feat: add sbom Makefile target

[MarkAtwood]

Important

Blocked on wolfSSL/wolfssl#10343 — scripts/gen-sbom has not merged to wolfssl master yet. Do not merge this PR before #10343 lands. Until then the target fails with a clear error pointing at the missing script.

Summary

Adds make sbom, make install-sbom and make uninstall-sbom targets that produce CycloneDX and SPDX SBOM files for wolfHSM, at parity with the canonical scripts/sbom.am fragment and wolfTPM's CMake target.

• Parses version from ChangeLog.md (e.g., # wolfHSM Release v1.4.0)
• Enumerates sources from src/*.c (sorted for reproducible output)
• Artifact hashing is done by gen-sbom --srcs (OmniBOR gitoid over the source list). --srcs was chosen over --lib deliberately: wolfHSM ships as source into embedded builds, so there is no canonical installed library to hash
• Records the effective build configuration by preprocessing wolfhsm/wh_settings.h with the same defines/include path as the test build (WOLFHSM_CFG_DIR selects the config; point it at your port's wolfhsm_cfg.h/user_settings.h)
• Lists wolfssl as a dependency component, version read from WOLFSSL_DIR; SBOM_DEP_WOLFSSL=no for WOLFHSM_CFG_NO_CRYPTO builds
• Validates the SPDX output with pyspdxtools and converts it to tag-value; SBOM_VALIDATE=no skips both
• Honors SOURCE_DATE_EPOCH (defaults to last git commit time) — repeated runs are byte-identical
• install-sbom places files under $(PREFIX)/share/doc/wolfhsm (opt-in; plain make/make install never touches SBOM files)
• No build system changes — pure Makefile target

Usage

make sbom WOLFSSL_DIR=/path/to/wolfssl/source
make install-sbom WOLFSSL_DIR=... [PREFIX=/usr/local] [DESTDIR=...]

Outputs: wolfhsm-<version>.cdx.json, wolfhsm-<version>.spdx.json, wolfhsm-<version>.spdx (tag-value)

Requirements

• WOLFSSL_DIR must point to a wolfssl source tree containing scripts/gen-sbom (feat: SBOM generation and OmniBOR build provenance (CRA compliance) wolfssl#10343, pending a future wolfssl release)
• python3, plus pip install spdx-tools unless SBOM_VALIDATE=no

Test plan

• make sbom WOLFSSL_DIR=... — CDX/SPDX/tag-value produced, correct version and source count
• Config macros recorded (33 WOLFHSM_CFG_*/wolfSSL options with the default test config)
• wolfssl dependency component present with correct version; absent with SBOM_DEP_WOLFSSL=no
• Reproducibility: two consecutive runs byte-identical
• SBOM_VALIDATE=no skips validation/tag-value; missing pyspdxtools fails with a clear error otherwise
• install-sbom/uninstall-sbom round-trip under DESTDIR
• make -n clean does not shell out (GNU make 3.81 parse regression check)

[Copilot]

Pull request overview

Note

Copilot couldn't run its full agentic review because no GitHub Actions runner was available. Make sure your repository has a runner available to run Copilot's review, or add a copilot-setup-steps.yml file specifying one with the runs-on attribute. See the docs for more details.

Adds a make sbom target to generate CycloneDX and SPDX SBOM artifacts for wolfHSM using wolfSSL’s scripts/gen-sbom.

Changes:

• Introduces an sbom Makefile target with checks for version parsing and WOLFSSL_DIR.
• Gathers sources from src/*.c and passes compiler defines + metadata into gen-sbom.
• Emits versioned wolfhsm-<version>.cdx.json and wolfhsm-<version>.spdx.json outputs.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[bigbrett]

Running make sbom WOLFSSL_DIR=../wolfssl against the latest wolfSSL master I get the following error:

ERROR: ../wolfssl/scripts/gen-sbom not found.                                                                                                                                                 
       Use a wolfSSL tree that includes SBOM support.                                                                                                                                         
make: *** [Makefile:73: sbom] Error 1                       

Did this not merge into wolfSSL master yet? If so it shouldn't merge into wolfHSM.

