Commit graph

7082 commits

Author SHA1 Message Date
Angus Gratton d8213d00db Let MBEDTLS_SSL_MAX_CONTENT_LEN to be split into outward & inward sizes
For the situation where the mbedTLS device has limited RAM, but the
other end of the connection doesn't support the max_fragment_length
extension. To be spec-compliant, mbedTLS has to keep a 16384 byte
incoming buffer. However the outgoing buffer can be made smaller without
breaking spec compliance, and we save some RAM.

See comments in include/mbedtls/config.h for some more details.

(The lower limit of outgoing buffer size is the buffer size used during
handshake/cert negotiation. As the handshake is half-duplex it might
even be possible to store this data in the "incoming" buffer during the
handshake, which would save even more RAM - but it would also be a lot
hackier and error-prone. I didn't really explore this possibility, but
thought I'd mention it here in case someone sees this later on a mission
to jam mbedTLS into an even tinier RAM footprint.)
2018-06-18 20:51:51 +10:00
Simon Butcher b6a5bffc4b Add ChangeLog entry for clang version fix. Issue #1072 2018-06-18 11:51:36 +01:00
Manuel Pégourié-Gonnard c7bc9e122f Fix a few typos 2018-06-18 10:30:30 +02:00
Simon Butcher cbe248a3f5 Merge remote-tracking branch 'public/pr/1727' into development 2018-06-17 17:37:29 +01:00
Simon Butcher 1f91575546 Merge remote-tracking branch 'public/pr/1681' into development 2018-06-17 17:35:54 +01:00
Simon Butcher 688a456f58 Merge remote-tracking branch 'public/pr/1734' into development 2018-06-17 17:35:34 +01:00
Simon Butcher cdc338bfd7 Merge remote-tracking branch 'public/pr/1735' into development 2018-06-17 17:35:15 +01:00
Simon Butcher 2711ad7505 Merge remote-tracking branch 'public/pr/1736' into development 2018-06-17 17:34:55 +01:00
Simon Butcher 600c5e6d20 Compilation warning fixes on 32b platfrom with IAR
Fix compilation warnings with IAR toolchain, on 32 bit platform.
Reported by rahmanih in #683

This is based on work by Ron Eldor in PR #750, some of which was independently
fixed by Azim Khan and already merged in PR #1646.
2018-06-17 17:24:56 +01:00
Jaeden Amero d9c71dad50 Revert "Turn on MBEDTLS_SSL_ASYNC_PRIVATE by default"
This reverts commit 5ffe41c8f9.

