fix: free encoded point on write failure in s2n_ecc_evp_write_params_point

[WesleyRosenblum]

Goal

Fix a small memory leak on an error path in s2n_ecc_evp_write_params_point.

Why

In the EVP_APIS_SUPPORTED code path, EVP_PKEY_get1_tls_encodedpoint allocates the encoded EC point, which is then written to the output stuffer. If s2n_stuffer_write_bytes failed, the POSIX_GUARD returned early before reaching the OPENSSL_free(encoded_point) call, leaking the allocation.

The write fails when the output stuffer is non-growable and lacks room for the point. This is a small (tens of bytes), pre-authentication, repeatable leak — DoS-only, no memory corruption.

How

Declare encoded_point with DEFER_CLEANUP(..., OPENSSL_free_pointer) so it is freed on every exit path (size mismatch, write failure, and success). The size check becomes POSIX_ENSURE and the write keeps its POSIX_GUARD, with no manual OPENSSL_free calls to forget. This matches the existing cleanup-function pattern already used in this file (EVP_PKEY_free, EC_KEY_free, etc.).

Callouts

• The #else (non-EVP) branch uses s2n_stuffer_raw_write and has no separate allocation to leak, so no change is needed there.

Testing

Added an error-path test to s2n_ecc_evp_test.c that, for every supported curve, allocates a non-growable stuffer one byte too small to hold the encoded point and asserts the write fails. This exercises the leak path; the leak is caught by valgrind/ASAN in CI. Existing success-path tests continue to pass.

Related

N/A

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

[maddeleine]

Suggested change

	EXPECT_FAILURE(s2n_ecc_evp_write_params_point(&test_params, &wire));
	EXPECT_ERROR_WITH_ERRNO(s2n_ecc_evp_write_params_point(&test_params, &wire), S2N_ERR_STUFFER_IS_FULL);

I think you should be able to assert on the exact error here. Pretty sure it should be stuffer is full error.