LEN_MAX`…: restore it.
* `MBEDTLS_TLS_DEFAULT_ALLOW_SHA1_IN_CERTIFICATES`: restore the version without a space to format the line like the others.
Look for any other invocation of test scripts that was removed: look for a change that removes a line (`^-.*…`) containing one of the names of a test script (without its path because it may be used with a relative path). Look for `ssl-opt.sh` and `compat.sh`, as well any file in `tests/scripts` that only exists in tls.
```
git diff 'HEAD^{/^Merge}~1' HEAD --diff-filter=M -- . ':!library/error.c' ':!library/version_features.c' ':!programs/test/query_config.c' ':!visualc' ':!*.pdf' ':!*.der' | grep -E "^-.*($(comm -23 <(git ls-tree -r --name-only 'HEAD^{/^Merge}~1') <(git ls-tree -r --name-only $(git merge-base upstream-crypto/development 'HEAD^{/^Merge}^2')) | sed -n 's!^tests/scripts/!!p' | sed 's/\./\\./g' | tr '\n' '|')ssl-opt\.sh|compat\.sh)" | grep -v '^---'
```
This only turns up changes in `basic-in-docker.sh`.
The merge of mbed-crypto removed some tls coverage. Restore it. Also
remove references to the `crypto` subdirectory brought by the mbedtls
side of the merge. In more detail:
* `tests/scripts/all.sh`:
* `fuzz` in comments (×2): restore it.
* `CTEST_OUTPUT_ON_FAILURE=1`: don't remove it.
* `cd crypto` for `make clean`: don't restore it.
* `cleanup`: do restore `programs/fuzz/Makefile`. Don't go into `crypto`. Keep only one copy of the calls to `rm` in `cmake_subproject`.
* Comment legacy options: don't remove it.
* `crypto/Makefile` and `pre_check_seedfile`: don't restore either. See below regarding the lack of need for `pre_check_seedfile`.
* blank line in `pre_print_configuration`: restore it.
* blank line before `#### Build and test`: restore it.
* SSL tests in `component_test_full_cmake_gcc_asan` and zlib components: restore it.
* `component_test_no_pem_no_fs` (×2): the merge placed two copies in different locations. Reconcile them: unset PSA storage like in crypto, and call `ssl-opt.sh` like in tls. Put the merged version at the tls location.
* `component_test_everest`: do add it at the tls location.
* `component_test_small_mbedtls_ssl_dtls_max_buffering`: restore the tls value.
* `component_test_new_ecdh_context`…: move `component_test_new_ecdh_context` before `component_test_everest` and add a calls to `compat.sh` and `ssl-opt.sh` like in `component_test_everest`. Remove the redundant crypto-only `component_test_everest`. Don't remove `component_test_psa_collect_statuses`.
* `component_test_full_cmake_clang`: don't remove `clang` in the `msg` call. Don't remove the call to `test_psa_constant_names.py`.
* `component_test_full_make_gcc_o0`: remove it. It's subsumed by `component_test_gcc_opt`.
* `component_build_deprecated`: don't remove anything.
* `component_test_memory_buffer_allocator`: restore `ssl-opt.sh`.
* `component_test_when_no_ciphersuites_have_mac`: restore it.
* `component_test_platform_calloc_macro`: don't restore `unset MBEDTLS_MEMORY_BUFFER_ALLOC_C` which is now redundant. Don't restore explicit flags instead of `$ASAN_CFLAGS`.
* `component_test_aes_fewer_tables`…: don't remove it.
* `component_test_m32_o1`: restore SSL testing.
* `component_test_m32_everest`: restore SSL testing.
* `component_test_min_mpi_window_size`…: don't remove it.
* `component_test_valgrind`: do restore the tls version of the comment.
* `run_component`: don't remove the seedfile creation. This is better than `pre_check_seedfile` (see below).
* `pre_check_seedfile`: don't restore it. `pre_check_seedfile` (from tls) creates a seedfile once and for all. This is not good enough if a component fails in such a way as to leave a broken seedfile, or if a component leaves a seedfile with a size that's wrong for the next component to run. Instead (from crypto), `run_component` creates a sufficiently large seedfile before each component.
