Re: [PATCH net-next v3 1/2] tls: annotate lockless access to sk->sk_err

[quanyeyang]

On Sunday, October 11th, 2026 at AM 8:10, netdev-bot+sashiko@xxxxxxxxxx <netdev-bot+sashiko@xxxxxxxxxx> wrote:

> Can the EBADMSG test in bpf_exec_tx_verdict() miss a synchronous
> encryption failure?

That window is pre-existing. tls_push_record() still reports the
crypto failure through sk_err and returns the raw crypto rc; the
caller peeks sk_err to decide cleanup. Folding the two unmarked
loads into one READ_ONCE() does not close a steal or overwrite of
that slot. Routing the failure back through the return value would
be a separate change, not this annotation.

> Should this patch carry Fixes: tags so the fix reaches stable?

No. This series is a net-next annotation of lockless sk_err access.
The consume-once change in tls_rx_rec_wait() matches what we already
did for TCP/MPTCP; it is not being backported here.