Turn off MBEDTLS_SSL_ASYNC_PRIVATE by default because we don't want to
change the ABI yet.
2018-06-15 20:31:26 +01:00
Simon Butcher 2dbecc04cc Merge remote-tracking branch 'public/pr/1602' into development 2018-06-15 20:15:00 +01:00
Simon Butcher 66a8903f4a Fix for missing len var when XTS config'd and CTR not
The AES XTS self-test was using a variable len, which was declared only when CTR
was enabled. Changed the declaration of len to be conditional on CTR and XTS.
2018-06-15 18:20:29 +01:00
Simon Butcher 9873696c34 Merge remote-tracking branch 'public/pr/710' into development 2018-06-15 14:18:11 +01:00
Simon Butcher 5f57f1e3cc Merge remote-tracking branch 'public/pr/1270' into development 2018-06-15 14:17:31 +01:00
Simon Butcher 675590519a Merge remote-tracking branch 'public/pr/1602' into development 2018-06-15 14:13:14 +01:00
Gilles Peskine d5d983e168 ssl_server2: handle mbedtls_x509_dn_gets failure
If mbedtls_x509_dn_gets fails, the server could end up calling printf
on an uninitialized buffer. Check if the function succeeds. Found by
Coverity.
2018-06-15 14:05:10 +02:00
Gilles Peskine ace05929e8 Fix harmless use of uninitialized memory in ssl_parse_encrypted_pms
In ssl_parse_encrypted_pms, some operational failures from
ssl_decrypt_encrypted_pms lead to diff being set to a value that
depended on some uninitialized unsigned char and size_t values. This didn't
affect the behavior of the program (assuming an implementation with no
trap values for size_t) because all that matters is whether diff is 0,
but Valgrind rightfully complained about the use of uninitialized
memory. Behave nicely and initialize the offending memory.
2018-06-14 13:26:52 +02:00
Gilles Peskine c306a059d1 SSL async tests: add a few test cases for error in decrypt
The code paths in the library are different for decryption and for
signature. Improve the test coverage by doing some error path tests
for decryption in addition to signature.
2018-06-14 13:26:52 +02:00
Gilles Peskine 4481744538 Fix memory leak in ssl_server2 with SNI + async callback
In ssl_server2, the private key objects are normally local variables
of the main function. However this does not hold for private keys in
the SNI configuration. When async callbacks are used, the test code
transfers the ownership of the private keys to the async callbacks.
Therefore the test code must free the SNI private keys through the
async callbacks (but it must not free the straight private keys this
way since they are not even heap-allocated).
2018-06-14 13:26:52 +02:00
Gilles Peskine e247989061 SNI + SSL async callback: make all keys async
When testing async callbacks with SNI, make all the keys async, not
just the first one. Otherwise the test is fragile with respect to
whether a key is used directly or through the async callbacks.
2018-06-14 13:26:52 +02:00
Gilles Peskine 2636fade52 ssl_async_resume: free the operation context on error 2018-06-14 13:26:52 +02:00
Gilles Peskine f5a9996088 ssl_server2: get op_name from context in ssl_async_resume as well 2018-06-14 13:26:52 +02:00
Gilles Peskine ef30742a27 Clarify "as directed here" in SSL async callback documentation 2018-06-14 13:26:51 +02:00
Gilles Peskine 7457933a04 SSL async callbacks documentation: clarify resource cleanup
Clarify when resume must clean up resources and when cancel is called.
2018-06-14 13:26:51 +02:00
Gilles Peskine 3dae1cfa3a Async callback: use mbedtls_pk_check_pair to compare keys
In the current test code, the object that is used as a public key in
the certificate also contains a private key. However this is because
of the way the stest code is built and does not demonstrate the API in
a useful way. Use mbedtls_pk_check_pair, which is not what real-world
code would do (since the private key would typically be in an external
cryptoprocessor) but is a more representative placeholder.
2018-06-14 13:26:51 +02:00
Gilles Peskine a668c60186 Rename mbedtls_ssl_async_{get,set}_data for clarity
Rename to mbedtls_ssl_get_async_operation_data and
mbedtls_ssl_set_async_operation_data so that they're about
"async operation data" and not about some not-obvious "data".
2018-06-14 13:26:51 +02:00
Gilles Peskine 02b86d0415 Fix copypasta in the async callback documentation 2018-06-14 13:26:51 +02:00
Gilles Peskine 166ce748cf SSL async callback: cert is not always from mbedtls_ssl_conf_own_cert
The certificate passed to async callbacks may not be the one set by
mbedtls_ssl_conf_own_cert. For example, when using an SNI callback,
it's whatever the callback is using. Document this, and add a test
case (and code sample) with SNI.
2018-06-14 13:26:51 +02:00
Gilles Peskine d6fbfde994 ssl_async_set_key: detect if ctx->slots overflows 2018-06-14 13:26:51 +02:00
Gilles Peskine 07981585d3 Fix missing continuation indicator in ssl-opt.sh 2018-06-14 13:26:51 +02:00
Gilles Peskine 20e2bdf4b0 SSL async tests: tighten a few log checks in some test cases 2018-06-14 13:26:51 +02:00
Gilles Peskine 94e153af77 Improve documentation of the async callback's crypto parameters 2018-06-14 13:26:51 +02:00
Gilles Peskine 37d417561d Add test case for SSL async resume after resume
Add a test case for SSL asynchronous signature where f_async_resume is
called twice. Verify that f_async_sign_start is only called once.

