Commit Graph

64 Commits

Author SHA1 Message Date
Jacob Barthelmeh
1b3a8f40b8 account for 8k keys with MySQL compatibility 2017-09-14 09:37:32 -06:00
dgarske
108f6a4958 Merge pull request #1077 from JacobBarthelmeh/StaticAnalysisTests
Static analysis tests
2017-08-02 15:08:32 -07:00
Jacob Barthelmeh
43b04af388 adjust sanity check in integer.c to avoid reading uninitialized values 2017-08-02 10:39:05 -06:00
Jacob Barthelmeh
062f3efa6d adjust integer.c for loop range 2017-08-02 10:29:34 -06:00
David Garske
8a11b78960 Fix for normal math mp_read_unsigned_bin where mp_zero was also resetting the raw source. This caused failures with QAT when using normal math. 2017-07-31 14:09:15 -07:00
David Garske
e8cf4b5ff0 Coverity fixes for TLS 1.3, async, small stack and normal math. 2017-05-09 09:13:21 -07:00
David Garske
c1640e8a3d Intel QuickAssist (QAT) support and async enhancements/fixes:
* Adds ./configure "--with-intelqa=../QAT1.6”, port files, memory management and README.md (see wolfcrypt/src/port/intel/).
* Added Intel QAT support for RSA public/private (CRT/non-CRT), AES CBC/GCM, ECDH/ECDSA, DH, DES3, SHA, SHA224, SHA256, SHA384, SHA512, MD5 and HMAC.
* wolfSSL async enabled all client and server: PKI, Encrypt/Decrypt, Hashing/HMAC and Certificate Sign/Verify.
* wolfSSL async support in functions: Encrypt, Decrypt, VerifyMAC, BuildMessage, ConfirmSignature, DoCertificate, ParseCertRelative, and MakeSignature.
* wolfCrypt test and benchmark async support added for all HW acceleration.
* wolfCrypt benchmark multi-threading support.
* Added QuickAssist memory overrides for XMALLOC, XFREE and XREALLOC. XREALLOC determines if existing pointer needs reallocated for NUMA.
* Refactor to make sure “heap” is available for async dev init.
* Added async support for all examples for connect, accept, read and write.
* Added new WC_BIGINT (in wolfmath.c) for async hardware support.
* Added async simulator tests for DES3 CBC, AES CBC/GCM.
* Added QAT standalone build for unit testing.
* Added int return code to SHA and MD5 functions.
* Refactor of the async stack variable handling, so async operations have generic args buffer area and cleanup function pointer.
* Combined duplicate code for async push/pop handling.
* Refactor internal.c to add AllocKey / FreeKey.
* Refactor of hash init/free in TLS to use InitHashes and FreeHashes.
* Refactor of the async event->context to use WOLF_EVENT_TYPE_ASYNC_WOLFSSL for WOLFSSL* and WOLF_EVENT_TYPE_ASYNC_WOLFCRYPT for WC_ASYNC_DEV*.
* Suppress error message for WC_PENDING_E.
* Implemented "wolfSSL_EVP_MD_CTX_init" to do memset.
* Cleanup of the openssl compat CTX sizes when async is enabled.
* Cleanup of AES, DES3, DH, SHA, MD5, DES3, DH, HMAC, MD5 for consistency and readability.
* Cleanup of the OPAQUE_LEN.
* Cleanup to use ENCRYPT_LEN instead of sizeof(ssl->arrays.preMasterSecret).
* Changed ssl->arrays.preMasterSecret to use XMALLOC (accelerates HW operations)
* Reduce verbosity with debug enabled for "GetMyVersion", "wolfSSL Using RSA OAEP padding" and "wolfSSL Using RSA PKCSV15 padding".
* Updated RSA un-padding error message so its different than one above it for better debugging.
* Added QAT async enables for each algorithm.
* Refactor of the async init to use _ex.
* Added WC_ASYNC_THRESH_NONE to allow bypass of the async thresholds for testing.
