[Ricardo Casallas (rcasallas-silabs)]

Summary

Upgrade mbedTLS support from 3.x to 4.x:

• Build paths updated
• Crypto function updated

Related issues

• MATTER-6297.

Testing

• Lighting-app built and commission with MG24, MG26, Six301 and Si907

[cursor]

PR Summary

High Risk
Touches commissioning crypto, DAC provisioning, certificate parsing, OTA decryption, and OpenThread link-layer security across EFR32 and SiWx917.

Overview
Upgrades Silicon Labs Matter builds to mbedTLS 4 and the split TF-PSA-Crypto layout: GN/SDK wiring now compiles crypto from mbedtls/mbedtls plus tf-psa-crypto, switches PSA config to TF_PSA_CRYPTO_CONFIG_FILE, and retargets Simplicity SDK paths (e.g. OpenThread under thread/ and openthread_forks, lwIP under network_stack).

Platform crypto is updated for mbedTLS 4 API breaks: legacy symbols move behind MBEDTLS_DECLARE_PRIVATE_IDENTIFIERS and mbedtls/private/*; HKDF is reimplemented where mbedtls_hkdf was removed; X.509/CSR public-key export uses PSA (mbedtls_pk_import_into_psa / mbedtls_pk_write_pubkey_psa); psa_open_key is replaced with attribute checks or direct destroy; provisioning and OTA paths use PSA keygen/signing and PSA AES-CTR instead of direct mbed APIs where required.

SiWx917 tinycrypt builds add siwx917_psa_tinycrypt_stubs.c so PK/X.509 can call a minimal PSA surface without the full TF-PSA core. Series 3 (simg301) adds simg301-clang.ld for the new flash/RAM/NVM layout. OpenThread on EFR32 picks up updated platform MAC/security sources from the SDK.

^{Reviewed by Cursor Bugbot for commit d812eaf. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit d812eaf. Configure here.}

[cursor]

OTA decrypt ignores cipher finish errors

High Severity

Decrypt never checks psa_cipher_finish and always returns success while advancing mIVOffset. PSA CTR can leave the last bytes in finish, and a failed finish must be aborted. Encrypted OTA can then apply a corrupt image or decrypt later chunks with a wrong counter. The same path also runs AES-CTR in place, unlike the working PSA decrypt helper that uses a separate output buffer.

 

^{Reviewed by Cursor Bugbot for commit d812eaf. Configure here.}