Trond Myklebust wrote: > On Mon, 2007-08-06 at 11:57 -0400, Chuck Lever wrote: >> Finalize support for setting up RPC client transports to remote RPC >> services addressed via IPv6. >> >> Based on work done by Gilles Quillard at Bull Open Source. >> >> Signed-off-by: Chuck Lever >> Cc: Aurelien Charbon >> --- >> >> net/sunrpc/xprtsock.c | 57 ++++++++++++++++++++++++++++++++++++++++--------- >> 1 files changed, 47 insertions(+), 10 deletions(-) >> >> diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c >> index 391b88d..fb55c1c 100644 >> --- a/net/sunrpc/xprtsock.c >> +++ b/net/sunrpc/xprtsock.c >> @@ -13,6 +13,9 @@ >> * (C) 1999 Trond Myklebust >> * >> * IP socket transport implementation, (C) 2005 Chuck Lever >> + * >> + * IPv6 support contributed by Gilles Quillard, Bull Open Source, 2005. >> + * >> */ >> >> #include >> @@ -1790,6 +1793,7 @@ static struct rpc_xprt *xs_setup_xprt(struct rpc_xprtsock_create *args, unsigned >> */ >> struct rpc_xprt *xs_setup_udp(struct rpc_xprtsock_create *args) >> { >> + struct sockaddr *addr = args->dstaddr; >> struct rpc_xprt *xprt; >> struct sock_xprt *transport; >> >> @@ -1798,15 +1802,11 @@ struct rpc_xprt *xs_setup_udp(struct rpc_xprtsock_create *args) >> return xprt; >> transport = container_of(xprt, struct sock_xprt, xprt); >> >> - if (ntohs(((struct sockaddr_in *)args->dstaddr)->sin_port) != 0) >> - xprt_set_bound(xprt); >> - >> xprt->prot = IPPROTO_UDP; >> xprt->tsh_size = 0; >> /* XXX: header size can vary due to auth type, IPv6, etc. */ >> xprt->max_payload = (1U << 16) - (MAX_HEADER << 3); >> >> - INIT_DELAYED_WORK(&transport->connect_worker, xs_udp_connect_worker4); >> xprt->bind_timeout = XS_BIND_TO; >> xprt->connect_timeout = XS_UDP_CONN_TO; >> xprt->reestablish_timeout = XS_UDP_REEST_TO; >> @@ -1819,7 +1819,28 @@ struct rpc_xprt *xs_setup_udp(struct rpc_xprtsock_create *args) >> else >> xprt_set_timeout(&xprt->timeout, 5, 5 * HZ); >> >> - xs_format_ipv4_peer_addresses(xprt); >> + switch (addr->sa_family) { >> + case AF_INET: >> + if (ntohs(((struct sockaddr_in *)addr)->sin_port) != 0) > > Hmm.... I'm converting these to compare sin_port to htons(0). There is > no need to byte-swap the variable every time we run this code if we can > just have the compiler byte-swap the constant. Not a hot path... but yes, that is true. Up to you if you want to change this as described, but my feeling is that would be a little harder to read. Your call, of course.