* Reformatted the benchmark results:
PKI: "RSA 2048 private HW 18522 ops took 1.003 sec, avg 0.054 ms, 18467.763 ops/sec"
Crypto/Hashing: SHA-256 SW 350 megs took 1.009 seconds, 346.946 MB/s Cycles per byte = 9.87
* Added min execution time for all benchmarks.
* Moved wc_*GetHash and wc_*RestorePos to appropriate files so use of isCopy flag is local.
* Fix for ECC sign status sometimes being invalid due to uninitialized ECC digest in benchmark.
* Added new DECLARE_VAR/FREE_VAR and DECLARE_ARRAY/FREE_ARRAY macros for helping setup test/benchmark variables to accelerate async.
* Added NO_SW_BENCH option to only run HW bench.
* Added support for PRNG to use hardware SHA256 if _wc devId provided.
* Fix to prevent curve tests from running against wrong curve sizes. Changed wc_ecc_set_curve to match on exact size.
* Added the wc_*GetHash calls to the wolfCrypt tests.
* Added async hardware start/stop to wolfSSL init/cleanup.
* Refactor to add wc_*Copy for hashing context (for async), which replaces wc_*RestorePos.
* Fixes for building with TI hashing (including: SHA224, missing new API’s and building with dummy build for non hw testing). Note: We need to add build test for this `./configure CFLAGS="-DWOLFSSL_TI_HASH -DTI_DUMMY_BUILD”`.
* Added arg checks on wc_*GetHash and wc_*Copy.
* Cleanup of the BuildMD5, BuildSHA, BuildMD5_CertVerify and BuildSHA_CertVerify functions.
* Added new ./configure --enable-asyncthreads, to allow enable/disable of the async threading support. If --enable-asynccrypt set this will be enabled by default if pthread is supported. Allows multi-threaded benchmarks with async simulator.
* Added checks for all hashing to verify valid ->buffLen.
* Fix for SHA512 scan-build warning about un-initialized “W_X”.
* Fix for valgrind un-initialized use of buffer in AllocDer (der->buffer) and BuildTlsFinished handshake_hash.
* Refactor of the benchmarking to use common function for start, check and finish of the stats.
* Fixed issue with ECC cache loading in multi-threading.
* Fix bug with AESNI not aligned code that assumes XMALLOC is 16-byte aligned.
* Added new WC_ASYNC_NO_… options to allow disabling of individual async algorithms. New defines are: WC_ASYNC_NO_CRYPT, WC_ASYNC_NO_PKI and WC_ASYNC_NO_HASH. Additionally each algorithm has a WC_ASYNC_NO_[ALGO] define.
* Added “wolfSSL_GetAllocators” API and fixed the wolfCrypt memcb_test so it restores callback pointers after test is complete (fixes issue with using custom allocators and test breaking it).
2017-04-10 14:45:05 -07:00
David Garske
141210dcc0 Fix warning with "implicit conversion loses integer precision”. 2017-03-16 14:56:03 -07:00
David Garske
2c890e6827 Fix mp_set_int to use calc for 32 const. Changed it to sizeof(b) instead of sizeof(long). 2017-03-15 12:34:55 -07:00
David Garske
5a24fd9237 Fix TFM mp_set_int to handle long. Enhance mp_set_int to use mp_set if less than max mp_digit. Added new MP_SET_CHUNK_BITS to eliminate hard coded const’s and allow build time adjustment. 2017-03-15 12:23:50 -07:00
Sean Parkinson
72728b21af Undo as mp_digit is not allowed to get as large as tested 2017-03-14 10:23:13 +10:00
Sean Parkinson
8ac2f5cb9c Windows warning about negating unsigned fix 2017-03-13 12:29:58 +10:00
Sean Parkinson
d4f0c79272 Cast for Windows 2017-03-13 12:18:45 +10:00
Sean Parkinson
614231f71c Fixes for extended configuration testing 2017-03-13 11:33:39 +10:00
toddouska