[dgarske]

#414 (review)

#414 (review)

[jackctj117]

Skoll Multi-Scan Review

Modes: review + review-securityOverall recommendation: COMMENT
Findings: 5 total — 5 posted, 0 skipped
5 finding(s) posted as inline comments (see file-level comments below)

Posted findings

• [Medium] [review] Default WOLFSSL_DIR is off by one level for the top-level Makefile — Makefile:62
• [Medium] [review+review-security] gen-sbom --options-h receives only compiler builtin macros, not wolfHSM/wolfSSL config — Makefile:87-99
• [Low] [review+review-security] VERSION shell-out runs on every make invocation, not just sbom — Makefile:63
• [Low] [review+review-security] Temp file uses hardcoded /tmp prefix, ignoring $TMPDIR — Makefile:87
• [Low] [review+review-security] README manual invocation diverges from the Makefile sbom target — README.md:37-44

Review generated by Skoll

[MarkAtwood]

@bigbrett Correct — gen-sbom hasn't merged to wolfssl master. It's in wolfssl PR wolfSSL/wolfssl#10343, which is still open, so no wolfssl release or master checkout has scripts/gen-sbom yet. The target can't create that file; what it can do is fail clearly. I updated the error to name the dependency and point you at setting WOLFSSL_DIR to a tree that has the script, and the README says the same. So the wolfHSM side is ready to use the moment #10343 lands; it just no-ops with a clear message until then. I did not vendor a copy of gen-sbom into wolfHSM. If you'd rather hold this PR until #10343 merges, flag it and I'll mark it blocked.

[MarkAtwood]

Force-pushed a cleanup: this branch was cut from a stale shallow base, which made ~7,800 lines of already-merged main content (armv8m-tz, hwkeystore, test files) appear as PR additions. Rebased onto current main; the PR now contains only the SBOM change: +133 in Makefile and README.

Verified on the rebased head: make sbom WOLFSSL_DIR=<wolfssl> WOLFHSM_CFG_DIR=test/config generates both artifacts (wolfhsm-1.4.0.cdx.json, wolfhsm-1.4.0.spdx.json) against wolfSSL PR #10343's gen-sbom.

@bigbrett @dgarske ready for re-review; your earlier comments on the unrelated files are moot.

[MarkAtwood]

Resolved the July 10 review threads: every fix discussed in them is carried in the rebuilt branch (d03b224 contains the cumulative Makefile/README state, verified end-to-end with make sbom). The conversation history references pre-rebuild commit ids; the content is all here.

[sameehj]

@MarkAtwood

direction is right, but the target should reuse gen-sbom the way the rest of the family does

Thanks Mark. The overall approach — a dedicated make sbom target that calls wolfSSL's gen-sbom — is the correct call for wolfHSM, since wolfHSM is a hand-written Makefile and can't include scripts/sbom.am like the autotools repos (wolfMQTT, wolfProvider, wolfCLU, wolfEngine, wolfssh, wolfscep). We already have a precedent for a non-automake reimplementation: wolfTPM's CMake sbom target, which was explicitly brought to parity with the canonical scripts/sbom.am. That's the bar this PR should hit.

Right now #414 diverges from the canonical generator in a few ways, which is what the review threads are circling:

It rolls its own source hash (SHA-256 fingerprint of src/*.c) instead of using gen-sbom's own input. gen-sbom already supports --lib (hash the built library from a staged install) and --srcs/--srcs-file (OmniBOR gitoid for embedded/source builds). Using a private hash means the wolfHSM SBOM checksum isn't comparable to the other products and doesn't describe the shipped artifact. Please pick --lib or --srcs deliberately and note which in the PR.

--options-h only receives compiler builtins, not wolfHSM/wolfSSL config macros (jackctj117's finding). The point of --options-h is to record the configured feature macros, so this needs the real CC -dM -E dump with wolfHSM's actual CPPFLAGS/CFLAGS + config header.

Missing parity items the fragment/CMake target have: install-sbom / uninstall-sbom, SOURCE_DATE_EPOCH for reproducible output, and the SPDX validation / tag-value step.

Sequencing: as bigbrett and dgarske noted, gen-sbom is still on wolfSSL PR #10343 and hasn't merged to master. The clear-error no-op behavior is fine, but this PR shouldn't merge before #10343 lands — let's mark it blocked-on that.

Net: keep the bespoke-Makefile approach, but rework it to parity with wolfTPM's target (drive everything through gen-sbom, capture real macros, add the install/reproducibility bits), and gate the merge on #10343. Happy to pair on the diff.

