> > > > @@ -859,6 +861,7 @@ static int __init tls_register(void)
> > > >
> > > > tls_sw_proto_ops = inet_stream_ops;
> > > > tls_sw_proto_ops.splice_read = tls_sw_splice_read;
> > > > + tls_sw_proto_ops.sendpage_locked = tls_sw_sendpage_locked,
> > > >
> > > > and additionally allowing MSG_NO_SHARED_FRAGS:
> > > >
> > > > int tls_sw_sendpage_locked(struct sock *sk, struct page *page,
> > > > int offset, size_t size, int flags)
> > > > {
> > > > if (flags & ~(MSG_MORE | MSG_DONTWAIT | MSG_NOSIGNAL |
> > > > - MSG_SENDPAGE_NOTLAST | MSG_SENDPAGE_NOPOLICY))
> > > > + MSG_SENDPAGE_NOTLAST |
> > > > MSG_SENDPAGE_NOPOLICY | MSG_NO_SHARED_FRAGS))
> > > > return -ENOTSUPP;
> > > >
> > >
> > > If you had added MSG_NO_SHARED_FRAGS to the existing tls_sw_sendpage
> > > would that have been sufficient?
> >
> > No, the stack trace I observed is
> >
> > tcp_bpf_sendmsg_redir
> > tcp_bpf_push_locked
> > tcp_bpf_push
> > kernel_sendpage_locked
> > sock->ops->sendpage_locked
> >
> > which never tries tls_sw_sendpage. Perhaps the relevant part is the
> > following in tcp_bpf_push?
> >
> > if (has_tx_ulp) {
> > flags |= MSG_SENDPAGE_NOPOLICY;
> > ret = kernel_sendpage_locked(sk,
> > page, off, size, flags);
> > } else {
> > ret = do_tcp_sendpages(sk, page, off, size, flags);
> > }
> >
>
> Got it, want to submit a fix? Or I can this is a bug.
>
> > Do let me know if there's anything I can help out with. Thanks for
> > your quick answer!
>
> Can you send out a fix for above sendpage_locked case?
Done:
net/tls: enable sk_msg redirect to tls socket egress
http://patchwork.ozlabs.org/patch/1196825/
It addresses both issues at the same time. This seemed preferable than
two separate patches. MSG_NO_SHARED_FRAGS precedes commit 0608c69c9a80
("bpf: sk_msg, sock{map|hash} redirect through ULP"), so no additional
Fixes tag for that.