0e64c7708f Merge pull request #769 from JacobBarthelmeh/Testing
Testing
2017-03-09 10:17:48 -08:00
Jacob Barthelmeh
a0effa6329 call mp_clear to match call to mp_init 2017-03-08 11:26:16 -07:00
David Garske
be42a575da Fix additional integer.c report of possible use of NULL dp (after normal math performance improvement to defer dp pointer alloc commit bdbb98ed20 2017-03-06 13:19:52 -08:00
David Garske
431f363520 Better fixes for suppressing scan-build warning with normal math enabled. 2017-03-03 07:35:26 -08:00
David Garske
67a8626430 Fix for scan-build warning with “->dp == NULL”. Scenario can’t happen, but adding sanity check to suppress warning. 2017-03-02 15:56:31 -08:00
Jacob Barthelmeh
f77458992e resolve windows warnings and add sanity check with PKCS12 parse 2017-02-28 14:33:07 -07:00
David Garske
9db6a27921 Fixes for scan-build warnings. Fix possible memory leak in wolfSSL_DH_new on failure. Add null checks in integer.c for destination to make sure “dp” grows when NULL (even though never happens in real-use). Added suppression of wc_port.c warning “Value stored to 'ret' is never read”. 2017-02-23 14:47:36 -08:00
David Garske
4cbfec1c7d Implemented ksdk_port fixes to handle mp_ response codes. Added KSDK support for normal math. Regression testing against K82 hardware (MMCAU/LTC) and software with normal and fast math. 2017-02-21 14:03:21 -08:00
David Garske
3008c888bf Fix mp_cmp_d logic to handle a->used == 0. Revert mp_copy and mp_set 0 size workarounds. 2017-02-21 14:03:21 -08:00
David Garske
9c7407d18c Added return codes to wc_InitDhKey, wc_InitDsaKey and mp_set. Added missing return code checks on mp_copy in ecc.c. Fixed build with DSA and no ECC where mp_set function def would be missing. 2017-02-21 14:03:21 -08:00
David Garske
da5825b94d Normal math speed-up to not allocate on mp_int and defer until mp_grow. Added memory tracker support to ./tests/unit.test. Fix memory leak with curve cache enabled, by adding to wolfSSL_Cleanup. 2017-02-21 13:59:38 -08:00
David Garske
4f53761faf Fix naming for integer.c min/max local variables to resolve reported “error: declaration of 'min' shadows a global declaration”. 2017-02-09 15:52:25 -08:00
dgarske
ee53853d2f Merge pull request #632 from toddouska/init_multi
fix non ecc_make_key init_mulit potential problems
2016-11-14 19:34:14 -08:00
toddouska
1a7fe0d4c5 fix non ecc_make_key init_mulit potential problems 2016-11-14 12:49:42 -08:00
David Garske
6d5485b88f Fix to "mp_init_multi" so failure ensures a later "mp_clear" won't free on un-initialized pointer. Applies to !USE_FAST_MATH only. No measurable benchmark difference. 2016-11-11 20:03:58 -08:00
toddouska
c2b55f69fa fix 32bit mp_add_d need 2016-07-18 12:49:31 -07:00
David Garske
dd52af0872 ECC cleanup / fixes. Improvements to ECC with fast math enabled to avoid mp_clear on stack variables. Refactor of ECC failure cleanup (fixes possible mem leaks with small stack enabled). Refactor of "fp_is*" response checks to use FP_YES or FP_NO. Pulled libtom enhancement/cleanup of fp_isprime. Fix for compressed keys import with custom curves (still having some issues though). 2016-06-21 15:27:51 -07:00
David Garske
aa1a405dd1 Fixes for compressed keys. Fix to fast math "mp_cnt_lsb" to return proper value, which fixes "mp_jacobi", which fixes "mp_sqrtmod_prime", which fixes compressed keys for 224-bit key. Removed workarounds for compressed keys. Added new configure option "--enable-compkey". Fixed issue with normal math and custom curves where "t2" could be free'd and used. Fixed issue with mp_dump in integer.c, with not allocating correctly sized buffer for toradix. 2016-06-21 14:06:02 -07:00
