On 07/12/18 11:14 AM, Vakul Garg wrote:
Hi Boris
Thanks for explaining.
Few questions/observations.
1. Isn't ' ctx->decrypted = true' a redundant statement in tls_do_decryption()?
The same has been repeated in tls_recvmsg() after calling decrypt_skb()?
2. Similarly, ctx->saved_data_ready(sk) seems not required in tls_do_decryption().
This is because tls_do_decryption() is already triggered from tls_recvmsg() i.e. from user space app context.
3. In tls_queue(), I think strp->sk->sk_state_change() needs to be replaced with ctx->saved_data_ready().
Yes, I think these 3 can all be changed. #2 would be required if
do_decryption ever is called not in user context, but that's not the
case currently.