C2000 example: HWAES, entropy and probe

[dgarske]

Adds three toggles to embedded/ti-c2000-f28p55x, pairing with the wolfSSL PR: HWAES=1 routes AES through the accelerator and defines WC_USE_DEVID so wolfcrypt_test and benchmark actually exercise the device rather than silently measuring software, ENTROPY=1 swaps the dev-only seed for the real source with on-target validation KATs, and ENTROPY_PROBE=1 builds a measurement-only image that dumps unconditioned samples over SCI for host min-entropy analysis. Also corrects the include order so -I./Header precedes -I$(WOLFROOT), since wolfSSL's documented user_settings.h workflow puts a copy at the wolfSSL tree root that otherwise shadowed the example's own.

Requires: wolfSSL/wolfssl#11202

[Copilot]

Pull request overview

Adds build-time toggles and supporting harness/tools for the TI C2000 F28P55x example to (1) route AES through the on-chip AESA accelerator, (2) switch from the dev-only DRBG seed to a real oscillator-jitter entropy source with on-target checks, and (3) build a measurement-only “entropy probe” image plus host-side analysis to estimate min-entropy.

Changes:

• Introduces HWAES=1, ENTROPY=1, and ENTROPY_PROBE=1 toggles and associated build plumbing (including include-path order fixes).
• Adds on-target harness routines for hardware-AES vs software cross-checking and entropy-source validation, plus a standalone entropy probe firmware path.
• Adds a numpy-based host tool to analyze probe captures and updates the port README with usage and measurement results.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
embedded/ti-c2000-f28p55x/tools/entropy_analyze.py	New host-side analyzer for ENTROPY_PROBE captures (min-entropy/bias/acf/chi-square).
embedded/ti-c2000-f28p55x/Source/wolf_main.c	Adds entropy validation, AESA-vs-SW AES tests, and a probe-only execution path.
embedded/ti-c2000-f28p55x/Source/entropy_probe.c	New measurement-only firmware that emits tagged raw/packed entropy samples over SCI.
embedded/ti-c2000-f28p55x/README.md	Documents new toggles, hardware AES behavior/perf, and entropy/probe methodology.
embedded/ti-c2000-f28p55x/Makefile	Adds toggle wiring, avoids duplicate AES core sources, reorders include paths, and adds probe build.
embedded/ti-c2000-f28p55x/Header/user_settings.h	Enables crypto-callback + C2000 AES/entropy configuration under the new toggles.

Suppressed comments (1)

embedded/ti-c2000-f28p55x/tools/entropy_analyze.py:260

• main() reads capture files via open(...).read() without closing the file handle. Using a context manager avoids leaking descriptors in long-running / repeated invocations (e.g., scripted batch analysis).

    path = sys.argv[1] if len(sys.argv) > 1 else "-"
    if path == "--selftest":
        selftest()
        return
    text = sys.stdin.read() if path == "-" else open(path, errors="replace").read()

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated no new comments.

Suppressed comments (5)

embedded/ti-c2000-f28p55x/Source/entropy_probe.c:145

• probe_adcSample() also uses a bounded wait but unconditionally reads the ADC result even if the interrupt never arrives. That can return a stale conversion and skew both the raw dump and the packed LSB stream. Consider checking for timeout and handling it explicitly (e.g., return a sentinel, increment an error counter, or print an error tag and stop).

    ADC_clearInterruptStatus(ADCC_BASE, ADC_INT_NUMBER1);
    ADC_forceSOC(ADCC_BASE, ADC_SOC_NUMBER0);

    for (guard = 0; guard < 1000000UL; guard++) {
        if (ADC_getInterruptStatus(ADCC_BASE, ADC_INT_NUMBER1)) {
            break;
        }
    }

    return ADC_readResult(ADCCRESULT_BASE, ADC_SOC_NUMBER0);
}

embedded/ti-c2000-f28p55x/Source/wolf_main.c:64

• random.h is already included under other feature toggles in this file; adding another conditional include here increases duplication and the chance of diverging include guards. Consider including <wolfssl/wolfcrypt/random.h> once in a common section (or centralizing it behind a single feature check) so it’s not repeated across multiple #ifdef blocks.

#ifdef WOLF_ENTROPY
#include <wolfssl/wolfcrypt/port/ti/ti-c2000-entropy.h>
#include <wolfssl/wolfcrypt/random.h>
#endif

embedded/ti-c2000-f28p55x/Source/entropy_probe.c:97

• probe_dccSample() uses a bounded wait loop, but it doesn’t check whether it actually observed DONE/ERROR before returning a value. If the mux is misconfigured or the peripheral stalls, the function will still return a counter-derived value (often 0) and silently pollute the capture. Consider detecting timeout (guard exhausted) and/or error status and emitting an explicit marker or aborting the probe so the host analysis can trust the samples.

This issue also appears on line 135 of the same file.

    /* Bounded wait, scaled to the window, so a bad mux cannot hang. */
    for (guard = 0; guard < (window * 256UL) + 100000UL; guard++) {
        if (DCC_getSingleShotStatus(base) || DCC_getErrorStatus(base)) {
            break;
        }

embedded/ti-c2000-f28p55x/Header/user_settings.h:572

• The section header comment says the RNG is “seeded by a DEV-ONLY test seed”, but this block is now conditional and can enable a real entropy source when WOLF_ENTROPY is set. Update the banner text to reflect the two modes (real entropy vs WOLFSSL_GENSEED_FORTEST) so it doesn’t contradict the code below.

/* ------------------------------------------------------------------------- */
/* RNG - real SHA-256 Hash-DRBG seeded by a DEV-ONLY test seed                */
/* ------------------------------------------------------------------------- */
#ifdef WOLF_ENTROPY

embedded/ti-c2000-f28p55x/Source/wolf_main.c:1144

• If wc_c2000_Entropy_GetRaw() fails, the code still counts ones over whatever happens to be in raw (static buffer), and prints that count alongside the failure. That makes the diagnostic output misleading (e.g., showing 0/2048 ones for a read error). Consider checking ret first and printing a distinct “GetRaw failed (ret=…)” line (or zeroing raw) before doing the popcount.

        ret = wc_c2000_Entropy_GetRaw(raw, (word32)sizeof(raw), src);
        ones = 0;
        for (i = 0; i < (word32)sizeof(raw); i++) {
            for (b = 0; b < 8; b++) {
                if ((raw[i] >> b) & 1) {

[Copilot]

🟡 Changes recommended

The new on-target entropy validation currently computes and reports the bit-balance result even when the raw-sample read fails, which can mask failures and produce misleading output.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Review details

• Files reviewed: 8/10 changed files
• Comments generated: 1
• Review effort level: Lite

[wolfSSL-Fenrir-bot]

Fenrir Automated Review — PR #617

Scan targets checked: wolfssl-examples-bugs, wolfssl-examples-src

Findings: 1
1 finding(s) posted as inline comments (see file-level comments below)

This review was generated automatically by Fenrir. Reported findings require changes before merge.

[wolfSSL-Fenrir-bot]

wolf_mldsa87_verify_test() is compiled but never called in SECUREBOOT builds · Dead/unreachable code

The compile guard excludes only WOLF_MLDSA_SIGN and WOLF_MLDSA_OCTETS, but main() dispatches WOLF_SECUREBOOT first (line 2480), so with SECUREBOOT=1 this function is defined and never referenced, carrying a second static wc_MlDsaKey (~14 KB) into .bss alongside sb_key.

Fix: Add && !defined(WOLF_SECUREBOOT) to the guard at line 887 and its matching #endif comment at line 973.