The non-blocking setup for X25519 and ECC in TLS was unconditionally
setting up nbCtx, which caused functions to return FP_WOULDBLOCK. However,
with INVALID_DEVID (the default), TLS has no async loop to handle
FP_WOULDBLOCK, only WC_PENDING_E via the async framework.
The fix follows the pattern used in asn.c: only set up nbCtx when the async
device is active (devId != INVALID_DEVID). With INVALID_DEVID, the code now
uses the blocking fallback (WC_ECC_NONBLOCK_ONLY) instead.
This prevents unit test timeouts when built with --enable-curve25519=nonblock
or --enable-ecc=nonblock.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
## Summary
- Add non-blocking (incremental) Curve25519 key generation and shared secret via `WC_X25519_NONBLOCK`, modeled after the existing ECC non-blocking pattern (`WC_ECC_NONBLOCK`)
- Implement `curve25519_nb()` and `fe_inv__distinct_nb()` in `fe_low_mem.c` as state-machine variants that return `FP_WOULDBLOCK` to yield after each field multiply
- Add `wc_curve25519_set_nonblock()` API to attach/detach non-blocking context to a key
- Integrate X25519 non-blocking with TLS 1.2/1.3 key share generation and shared secret in `tls.c` and `internal.c` (behind `WC_X25519_NONBLOCK && WOLFSSL_ASYNC_CRYPT_SW`)
- Add `--enable-curve25519=nonblock` configure option (auto-enables `--enable-asynccrypt` and `--enable-asynccrypt-sw`)
- Add X25519 async software dispatch cases in `async.c` and types in `async.h`
- Fix async guard in `curve25519.c` to require `WOLFSSL_ASYNC_CRYPT_SW` (matching other algorithms)
- Overhaul `examples/async/` client/server: non-blocking I/O via `WOLFSSL_USER_IO`, standalone `Makefile`, X25519/ECC mode selection, CI-friendly ready-file sync
- Add `examples/configs/user_settings_curve25519nonblock.h` and CI coverage in `os-check.yml` and new `async-examples.yml` workflow
- Add wolfcrypt test and API test coverage for X25519 non-blocking
Mostly combinations of NO_WOLFSSL_CLIENT, NO_WOLFSSL_SERVER and
WOLFSSL_NO_CLIENT_AUTH were failing.
Added configurations to CI loop.
wc_AesGcmDecryptFinal: use WC_AES_BLOCK_SIZE to satisfy compiler.
* More PQC configurations
* More CMake setups
* Fix various bugs uncovered by these tests
Added some missing feature additions to CMake to make the example
`user_settings_all.` config file work for the CI test.
When both TLS 1.3 and Brainpool curves are enabled, three new groups can
be used for the ECDHE key exchange according to RFC 8734:
* WOLFSSL_ECC_BRAINPOOLP256R1TLS13 (31)
* WOLFSSL_ECC_BRAINPOOLP384R1TLS13 (32)
* WOLFSSL_ECC_BRAINPOOLP512R1TLS13 (33)
Also ensure that the existing TLS 1.2 curves are sent properly.
The TLS client application is updated to support handshakes via
Brainpool curves using the new argument "--bpKs".
Change TLS 1.3 handling to return missing_extension alert code when
- KeyShare is present but SupportedGroups is missing and
- SupportedGroups is present but KeyShare is missing
Added tests for this.
x509.c: realloc may fail and therefore need to store result in a
temporary so the old pointer is not lost.
tls.c: free the name if it is not pushed on to the stack of peer CA
names. Failure to push can be from memory allocation failure.
aes.c: Don't compile XTS decrypt functions without HAVE_AES_DECRYPT.
Fix tests to have better pre-processor protection.
Exposes dynamic TLS certificate loading and OCSP stapling to allow applications to load certs lazily.
The server no longer needs to load the CA to staple OCSP responses.
Adds a certificate setup callback (WOLFSSL_CERT_SETUP_CB)
Adds an OCSP status callback to load OCSP responses directly
Adds `wc_NewOCSP`, `wc_FreeOCSP`, and `wc_CheckCertOcspResponse`
Don't call verify twice on the same error
Send correct alert on status response error
Sean Parkinson
David Garske
Juliusz Sosinowicz
Kareem
Tobias Frauenschläger
Daniel Pouzzner
Ruby Martin
Eric Blankenhorn
Paul Adelsbach
JacobBarthelmeh
Anthony Hu
effbiae
Reda Chouk
Mattia Moffa
Josh Holtrop
Hideki Miyazaki