[MarkAtwood]

@sameehj parity items are in as of 073f4eb:

• SPDX validation + tag-value: pyspdxtools validates the SPDX output and emits wolfhsm-<ver>.spdx. SBOM_VALIDATE=no skips both (same gate name #10343 is growing for the Gentoo case).
• SOURCE_DATE_EPOCH: honored, defaults to last git commit time. Verified two consecutive runs produce byte-identical output.
• install-sbom / uninstall-sbom: opt-in, $(PREFIX)/share/doc/wolfhsm (PREFIX=/usr/local default, DESTDIR honored). No uninstall-hook chaining because the root Makefile has no install/uninstall targets to hook.
• Dependency component: the SBOM now lists wolfssl with the version read from WOLFSSL_DIR (--dep-wolfssl); SBOM_DEP_WOLFSSL=no for WOLFHSM_CFG_NO_CRYPTO builds.

On the artifact hash: the July 14 rebuild already dropped the private fingerprint — hashing goes through gen-sbom --srcs (sorted src/*.c, OmniBOR gitoid). That over --lib is deliberate: wolfHSM ships as source into embedded builds, so there is no canonical installed library to hash. Now stated in the PR description.

Config macros: also in since the rebuild — the recipe preprocesses wolfhsm/wh_settings.h with the same defines/include path as the test build, so the SBOM records effective WOLFHSM_CFG_* + wolfSSL options (33 config macros with the default test config), not compiler builtins.

Sequencing: agreed. Marked blocked on wolfSSL/wolfssl#10343 at the top of the description — @bigbrett that also answers your question: gen-sbom is still on that PR, this one must not merge before it.

[sameehj]

Update: migrate to vendored wolfGlass driver

Replaces the hand-rolled gen-sbom Makefile recipe with vendored tools/sbom from wolfGlass 9bdf5b7.

Critical fix vs earlier draft

Config capture uses SBOM_OPTIONS_H (preprocess wh_settings.h with -I/-include), not SBOM_CFLAGS. The driver's --cflags path drops -I/-include and previously produced a 2-property empty SBOM.

Also in this update

• WOLFHSM_CFG_DIR is required (no silent test/config default)
• Core-only CI canary with --min-properties 50 and --require-dep-version wolfssl
• Docs moved to docs/SBOM.md; follow-ups listed in docs/SBOM-FOLLOWUPS.md
• .gitignore un-ignores tools/sbom/build/ (bare Build/ matched it on macOS)

Local smoke: 1793 properties, wolfssl dep version 5.9.2.

[MarkAtwood]

gen-sbom is on wolfssl master now (#10343), so make sbom WOLFSSL_DIR=../wolfssl WOLFHSM_CFG_DIR=<config> works against master. The error you hit is gone.

Both failing checks are fixed.

scan-build: the WOLFHSM_CFG_DIR guard was a plain $(error) at file scope. Make evaluates that when it reads the Makefile, so it killed every target, not just the sbom ones. make scan died at Makefile:80. It now only fires when an sbom goal is actually requested.

sbom_canary: gcc and clang gave different SBOMs. -dM dumps the compiler's own builtins. gen-sbom drops __-prefixed names but keeps system macros whose values point at them, so ATOMIC_BOOL_LOCK_FREE expands to __GCC_ATOMIC_BOOL_LOCK_FREE on gcc and __CLANG_ATOMIC_BOOL_LOCK_FREE on clang. Function-like macro bodies differ too (idx ##VAR vs idx##VAR). The capture now keeps only object-like project config macros: 230 instead of 2138, identical under both compilers. All the WOLFHSM_CFG_, WOLFSSL_, HAVE_* and NO_* macros are still recorded. What went away is glibc stuff like BUFSIZ.

The filter is a prefix allowlist, which will rot. I tried it without one (all-caps, object-like, skip __) and it still differed by 34 lines, so shape alone isn't enough to tell config from system macros. Follow-up #5 in docs/SBOM-FOLLOWUPS.md is to read the macro list from the settings headers instead, probably in the shared driver.

README: cut to one bullet in the resources list, as you asked. docs/SBOM.md already covers the WOLFHSM_CFG_DIR requirement.

048c405 is green.