On Wed, 20 May 2026 22:58:39 -0400 Christopher Lusk wrote: > When the BPF sk_msg verdict sets apply_bytes smaller than the current > open record, tls_push_record() splits ctx->open_rec into the record > being encrypted and a remainder record. The synchronous path reattaches > the remainder to ctx->open_rec before continuing.
The current understanding is that this code has no real users. So let's try clear async_capable if BPF is attached and avoid all these bugs in record handling, please? The savings from zero copy are negligible.