Merge `unremove-non-crypto` into `mbedtls/development`. The branch
`unremove-non-crypto` was obtained by starting from `mbed-crypto/development`,
then reverting many commits that removed X.509 and TLS functionality when Mbed
Crypto forked from Mbed TLS (the “unremoval”), then make a few tweaks to
facilitate the merge.
The unremoval step restored old versions of some tls files. If a file doesn't
exist in mbed-crypto, check out the mbedtls version, regardless of what
happened during the unremoval of tls files in the crypto tree. Also
unconditionally take the mbedtls version of a few files where the
modifications are completely project-specific and are not relevant in
mbed-crypto:
* `.github/issue_template.md`: completely different. We may want to reconcile
them independently as a follow-up.
* `.travis.yml`: would only be reverted to an earlier tls version.
* `README.md`: completely different. We may want to reconcile them
independently as a follow-up.
* `doxygen/input/doc_mainpage.h`: the changes in crypto were minimal and not
relevant except as a stopgap as mbed-crypto did not have its own product
versioning in the Doxygen documentation.
* `tests/.jenkins/Jenkinsfile`: completely different.
* `tests/data_files/Makefile`: there were no changes in mbed-crypto,
but the unremoval step restored an old version.
Shell script for everything to do after the merge apart from the conflict
resolution:
```
tls_files=($(comm -23 <(git ls-tree -r --name-only HEAD) <(git ls-tree -r --name-only $(git merge-base upstream-crypto/development MERGE_HEAD))))
tls_files+=($tls_files .github/issue_template.md .travis.yml README.md doxygen/input/doc_mainpage.h tests/.jenkins/Jenkinsfile tests/data_files/Makefile)
git checkout --theirs HEAD -- $tls_files
git add -- $tls_files
```
Resolve the remaining conflicts:
* `library/CMakeLists.txt`:
* Keep the TLS definition of `src_crypto`
* `USE_SHARED_MBEDTLS_LIBRARY`: keep all three libraries, with both
`include` and `crypto/include` in `target_include_directories`, all with
version `2.21.0`.
* `programs/Makefile`:
* Reconcile the APPS lists (add/add from a differently-formatted common
ancestor): insert the `psa/*` from crypto into the tls list.
* Keep the `fuzz` target defined only in tls version.
* Keep the recipe (only in tls version) cleaning `ssl_pthread_server`
stuff for the `clean` target.
* `scripts/config.py`:
* `include_in_full`: add/add conflict. Keep both.
* `tests/scripts/all.sh`:
* `component_test_no_use_psa_crypto_full_cmake_asan`: partially old
version in crypto. Take the tls version.
* `component_test_malloc_0_null` and more: take
`component_test_malloc_0_null` from crypto (with `config.py` rather than
`config.pl`, and with `$ASAN_FLAGS` rather than an explicit list), but
add the call to `ssl-opt.sh` from tls. Take the other components from
crypto.
With this commit, building and running the unit tests with both `make ` and
`cmake` work in the default configuration on Linux. Other platforms, build
systems and configurations are likely not to work, and there is some
regression in test coverage.
There is some loss of functionality because the unremoval step restored older
versions of tls content. This commit contains the latest tls version of
tls-only files, but some changes from the tls side in files that existed on
both sides have regressed. Most problematic changes are hunks that remove some
tls-specific feature and contain either a C preprocessor symbol identifying a
tls-specific module or option, or the name of a tls-specific file. Hunks
that remove a tls-specific preprocessor symbol can be identified with the
regular expression `^-.*MBEDTLS_(ERR_)?(PKCS11|X509|NET|SSL)_`.
Subsequent commits will revert a few parts of the patch from this merge commit
in order to restore the tls functionality that it removes, ensure that the
test coverage includes what was covered in either branch, and fix test
failures.
Even if other higher-level libraries were added, these programs would
only link with the crypto library, which is the one that contains
platform functions.
This reverts commit bea98b4581.
Conflicts:
* programs/Makefile:
* APPS: the layout of the definition has changed. re-add dh_client
and dh_server appropriately.
Run scripts/generate_visualc_files.pl to account for the added programs.
