This fix looks good, just a few small comments:
2026-07-06, 13:34:01 +0200, Ralf Lici wrote:
> ovpn_tcp_init initializes the TCP proto and proto_ops templates used
> when a TCP socket is attached to an ovpn peer but ovpn_init registers
> the rtnl link ops and generic netlink family before initializing those
> templates. Once either interface is visible, userspace can create an
If only the rtnl part is registered, we can create a netdevice but not
attach a socket to it, so the ovpn_tcp_* ops should not be reachable.
> ovpn device and configure a TCP socket while the TCP templates are still
> zero-initialized.
>
> Initialize the TCP templates before publishing the rtnl and netlink
> interfaces, so externally reachable setup paths can only observe
> initialized TCP state.
>
> Fixes: 11851cbd60ea ("ovpn: implement TCP transport")
> Signed-off-by: Ralf Lici <[email protected]>
> ---
> New patch added in v5.
>
> drivers/net/ovpn/main.c | 10 +++++++---
> 1 file changed, 7 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/net/ovpn/main.c b/drivers/net/ovpn/main.c
> index 9993c1dfe471..c2c3af0c3c55 100644
> --- a/drivers/net/ovpn/main.c
> +++ b/drivers/net/ovpn/main.c
> @@ -233,8 +233,14 @@ static struct rtnl_link_ops ovpn_link_ops = {
>
> static int __init ovpn_init(void)
> {
> - int err = rtnl_link_register(&ovpn_link_ops);
> + int err;
>
> + /* initialize the TCP protos before the module is exposed through rtnl
> + * or netlink
> + */
I don't think that comment adds much. The commit message explains why
this needs to be moved (and why this had to be moved, if someone
touches this later)
> + ovpn_tcp_init();
> +
> + err = rtnl_link_register(&ovpn_link_ops);
--
Sabrina
_______________________________________________
Openvpn-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/openvpn-devel
