On 11/1/21 16:34, Leonard Crestez wrote:
> This is mainly intended to protect against local privilege escalations
> through a rarely used feature so it is deliberately not namespaced.
>
> Enforcement is only at the setsockopt level, this should be enough to
> ensure that the tcp_authopt_needed static key never turns on.
>
> No effort is made to handle disabling when the feature is already in
> use.
>
> Signed-off-by: Leonard Crestez <cdleonard@xxxxxxxxx>
> ---
[..]
> diff --git a/net/ipv4/tcp_authopt.c b/net/ipv4/tcp_authopt.c
> index 5e80e5e5e36e..7c49dcce7d24 100644
> --- a/net/ipv4/tcp_authopt.c
> +++ b/net/ipv4/tcp_authopt.c
> @@ -3,10 +3,15 @@
> #include <linux/kernel.h>
> #include <net/tcp.h>
> #include <net/tcp_authopt.h>
> #include <crypto/hash.h>
>
> +/* This is mainly intended to protect against local privilege escalations through
> + * a rarely used feature so it is deliberately not namespaced.
> + */
> +int sysctl_tcp_authopt;
Could you add pr_warn_once() for setsockopt() without this set, so that
it's visible in dmesg for a user that gets -EPERM.
Thanks,
Dmitry
