Re: git: d195b3783fa4 - main - sctp: fix socket type created by sctp_peeloff()

On Tue, Feb 03, 2026 at 08:24:10PM +0100, Michael Tuexen wrote:
M> > M>   VNET_ASSERT(head->so_vnet != NULL, ("%s:%d so_vnet is NULL, head=%p",
M> > M>      __func__, __LINE__, head));
M> > M> + KASSERT(head->so_type == SOCK_SEQPACKET,
M> > M> +    ("%s: unexpecte so_type: %d", __func__, head->so_type));
M> > M>   so = soalloc(head->so_vnet);
M> > M>   if (so == NULL) {
M> > M>   log(LOG_DEBUG, "%s: pcb %p: New socket allocation failure: "
M> > M> @@ -1311,7 +1313,7 @@ sopeeloff(struct socket *head)
M> > M>      __func__, head->so_pcb);
M> > M>   return (NULL);
M> > M>   }
M> > M> - so->so_type = head->so_type;
M> > M> + so->so_type = SOCK_STREAM;
M> > M>   so->so_options = head->so_options;
M> > M>   so->so_linger = head->so_linger;
M> > M>   so->so_state = (head->so_state & SS_NBIO) | SS_ISCONNECTED;
M> > 
M> > This creates a socket where:
M> > 
M> > so->so_type != so->so_proto->pr_type.
M> > 
M> > I'm not sure this is a good idea.  I was actually looking into removing so_type
M> > at all.  What does SCTP idea is about this peel-off thing?  If the resulting
M> > socket is a stream one, shouldn't its so_proto point at sctp_stream_protosw?
M> Yes, that makes sense.
M> But this is now a generic routine (I think it was an SCTP specific one in the past),
M> how can I refer to it without using sctp_stream_protosw?

It was never generic, albeit put into generic file.  It lives under #ifdef
SCTP. I'm all for moving it entirely into SCTP related file.

-- 
Gleb Smirnoff