David Garske
5703e5eadb ECC changes to support custom curves. Added new "WOLFSSL_CUSTOM_CURVES" option to support non-standard ECC curves in ecc_is_point and ecc_projective_dbl_point. Refactor to load and pass curve "a" parameter down through ECC functions. Relocated mp_submod and added mp_addmod. Refactor to pass mp variable directly (not pointer) for montgomery variable. Fix in mp_jacobi to also handle case of a == 0. Cleanup of *_ecc_mulmod and wc_ecc_make_key_ex error handling. Cleanup of ecc_map for handling normal, fast and alt_ecc math for optimization of performance and allowing reduced ecc_size. 2016-06-15 08:41:51 -07:00
David Garske
87c00eb3f3 Math updates from libtom. Fixes in mp_read_radix for to include char 36 and clear the destination before checks. Fix to clear tmp value on fp_sub_d and fp_add_d. Fixes in assembly for x86-32 INNERMUL, PPC32 INNERMUL and PROPCARRY and x86-64 SQRADD. Added mp_isneg. Refactor of mp_iszero, mp_iseven, mp_isodd and mp_isneg to check using MP_YES or MP_NO. Changed fp_read_unsigned_bin "b" arg to const. Removal of the "register" on stack variables to let compiler determine best optimization. 2016-06-14 16:56:22 -07:00
David Garske
2ad9d41641 Performance improvement for fast math mp_clear to use fp_zero (memset) instead of fp_clear(ForceZero). Added new mp_forcezero function for clearing/free'ing sensitive private key data. Changed ECC and RSA to use mp_forcezero to explicitly handle private key clearing. 2016-05-12 11:50:34 -07:00
David Garske
440956f8d4 Fixed new issue with ecc_projective_add_point not free'ing the local x, y, z if fast math was disabled. Formatting cleanup in integer.c. 2016-05-11 08:47:30 -07:00
David Garske
5cbc4bdf29 Added new "WOLFSSL_DEBUG_MATH", which enables use of "mp_dump" to display information about an mp_int. 2016-05-09 09:51:11 -07:00
Jacob Barthelmeh
e99a5b0483 prepare for release v3.9.0 2016-03-17 16:02:13 -06:00
John Safranek
b6017c59ba Merge pull request #275 from dgarske/WolfErrorTypo
Fixed spelling errors
2016-02-01 09:00:10 -08:00
David Garske
f8876854f4 Spelling fixes in comments and error strings (ALGO_ID_E, ASN_TIME_E and WOLFSSL_ERROR function). 2016-01-29 16:13:09 -08:00
toddouska
fa25e5d09d fix normal math off by one loop error in fast_s_mp_mul_high_digs 2016-01-29 13:54:41 -08:00
Chris Conlon
0801eeac8d update support for Freescale KSDK 1.3.0 2016-01-25 10:41:26 -07:00
kaleb-himes
38392ce56a safeguards to avoid de-referencing a null pointer 2016-01-06 10:12:52 -07:00
Chris Conlon
b153ac002c fix Visual Studio warnings 2015-12-21 16:11:02 -07:00
kaleb-himes
c3a249009f allow openssl extra tests if configuration supports 2015-11-10 15:29:05 -07:00
toddouska
ad51d4ba09 make sure fast invmod lowers result in too big case 2015-10-02 15:38:22 -07:00
Chris Conlon
0cd893a51b Freescale: Use new I/O where applicable 2015-08-07 16:22:31 -06:00
Chris Conlon
5d40c5f566 Rename RNG to WC_RNG for Freescale, add NO_OLD_RNGNAME define to completely remove RNG type usage 2015-08-07 11:53:19 -06:00
kaleb-himes
121a24ba15 update logic for ECC FASTMATH KEY-GEN and COMP-KEY 2015-08-04 15:11:01 -06:00
toddouska
0a975eaff9 fix valgrind warning mp_add_d 2015-07-24 15:34:56 -07:00