Closed suihkulokki closed 7 months ago
Hi @suihkulokki
I've put up a pull request that fixes the inline assembly code for ChaCha20: #7182
Let us know if this fixes your issue.
Also if you see any more of these issue, please let us know and we will fix them as well.
Thanks, Sean
-- Sean Parkinson, wolfSSL Senior Software Engineer
Hi @SparkiDev
Thank you for your quick response. I can verify with that patch I no longer get unaligned access errors when compiled with arm assembler.
Hi @suihkulokki
Glad we could fix this quickly for you! The pull request has been merged and the fix will go out in the next release.
Please raise a new issue if you have anything more like this.
Sean
@SparkiDev I have a question here aes-gcm-256 assembly code is not available on armv7,is this a compiling problem?
Hi @20083017,
When compiling for ARM, you need to enable the assembly: --enable-armasm. If the host is armv7l, the NEON code will be compiled and the AES-GCM assembly will be compiled in.
Most versions of ARM CPUs don't have support for SHA-512/SHA-3/SM3/SM4 hardware crypto instructions and therefore they are called out explicitly in the configuration output and required to be explicitly requested.
Sean
Contact Details
Version
5.6.6
Description
I am evaluating wolfssl for commercial use. Tests have been successful, but low performance. Attempts to enable ARM assembler for the target platform (ARMv7 cortex-a9 i.mx6. Less so. ARM optimized version crashes with an alignment exception from the kernel when running unit tests on target system:
Reproduction steps
Relevant log output