Is AES from wep_supplicant threadsafe?
Posted: Sun Nov 12, 2017 1:37 pm
Hello, everybody!
I am trying to use aes-ccm algorithm extending ported wpa_supplicant interface with files from original:
https://w1.fi/cgit/hostap/tree/wlantest
- ccmp.c
https://w1.fi/cgit/hostap/tree/src/crypto
- aes-ccm.h
- aes-ccm.c
I use release/v2.0 and I guess that it's wpa_supplicant port (probably for all releases) is made as interface to mbedtls ciphers implementation.
It looks like working, but very often authentication of messages fails. It does not depend on packet size, but looks like depends from calls rate: higher rate - more errors. Thus I suspect that it could be an issue with not thread-safe access to underlying AES processing layer. I don't have race condition at my calls level.
I also don't have any task of higher priority and this operation runs are executing from wifi promiscuous callback.
Where to look at to solve this?
I am trying to use aes-ccm algorithm extending ported wpa_supplicant interface with files from original:
https://w1.fi/cgit/hostap/tree/wlantest
- ccmp.c
https://w1.fi/cgit/hostap/tree/src/crypto
- aes-ccm.h
- aes-ccm.c
I use release/v2.0 and I guess that it's wpa_supplicant port (probably for all releases) is made as interface to mbedtls ciphers implementation.
It looks like working, but very often authentication of messages fails. It does not depend on packet size, but looks like depends from calls rate: higher rate - more errors. Thus I suspect that it could be an issue with not thread-safe access to underlying AES processing layer. I don't have race condition at my calls level.
I also don't have any task of higher priority and this operation runs are executing from wifi promiscuous callback.
Where to look at to solve this?