This serves as a non-regression test for a bug where f_async_sign_start
was only called once, which turned out to be due to a stale build
artifacts with mismatched numerical values of
MBEDTLS_ERR_SSL_ASYNC_IN_PROGRESS.
2018-06-14 13:26:51 +02:00
Gilles Peskine 4d9ec4dcf7 Fix uninitialized variable in ssl_server2 2018-06-14 13:26:51 +02:00
Gilles Peskine 6331d78675 Don't use the printf format %zd
We target C89 libc, so don't use %zd or %zu. Just use %u, and make
slot numbers `unsigned` for simplicity.
2018-06-14 13:26:51 +02:00
Gilles Peskine 03af7f6ae7 Change boolean bitfield to unsigned
Reminder to self: 1 is not a valid value in a 1-bit bitfield. It's
undefined behavior and gcc -ansi -pedantic helpfully complains about it.
2018-06-14 13:26:51 +02:00
Jaeden Amero ff2f493432 config: List cipher modes in alphabetical order
Keeping the cipher modes list in alphabetical order makes it easier to
find things. Move OFB and XTS to their appropriate locations in the
list.
2018-06-14 11:42:27 +01:00
Simon Butcher 4ab8acd696 Merge remote-tracking branch 'public/pr/1075' into development 2018-06-14 10:34:44 +01:00
Simon Butcher 6ac1cf6f5f Merge remote-tracking branch 'public/pr/1182' into development 2018-06-14 10:33:29 +01:00
Simon Butcher e5cd86874e Add ChangeLog entry for PR #1555
Adds ChangeLog entry for PR #1555, 'Only redefine _WIN32_WINNT macro when <
0x0501'
2018-06-14 10:30:19 +01:00
Simon Butcher c1bf1aaee5 Merge remote-tracking branch 'public/pr/1555' into development 2018-06-14 10:24:56 +01:00
Simon Butcher 73a4b80475 Merge remote-tracking branch 'public/pr/1672' into development 2018-06-14 10:24:02 +01:00
Simon Butcher 9a8db56959 Merge remote-tracking branch 'public/pr/1730' into development 2018-06-14 10:23:29 +01:00
Simon Butcher 2ff0e52087 Fix missing preprocessor condition in AES self-test
The AES OFB self-test made use of a variable `offset` but failed to have a
preprocessor condition around it, so unless CTR and CBC were enabled, the
variable would be undeclared.
2018-06-14 09:57:07 +01:00
Philippe Antoine c03059db42 Simplify code in mbedtls_x509_csr_parse 2018-06-14 07:35:11 +02:00
Gilles Peskine 0a8352b4c2 Fix harmless use of uninitialized memory in ssl_parse_encrypted_pms
In ssl_parse_encrypted_pms, some operational failures from
ssl_decrypt_encrypted_pms lead to diff being set to a value that
depended on some uninitialized unsigned char and size_t values. This didn't
affect the behavior of the program (assuming an implementation with no
trap values for size_t) because all that matters is whether diff is 0,
but Valgrind rightfully complained about the use of uninitialized
memory. Behave nicely and initialize the offending memory.
2018-06-13 18:21:25 +02:00
Gilles Peskine 725f1cb6bd SSL async tests: add a few test cases for error in decrypt
The code paths in the library are different for decryption and for
signature. Improve the test coverage by doing some error path tests
for decryption in addition to signature.
2018-06-13 18:10:41 +02:00
Gilles Peskine 3f3ada8839 Fix memory leak in ssl_server2 with SNI + async callback
In ssl_server2, the private key objects are normally local variables
of the main function. However this does not hold for private keys in
the SNI configuration. When async callbacks are used, the test code
transfers the ownership of the private keys to the async callbacks.
Therefore the test code must free the SNI private keys through the
async callbacks (but it must not free the straight private keys this
way since they are not even heap-allocated).
2018-06-13 18:09:28 +02:00
Gilles Peskine 157f6d8f74 SNI + SSL async callback: make all keys async
When testing async callbacks with SNI, make all the keys async, not
just the first one. Otherwise the test is fragile with respect to
whether a key is used directly or through the async callbacks.
2018-06-13 18:06:51 +02:00
Jaeden Amero d906b818f3 tests: cipher: Allocate enough space for XTS keys
XTS keys can be double the size, since XTS uses two keys concatenated
together as a key (one for the tweak, one for encryption).
2018-06-13 12:16:25 +01:00