anatol opened a new pull request, #21646:
URL: https://github.com/apache/nuttx/pull/21646

   ## Summary
     
     `TD4` holds `uint8_t`, which is promoted to a signed `int`, so `<< 24` of 
a byte >= 0x80 shifts into the sign bit. UBSan flags this as 
`shift-out-of-bounds`. It reports on every rtl8188f WPA decrypt (201
     reports in `TestSanitizedImages/ubsan`).
     
     The shifted value is only used as part of the `uint32_t` result of the XOR 
with the round key, so the behavior is harmless in practice. This change adds 
`__attribute__((no_sanitize("shift")))` to
     `rijndaeldecrypt()` so UBSan no longer checks shifts in that function. It 
also adds a comment explaining why.
     
     I used the attribute because it is the smallest change and leaves the 
cipher code itself untouched. If maintainers prefer, I can instead cast the 
`TD4[...]` lookups to `uint32_t` before shifting.
     
     ## Impact
     
     - Users: none. Generated code is unchanged apart from the removed UBSan 
instrumentation in this one function.
     - Build process: none for non-sanitized builds. With 
`-fsanitize=undefined` the `shift` check is skipped for `rijndaeldecrypt()` 
only. All other checks stay enabled there.
     - Hardware / documentation / security / compatibility: no impact.
     
     ## Testing
     
     Host: Arch Linux x86_64, GCC 16.2
     Board: K230
     Build: with `-fsanitize=undefined` gcc flag
     
     Ran the UBSan-enabled image on the board and exercised the wireless WPA 
decrypt path.
     
     Before (without this change), UBSan reports on each decrypt:
     
    `UBSAN: shift-out-of-bounds in crypto/rijndael.c`
     
     After (with this change): no UBSan reports from `crypto/rijndael.c` on the 
same run.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to